sergiotarxz
/
libvips
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity
libvips/libvips/foreign/dzsave.c

3203 lines
80 KiB
C
Raw Normal View History

started hacking in dzsave
2012-03-21 15:53:53 +01:00
/* save to deep zoom format
*
* 21/3/12
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
* - from the tiff pyramid writer
more dzsave overlap fixups works with odd-sized tiles, overlaps handled correctly
2012-07-07 14:58:50 +02:00
* 5/7/12 (thanks Alexander Koshman)
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
* - make tiles down to 1x1 pixels
more dzsave overlap fixups works with odd-sized tiles, overlaps handled correctly
2012-07-07 14:58:50 +02:00
* - oop make right-hand edge tiles
* - improve overlap handling
add threaded write of deepzoom tiles
2012-07-07 19:42:12 +02:00
* 7/7/12
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
* - threaded write
fixes to deep zoom write - shrink down to a 1x1 pixel tile, even for very long and thin images - round image size up on shrink - write a .dzi file with the pyramid params - default tile size and overlap now matches the openslide writer thanks to Benjamin Gilbert for pointing out the errors
2012-08-06 16:14:59 +02:00
* 6/8/12 (thanks to Benjamin Gilbert for pointing out the errors)
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
* - shrink down to a 1x1 pixel tile, even for very long and thin images
* - round image size up on shrink
* - write a .dzi file with the pyramid params
* - default tile size and overlap now matches the openslide writer
more improvements to dzsave now writes x_files and x.dzi, as per spec (thanks Benjamin) deprecate tile_width and tile_height, just have tile_size now
2012-08-07 14:29:03 +02:00
* 7/8/12 (thanks to Benjamin Gilbert again for more testing)
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
* - reorganise the directory structure
* - rename to basename and tile_size
* - deprecate tile_width/_height and dirname
sdzsave fix if dzsave was running a scanline at a time and height was odd, the final set of tiles were not being written -- thanks Martin
2012-10-01 15:08:59 +02:00
* 1/10/12
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
* - did not write low pyramid layers for images with an odd number of
* scan lines (thanks Martin)
fix .dzi format string dzsave left the save options attached to the Format string in the .dzi file also, version bump
2012-10-02 14:11:43 +02:00
* 2/10/12
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
* - remove filename options from format string in .dzi (thanks Martin)
dzsave can write zoomify and google maps use "layout" to pick the directory layout, deepzoom / zoomify / google all allowed
2012-10-03 15:26:59 +02:00
* 3/10/12
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
* - add zoomify and google maps output
add a @background option to embed and dzsave you can now set the background colour for edge tiles in dzsave and embed, eg.: vips dzsave k2.jpg x --background "1 2 3"
2012-10-10 14:22:30 +02:00
* 10/10/12
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
* - add @background option
add @depth option to dzsave so you can ask for a 1 layer pyramid
2012-11-01 19:11:46 +01:00
* 1/11/12
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
* - add @depth option
add --centre option to dzsave with --centre turned on, in google maps mode dzsave will centre tiles rather than placing them at the top-left also, make sure blank.png is set to @background
2013-01-21 15:48:14 +01:00
* 21/1/13
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
* - add @centre option
fix another dzsave corner case better padding calculation fixes yet another corner case for dzsave, thanks Martin
2013-02-26 13:32:01 +01:00
* 26/2/13
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
* - fix another corner case, thanks Martin
add --angle option to dzsave
2013-05-29 15:06:23 +02:00
* 29/5/13
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
* - add --angle option
faster --centre logic for dzsave don't test pixel-by-pixel for background tiles, instead track the rect of real pixels and test for intersection this might help performance scaling on many-core CPUs since we are no longer spinning up a threadgroup for each output tile thanks Kacey
2013-06-19 15:03:39 +02:00
* 19/6/13
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
* - faster --centre logic, thanks Kacey
dzsave will write a zip file but see the TODO in dzsave.c - no way to pick a tile format, always jpeg - no way to pick a container format, always zip - needs testing - examine memuse, how well do we work for large zips - can we make a zip over 4gb?
2014-04-19 16:15:25 +02:00
* 18/4/14
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
* - use libgsf for output so we can write to .zip etc. as well as the
* filesystem
set Type on memory strip we need to set Type on memory strips so we can convert to the output format correctly, thanks philipgiuliani see https://github.com/jcupitt/libvips/issues/120
2014-05-08 15:24:22 +02:00
* 8/5/14
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
* - set Type on strips so we can convert for save correctly, thanks
* philipgiuliani
save openslide metadata to .dzi files see https://github.com/jcupitt/libvips/issues/137 todo: - save to zoomify ImageProperties.xml as well? - any other useful fields?
2014-06-25 12:26:05 +02:00
* 25/6/14
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
* - stop on zip write >4gb, thanks bgilbert
* - save metadata, see https://github.com/libvips/libvips/issues/137
fix dzsave date attr on Windows see https://github.com/jcupitt/libvips/issues/157 thanks John
2014-08-18 22:30:26 +02:00
* 18/8/14
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
* - use g_ date funcs, helps Windows
use vips_region_shrink()
2015-02-14 14:48:08 +01:00
* 14/2/15
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
* - use vips_region_shrink()
use a better name for the dzsave temp dir see https://github.com/jcupitt/libvips/issues/242 dzsave with deepzoom layout and fs container writes a temp dir called dzsave-temp-324857 or similar, and makes sure that this directory does not exist after write, the foo_files and foo.dzi files are picked out to the current dir and the temp removed
2015-02-22 16:59:31 +01:00
* 22/2/15
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
* - use a better temp dir name for fs dz output
allow dzsave zip output > 4gb thanks benjamin
2015-06-08 10:35:25 +02:00
* 8/8/15
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
* - allow zip > 4gb if we have a recent libgsf
working, at least with bug case working, at least with robclouth test case see https://github.com/jcupitt/libvips/issues/326#issuecomment-138888981
2015-09-09 18:27:19 +02:00
* 9/9/15
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
* - better overlap handling, thanks robclouth
working! doesn't seem to slow us down significantly ... before: $ time vips dzsave CMU-1.svs x --layout google --background "243 243 243 255" real 1m1.940s user 2m15.004s sys 0m37.092s after: $ time vips dzsave CMU-1.svs x --layout google --background "243 243 243 255" strip_work: skipping blank tile 176 x 67 strip_work: skipping blank tile 21 x 112 real 1m3.503s user 2m16.012s sys 0m40.328s
2015-11-24 18:08:56 +01:00
* 24/11/15
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
* - don't write almost blank tiles in google mode
always strip dzsave tile metadata dzsave now uses :strip => true for all tiles. There's no point adding all the metadata to every tile, so hopefully this won't break anything. Thanks Benjamin, see: https://github.com/jcupitt/libvips/issues/349
2015-11-25 14:29:23 +01:00
* 25/11/15
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
* - always strip tile metadata
fix up dzsave overlap handling There was a mixup with the previous fix to dzsave overlap handling, correct it and update the test suite. In the previous revision, dzsave overlapped tiles by overlap and sized them by tile_size. In fact, tiles should be sized as (tile_size + overlap * 2), ie. tile_size refers to the number of unique pixels per tile. See https://github.com/jcupitt/libvips/issues/357
2015-12-17 12:54:38 +01:00
* 16/12/15
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
* - fix overlap handling again, thanks erdmann
remove is_zip from dzsave we can use GSF_IS_OUTFILE_ZIP() instead
2016-06-08 15:46:45 +02:00
* 8/6/16 Felix Bünemann
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
* - add @compression option
new overlap system for dzsave again ... sigh see https://github.com/jcupitt/libvips/issues/512
2016-09-05 19:17:30 +02:00
* 5/9/16
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
* - more overlap changes to help gmaps mode
notes on vips-properties changes
2016-09-08 13:45:50 +02:00
* 8/9/16 Felix Bünemann
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
* - move vips-properties out of subdir for gm and zoomify layouts
add dzsave_buffer zip only ... see https://github.com/jcupitt/libvips/issues/415
2016-10-15 20:01:34 +02:00
* 15/10/16
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
* - add dzsave_buffer
better >4gb detection for zip dzsave older libgsfs can't save zip64 and will fail silently for very large output trees ... improve the slightly sketchy >4gb detection in dzsave this a a version of the patch in https://github.com/jcupitt/libvips/pull/462
2016-11-11 17:24:18 +01:00
* 11/11/16 Felix Bünemann
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
* - better >4gb detection for zip output on older libgsfs
better dzsave with zip output to a file close down output earlier to help mark-sweep bindings
2017-08-18 13:31:24 +02:00
* 18/8/17
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
* - shut down the output earlier to flush zip output
dzsave outputs extra tiles along right and bottom the deepzoom spec wants tiles to be written along the right and bottom, even of those tiles contain no new pixels this patch seems to work, but needs testing see: https://github.com/jcupitt/libvips/issues/795
2017-11-24 16:54:22 +01:00
* 24/11/17
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
* - output overlap-only tiles on edges for better deepzoom spec
* compliance
write scan-properties.xml to szi see https://github.com/jcupitt/libvips/issues/853
2018-01-26 18:20:58 +01:00
* 6/1/18
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
* - add scan-properties.xml for szi output
* - write all associated images
add skip-blanks option to dzsave Skip-blanks was a feature of google layout. This patch makes it into an option you can control which defaults on in gm mode. See https://github.com/libvips/libvips/issues/1185
2018-12-19 15:52:33 +01:00
* 19/12/18
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
* - add @skip_blanks
add no_strip option to dzsave it's sometimes useful to have pyramid tiles with ICC profiles see https://github.com/libvips/libvips/issues/1451
2019-10-21 17:46:20 +02:00
* 21/10/19
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
* - add @no_strip
add iiif layout to dzsave
2019-11-09 21:40:39 +01:00
* 9/11/19
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
* - add IIIF layout
better iiif tile naming We were clipping edge tiles against the size of the scaled up layer. Instead, clip against the size of the top level image. See https://github.com/libvips/libvips/issues/1621
2020-04-24 20:29:35 +02:00
* 24/4/20 [IllyaMoskvin]
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
* - better IIIF tile naming
Add iiif3 (#2483) * add iiif3 support to dzsave some small things have changed from v2
2021-10-19 15:18:40 +02:00
* 15/10/21 martimpassos
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
* - add IIIF3 layout
dzsave.c: Remove trailing comma from IIIFv3 folder names (#2590) Fixes #2589. Signed-off-by: Aaron <admin@datahoarder.dev>
2021-12-21 10:43:48 +01:00
* 21/12/21 whalehub
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
* - remove trailing comma from IIIFv3 folder names
deprecate "properties", register ".szi" as a suff
2022-03-29 18:50:48 +02:00
* 29/3/22
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
* - always write a properties file
* - add .szi as a registered suffix
* 9/5/22
* - add dzsave_target
dzsave will write a zip file but see the TODO in dzsave.c - no way to pick a tile format, always jpeg - no way to pick a container format, always zip - needs testing - examine memuse, how well do we work for large zips - can we make a zip over 4gb?
2014-04-19 16:15:25 +02:00
*/
started hacking in dzsave
2012-03-21 15:53:53 +01:00
/*
This file is part of VIPS.
VIPS is free software; you can redistribute it and/or modify
it under the terms of the GNU Lesser General Public License as published by
the Free Software Foundation; either version 2 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU Lesser General Public License for more details.
You should have received a copy of the GNU Lesser General Public License
along with this program; if not, write to the Free Software
Update address for Free Software Foundation Also update LGPLv2.1 license text from the copy currently on the GNU website.
2013-03-07 06:40:19 +01:00
Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
02110-1301 USA
started hacking in dzsave
2012-03-21 15:53:53 +01:00
*/
/*
These files are distributed with VIPS - http://www.vips.ecs.soton.ac.uk
*/
hacking
2013-02-25 15:54:39 +01:00
/*
fix another dzsave corner case better padding calculation fixes yet another corner case for dzsave, thanks Martin
2013-02-26 13:32:01 +01:00
This is difficult to test, there are so many options.
hacking
2013-02-25 15:54:39 +01:00
fix another dzsave corner case better padding calculation fixes yet another corner case for dzsave, thanks Martin
2013-02-26 13:32:01 +01:00
It's failed in the past in these cases. These have layers with strips which
exactly align with image edges, or which have orphan scanlines which need
adding for the shrink.
hacking
2013-02-25 15:54:39 +01:00
fix another dzsave corner case better padding calculation fixes yet another corner case for dzsave, thanks Martin
2013-02-26 13:32:01 +01:00
1. $ header test.v
test.v: 14016x16448 uchar, 3 bands, srgb, openin VipsImage (0x11e7060)
$ time vips dzsave test.v x --overlap 0
hacking
2013-02-25 15:54:39 +01:00
fix another dzsave corner case better padding calculation fixes yet another corner case for dzsave, thanks Martin
2013-02-26 13:32:01 +01:00
Not all layers written.
2. $ header ~/Desktop/leicaimage.scn
/home/john/Desktop/leicaimage.scn: 4225x7905 uchar, 4 bands, rgb
Not all layers written.
3. $ header ~/leicatest1.scn
/home/john/leicatest1.scn: 11585x8449 uchar, 4 bands, rgb
Not all layers written.
various combinations of odd and even tile-size and overlap need testing too.
fix up dzsave overlap handling There was a mixup with the previous fix to dzsave overlap handling, correct it and update the test suite. In the previous revision, dzsave overlapped tiles by overlap and sized them by tile_size. In fact, tiles should be sized as (tile_size + overlap * 2), ie. tile_size refers to the number of unique pixels per tile. See https://github.com/jcupitt/libvips/issues/357
2015-12-17 12:54:38 +01:00
Overlap handling
Use 254 pixels as default tile size I remember @benvanik, a friend and colleague, who worked on our DZI renderer (Seadragon) told us that, I believe, it’s better for GPUs to have power of two sized tiles, so 254px + 2 x 1px tile overlap will give you 256 pixel.s > To keep my tiles nicely sized, I shrink in instead of expand out – that means my tiles are really 254×254 with a 1px border, so the images are 256x256px with 254×254 of useful imagery. If you wanted larger tiles, you’d go 510×510 with 1px border making 512×512 tiles. > — Source: http://www.noxa.org/blog/2009/11/29/megatextures-in-webgl-2/
2015-12-21 06:43:41 +01:00
For deepzoom, tile-size == 254 and overlap == 1 means that edge tiles are
255 x 255 (though less at the bottom right) and non-edge tiles are 256 x
256. Tiles are positioned across the image in tile-size steps. This means
fix up dzsave overlap handling There was a mixup with the previous fix to dzsave overlap handling, correct it and update the test suite. In the previous revision, dzsave overlapped tiles by overlap and sized them by tile_size. In fact, tiles should be sized as (tile_size + overlap * 2), ie. tile_size refers to the number of unique pixels per tile. See https://github.com/jcupitt/libvips/issues/357
2015-12-17 12:54:38 +01:00
(confusingly) that two adjoining tiles will have two pixels in common.
This has caused bugs in the past.
fix another dzsave corner case better padding calculation fixes yet another corner case for dzsave, thanks Martin
2013-02-26 13:32:01 +01:00
*/
hacking
2013-02-25 15:54:39 +01:00
started hacking in dzsave
2012-03-21 15:53:53 +01:00
/*
#define DEBUG_VERBOSE
more dzsave overlap fixups works with odd-sized tiles, overlaps handled correctly
2012-07-07 14:58:50 +02:00
#define VIPS_DEBUG
new overlap system for dzsave again ... sigh see https://github.com/jcupitt/libvips/issues/512
2016-09-05 19:17:30 +02:00
#define DEBUG
update TODO
2016-09-07 11:18:06 +02:00
*/
started hacking in dzsave
2012-03-21 15:53:53 +01:00
#ifdef HAVE_CONFIG_H
#include <config.h>
#endif /*HAVE_CONFIG_H*/
Cleanup gettext handling (#2695) Use GLib's i18n support instead of copying and pasting that logic into its own header. This deprecates the vips/intl.h header in favour of glib/gi18n.h.
2022-02-27 16:27:28 +01:00
#include <glib/gi18n-lib.h>
started hacking in dzsave
2012-03-21 15:53:53 +01:00
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
dzsave: add missing include directive for errno/EEXIST (#3025) Resolves: #3024.
2022-09-05 10:58:52 +02:00
#include <errno.h>
started hacking in dzsave
2012-03-21 15:53:53 +01:00
#include <vips/vips.h>
dzsave can write zoomify and google maps use "layout" to pick the directory layout, deepzoom / zoomify / google all allowed
2012-10-03 15:26:59 +02:00
#include <vips/internal.h>
make libgsf-1 an optional dependency dzsave vanishes if you don't have it, but you can still build vips
2014-05-09 14:13:18 +02:00
#ifdef HAVE_GSF
block deprecation warnings from libgsf with an uglu gcc progma
2020-11-29 15:41:19 +01:00
/* Disable deprecation warnings from gsf. There are loads, and still not
* patched as of 12/2020.
*/
#pragma GCC diagnostic push
#pragma GCC diagnostic ignored "-Wdeprecated-declarations"
make libgsf-1 an optional dependency dzsave vanishes if you don't have it, but you can still build vips
2014-05-09 14:13:18 +02:00
#include <gsf/gsf.h>
block deprecation warnings from libgsf with an uglu gcc progma
2020-11-29 15:41:19 +01:00
#pragma GCC diagnostic pop
make libgsf-1 an optional dependency dzsave vanishes if you don't have it, but you can still build vips
2014-05-09 14:13:18 +02:00
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
/* A GSF output object that can write to a VipsTarget.
*/
typedef struct _GsfOutputTarget {
GsfOutput output;
VipsTarget *target;
} GsfOutputTarget;
typedef struct {
GsfOutputClass output_class;
} GsfOutputTargetClass;
G_DEFINE_TYPE( GsfOutputTarget, gsf_output_target, GSF_OUTPUT_TYPE );
static gboolean
gsf_output_target_close( GsfOutput *output )
{
GsfOutputTarget *output_target = (GsfOutputTarget *) output;
if( output_target->target ) {
fix deprecation warning
2022-05-21 13:43:57 +02:00
/* No easy way to report errors here, sadly.
*/
(void) vips_target_end( output_target->target );
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
VIPS_UNREF( output_target->target );
return( TRUE );
}
return( FALSE );
}
static void
gsf_output_target_finalize( GObject *obj )
{
GsfOutputTarget *output_target = (GsfOutputTarget *) obj;
fix deprecation warning
2022-05-21 13:43:57 +02:00
(void) gsf_output_target_close( GSF_OUTPUT( output_target ) );
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
fix dzsave_target on older libgsf Older libgsf needs TRUE back from the GsfOutput seek method. see https://github.com/libvips/libvips/issues/2878
2022-06-21 19:00:07 +02:00
G_OBJECT_CLASS( gsf_output_target_parent_class )->finalize( obj );
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
}
static gboolean
gsf_output_target_write( GsfOutput *output,
size_t num_bytes, guint8 const *buffer )
{
GsfOutputTarget *output_target = (GsfOutputTarget *) output;
if( vips_target_write( output_target->target, buffer, num_bytes ) )
return( FALSE );
return( TRUE );
}
static gboolean
gsf_output_target_seek( GsfOutput *output, gsf_off_t offset, GSeekType whence )
{
Fix `dzsave_target` for ZIP output (#2927)
2022-07-15 20:39:56 +02:00
#ifdef HAVE_GSF_ZIP64
/* No seek needed.
*/
return( FALSE );
#else
GsfOutputTarget *output_target = (GsfOutputTarget *) output;
int stdio_whence;
switch( whence ) {
case G_SEEK_CUR: stdio_whence = SEEK_CUR; break;
case G_SEEK_END: stdio_whence = SEEK_END; break;
case G_SEEK_SET: stdio_whence = SEEK_SET; break;
default:
g_assert_not_reached();
}
if( vips_target_seek( output_target->target,
offset, stdio_whence ) == -1 )
return( FALSE );
fix dzsave_target on older libgsf Older libgsf needs TRUE back from the GsfOutput seek method. see https://github.com/libvips/libvips/issues/2878
2022-06-21 19:00:07 +02:00
/* This will make our parent class handle the seek.
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
*/
fix dzsave_target on older libgsf Older libgsf needs TRUE back from the GsfOutput seek method. see https://github.com/libvips/libvips/issues/2878
2022-06-21 19:00:07 +02:00
return( TRUE );
Fix `dzsave_target` for ZIP output (#2927)
2022-07-15 20:39:56 +02:00
#endif
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
}
static void
gsf_output_target_init( GsfOutputTarget *output )
{
}
static void
gsf_output_target_class_init( GsfOutputTargetClass *class )
{
GObjectClass *gobject_class = G_OBJECT_CLASS( class );
GsfOutputClass *output_class = GSF_OUTPUT_CLASS( class );
gobject_class->finalize = gsf_output_target_finalize;
output_class->Close = gsf_output_target_close;
output_class->Write = gsf_output_target_write;
output_class->Seek = gsf_output_target_seek;
}
static GsfOutput *
gsf_output_target_new( VipsTarget *target )
{
GsfOutputTarget *output;
output = g_object_new( gsf_output_target_get_type(), NULL );
output->target = target;
g_object_ref( target );
return( GSF_OUTPUT( output ) );
}
dzsave: improve compatibility with libgsf < 1.14.29 (#3017) * Implement `GsfOutfileStdio` by ourselves * Add missing `->root` init * Fix styling * Update ChangeLog
2022-08-30 18:10:38 +02:00
/* A GSF output object that can write to a directory.
*/
typedef struct _GsfOutputDir {
GsfOutfile output;
char *root;
} GsfOutputDir;
typedef struct {
GsfOutfileClass output_class;
} GsfOutputDirClass;
G_DEFINE_TYPE( GsfOutputDir, gsf_output_dir, GSF_OUTFILE_TYPE );
static gboolean
gsf_output_dir_close( GsfOutput *output )
{
return( TRUE );
}
static void
gsf_output_dir_finalize( GObject *obj )
{
GsfOutputDir *output_dir = (GsfOutputDir *) obj;
g_free( output_dir->root );
G_OBJECT_CLASS( gsf_output_dir_parent_class )->finalize( obj );
}
static GsfOutfile *
gsf_output_dir_new_valist( char const *root,
char const *first_property_name, va_list var_args )
{
GsfOutputDir *output_dir;
if( g_mkdir( root, 0777 ) != 0 && errno != EEXIST ) {
int save_errno = errno;
char *utf8name = g_filename_display_name( root );
vips_error( "dzsave",
_( "unable to create directory \"%s\", %s" ),
utf8name, g_strerror( save_errno ) );
g_free( utf8name );
return( NULL );
}
output_dir = (GsfOutputDir *) g_object_new_valist( GSF_OUTFILE_STDIO_TYPE,
first_property_name, var_args );
output_dir->root = g_strdup( root );
gsf_output_set_name_from_filename( GSF_OUTPUT( output_dir ), root );
return( GSF_OUTFILE( output_dir ) );
}
static GsfOutput *
gsf_output_dir_new_child( GsfOutfile *parent,
char const *name, gboolean is_dir,
char const *first_property_name,
va_list args )
{
GsfOutputDir *output_dir = (GsfOutputDir *) parent;
GsfOutput *child;
char *path = g_build_filename( output_dir->root, name, NULL );
if( is_dir )
child = (GsfOutput *) gsf_output_dir_new_valist( path,
first_property_name, args );
else
child = gsf_output_stdio_new_valist( path, NULL,
first_property_name, args );
g_free( path );
return( child );
}
static void
gsf_output_dir_init( GsfOutputDir *output )
{
output->root = NULL;
}
static void
gsf_output_dir_class_init( GsfOutputDirClass *class )
{
GObjectClass *gobject_class = G_OBJECT_CLASS( class );
GsfOutputClass *output_class = GSF_OUTPUT_CLASS( class );
GsfOutfileClass *outfile_class = GSF_OUTFILE_CLASS( class );
gobject_class->finalize = gsf_output_dir_finalize;
output_class->Close = gsf_output_dir_close;
output_class->Seek = NULL;
output_class->Write = NULL;
output_class->Vprintf = NULL;
outfile_class->new_child = gsf_output_dir_new_child;
}
static GsfOutput *
gsf_output_dir_new( char const *root, ... )
{
GsfOutfile *output;
va_list var_args;
va_start( var_args, root );
output = gsf_output_dir_new_valist( root, NULL, var_args );
va_end( var_args );
return( GSF_OUTPUT( output ) );
}
dzsave will write a zip file but see the TODO in dzsave.c - no way to pick a tile format, always jpeg - no way to pick a container format, always zip - needs testing - examine memuse, how well do we work for large zips - can we make a zip over 4gb?
2014-04-19 16:15:25 +02:00
/* Simple wrapper around libgsf.
*
make libgsf-1 an optional dependency dzsave vanishes if you don't have it, but you can still build vips
2014-05-09 14:13:18 +02:00
* We need to be able to do scattered writes to structured files. So while
dzsave will write a zip file but see the TODO in dzsave.c - no way to pick a tile format, always jpeg - no way to pick a container format, always zip - needs testing - examine memuse, how well do we work for large zips - can we make a zip over 4gb?
2014-04-19 16:15:25 +02:00
* building a zip (for example) we need to be able to write to file/a/b.jpg,
* then to file/c/d.jpg, then back to file/a/e.jpg. This is tricky with the
* libgsf API which is happier doing writes in order.
*
* Put an API over libgsf to track refs to all directories and finish/close
* them.
*/
/* Need to track the directory tree we are writing, with a ref for each
* GsfOutput.
*/
typedef struct _VipsGsfDirectory {
struct _VipsGsfDirectory *parent;
fix memleak in dzsave tree elements were not being freed, thanks lovell see https://github.com/jcupitt/libvips/issues/837
2018-01-01 12:40:04 +01:00
char *name;
dzsave will write a zip file but see the TODO in dzsave.c - no way to pick a tile format, always jpeg - no way to pick a container format, always zip - needs testing - examine memuse, how well do we work for large zips - can we make a zip over 4gb?
2014-04-19 16:15:25 +02:00
/* List of child directories, if any.
*/
GSList *children;
/* The GsfOutput we use for this object.
*/
GsfOutput *out;
sort out dzsave paths, auto-enable zip rationalised output paths for dzsave enable zip automatically if output name ends in .zip
2014-05-06 15:28:41 +02:00
/* The root node holds the enclosing zip file or FS root ... finish
* this on cleanup.
*/
GsfOutput *container;
stop dzsave zip write at 4gb see https://github.com/jcupitt/libvips/issues/138 not great, but better than producing a corrupt file
2014-06-25 15:34:13 +02:00
better >4gb detection for zip dzsave older libgsfs can't save zip64 and will fail silently for very large output trees ... improve the slightly sketchy >4gb detection in dzsave this a a version of the patch in https://github.com/jcupitt/libvips/pull/462
2016-11-11 17:24:18 +01:00
/* Track number of files in tree and total length of filenames. We use
* this to estimate zip size to spot a >4gb write.
*/
size_t file_count;
size_t filename_lengths;
Add dzsave zip compression support
2016-06-04 02:14:46 +02:00
/* Set deflate compression level for zip container.
*/
small dzsave cleanups - rename compression_level as deflate_level to match libgsf convention - use libgsf enums for GSF_ZIP_STORED, etc. - configure-time checks for zip64 and deflate-level - off-topic: rename NEED_TYPE_INIT as HAVE_TYPE_INIT for consistency
2016-06-09 10:36:37 +02:00
gint deflate_level;
Add dzsave zip compression support
2016-06-04 02:14:46 +02:00
dzsave will write a zip file but see the TODO in dzsave.c - no way to pick a tile format, always jpeg - no way to pick a container format, always zip - needs testing - examine memuse, how well do we work for large zips - can we make a zip over 4gb?
2014-04-19 16:15:25 +02:00
} VipsGsfDirectory;
fix a few more function casts see https://github.com/libvips/libvips/pull/1697#issuecomment-657602404
2020-07-14 11:25:35 +02:00
static void *vips_gsf_tree_close( VipsGsfDirectory *tree );
static void *
fix a compiler warning
2020-07-14 13:56:47 +02:00
vips_gsf_tree_close_cb( void *item, void *a, void *b )
fix a few more function casts see https://github.com/libvips/libvips/pull/1697#issuecomment-657602404
2020-07-14 11:25:35 +02:00
{
Fix a typo
2020-07-14 18:56:25 +02:00
VipsGsfDirectory *tree = (VipsGsfDirectory *) item;
fix a compiler warning
2020-07-14 13:56:47 +02:00
fix a few more function casts see https://github.com/libvips/libvips/pull/1697#issuecomment-657602404
2020-07-14 11:25:35 +02:00
return( vips_gsf_tree_close( tree ) );
}
dzsave will write a zip file but see the TODO in dzsave.c - no way to pick a tile format, always jpeg - no way to pick a container format, always zip - needs testing - examine memuse, how well do we work for large zips - can we make a zip over 4gb?
2014-04-19 16:15:25 +02:00
/* Close all dirs, non-NULL on error.
*/
static void *
vips_gsf_tree_close( VipsGsfDirectory *tree )
{
fix a few more function casts see https://github.com/libvips/libvips/pull/1697#issuecomment-657602404
2020-07-14 11:25:35 +02:00
vips_slist_map2( tree->children, vips_gsf_tree_close_cb, NULL, NULL );
sort out dzsave paths, auto-enable zip rationalised output paths for dzsave enable zip automatically if output name ends in .zip
2014-05-06 15:28:41 +02:00
better dzsave with zip output to a file close down output earlier to help mark-sweep bindings
2017-08-18 13:31:24 +02:00
if( tree->out ) {
if( !gsf_output_is_closed( tree->out ) &&
!gsf_output_close( tree->out ) ) {
vips_error( "vips_gsf",
"%s", _( "unable to close stream" ) );
return( tree );
}
sort out dzsave paths, auto-enable zip rationalised output paths for dzsave enable zip automatically if output name ends in .zip
2014-05-06 15:28:41 +02:00
fix memleak in dzsave tree elements were not being freed, thanks lovell see https://github.com/jcupitt/libvips/issues/837
2018-01-01 12:40:04 +01:00
VIPS_UNREF( tree->out );
dzsave will write a zip file but see the TODO in dzsave.c - no way to pick a tile format, always jpeg - no way to pick a container format, always zip - needs testing - examine memuse, how well do we work for large zips - can we make a zip over 4gb?
2014-04-19 16:15:25 +02:00
}
sort out dzsave paths, auto-enable zip rationalised output paths for dzsave enable zip automatically if output name ends in .zip
2014-05-06 15:28:41 +02:00
if( tree->container ) {
better dzsave with zip output to a file close down output earlier to help mark-sweep bindings
2017-08-18 13:31:24 +02:00
if( !gsf_output_is_closed( tree->container ) &&
!gsf_output_close( tree->container ) ) {
vips_error( "vips_gsf",
"%s", _( "unable to close stream" ) );
return( tree );
}
fix memleak in dzsave tree elements were not being freed, thanks lovell see https://github.com/jcupitt/libvips/issues/837
2018-01-01 12:40:04 +01:00
VIPS_UNREF( tree->container );
sort out dzsave paths, auto-enable zip rationalised output paths for dzsave enable zip automatically if output name ends in .zip
2014-05-06 15:28:41 +02:00
}
fix memleak in dzsave tree elements were not being freed, thanks lovell see https://github.com/jcupitt/libvips/issues/837
2018-01-01 12:40:04 +01:00
VIPS_FREEF( g_slist_free, tree->children );
VIPS_FREE( tree->name );
VIPS_FREE( tree );
dzsave will write a zip file but see the TODO in dzsave.c - no way to pick a tile format, always jpeg - no way to pick a container format, always zip - needs testing - examine memuse, how well do we work for large zips - can we make a zip over 4gb?
2014-04-19 16:15:25 +02:00
return( NULL );
}
/* Make a new tree root.
*/
static VipsGsfDirectory *
small dzsave cleanups - rename compression_level as deflate_level to match libgsf convention - use libgsf enums for GSF_ZIP_STORED, etc. - configure-time checks for zip64 and deflate-level - off-topic: rename NEED_TYPE_INIT as HAVE_TYPE_INIT for consistency
2016-06-09 10:36:37 +02:00
vips_gsf_tree_new( GsfOutput *out, gint deflate_level )
dzsave will write a zip file but see the TODO in dzsave.c - no way to pick a tile format, always jpeg - no way to pick a container format, always zip - needs testing - examine memuse, how well do we work for large zips - can we make a zip over 4gb?
2014-04-19 16:15:25 +02:00
{
VipsGsfDirectory *tree = g_new( VipsGsfDirectory, 1 );
tree->parent = NULL;
tree->name = NULL;
tree->children = NULL;
tree->out = out;
sort out dzsave paths, auto-enable zip rationalised output paths for dzsave enable zip automatically if output name ends in .zip
2014-05-06 15:28:41 +02:00
tree->container = NULL;
better >4gb detection for zip dzsave older libgsfs can't save zip64 and will fail silently for very large output trees ... improve the slightly sketchy >4gb detection in dzsave this a a version of the patch in https://github.com/jcupitt/libvips/pull/462
2016-11-11 17:24:18 +01:00
tree->file_count = 0;
tree->filename_lengths = 0;
small dzsave cleanups - rename compression_level as deflate_level to match libgsf convention - use libgsf enums for GSF_ZIP_STORED, etc. - configure-time checks for zip64 and deflate-level - off-topic: rename NEED_TYPE_INIT as HAVE_TYPE_INIT for consistency
2016-06-09 10:36:37 +02:00
tree->deflate_level = deflate_level;
dzsave will write a zip file but see the TODO in dzsave.c - no way to pick a tile format, always jpeg - no way to pick a container format, always zip - needs testing - examine memuse, how well do we work for large zips - can we make a zip over 4gb?
2014-04-19 16:15:25 +02:00
return( tree );
}
static void *
Fix function pointer cast issues It is undefined behavior in C and C++ to cast a function pointer to another type and call it that way. This does work in most native platforms, however, despite it being UB, but in WASM it can fail. See: https://emscripten.org/docs/porting/guidelines/function_pointer_issues.html
2020-06-30 12:10:30 +02:00
vips_gsf_child_by_name_sub( VipsGsfDirectory *dir, const char *name, void *b )
dzsave will write a zip file but see the TODO in dzsave.c - no way to pick a tile format, always jpeg - no way to pick a container format, always zip - needs testing - examine memuse, how well do we work for large zips - can we make a zip over 4gb?
2014-04-19 16:15:25 +02:00
{
if( strcmp( dir->name, name ) == 0 )
return( dir );
return( NULL );
}
/* Look up a child by name.
*/
static VipsGsfDirectory *
vips_gsf_child_by_name( VipsGsfDirectory *dir, const char *name )
{
return( vips_slist_map2( dir->children,
(VipsSListMap2Fn) vips_gsf_child_by_name_sub,
(char *) name, NULL ) );
}
/* Make a new directory.
*/
static VipsGsfDirectory *
vips_gsf_dir_new( VipsGsfDirectory *parent, const char *name )
{
VipsGsfDirectory *dir = g_new( VipsGsfDirectory, 1 );
g_assert( !vips_gsf_child_by_name( parent, name ) );
dir->parent = parent;
dir->name = g_strdup( name );
dir->children = NULL;
sort out dzsave paths, auto-enable zip rationalised output paths for dzsave enable zip automatically if output name ends in .zip
2014-05-06 15:28:41 +02:00
dir->container = NULL;
better >4gb detection for zip dzsave older libgsfs can't save zip64 and will fail silently for very large output trees ... improve the slightly sketchy >4gb detection in dzsave this a a version of the patch in https://github.com/jcupitt/libvips/pull/462
2016-11-11 17:24:18 +01:00
dir->file_count = 0;
dir->filename_lengths = 0;
small dzsave cleanups - rename compression_level as deflate_level to match libgsf convention - use libgsf enums for GSF_ZIP_STORED, etc. - configure-time checks for zip64 and deflate-level - off-topic: rename NEED_TYPE_INIT as HAVE_TYPE_INIT for consistency
2016-06-09 10:36:37 +02:00
dir->deflate_level = parent->deflate_level;
add @container option to dzsave
2014-04-25 11:26:23 +02:00
remove is_zip from dzsave we can use GSF_IS_OUTFILE_ZIP() instead
2016-06-08 15:46:45 +02:00
if( GSF_IS_OUTFILE_ZIP( parent->out ) )
add @container option to dzsave
2014-04-25 11:26:23 +02:00
dir->out = gsf_outfile_new_child_full(
(GsfOutfile *) parent->out,
name, TRUE,
small dzsave cleanups - rename compression_level as deflate_level to match libgsf convention - use libgsf enums for GSF_ZIP_STORED, etc. - configure-time checks for zip64 and deflate-level - off-topic: rename NEED_TYPE_INIT as HAVE_TYPE_INIT for consistency
2016-06-09 10:36:37 +02:00
"compression-level", GSF_ZIP_STORED,
add @container option to dzsave
2014-04-25 11:26:23 +02:00
NULL );
else
dir->out = gsf_outfile_new_child(
(GsfOutfile *) parent->out,
name, TRUE );
add some tests
2016-09-08 11:41:01 +02:00
g_assert( dir->out );
dzsave will write a zip file but see the TODO in dzsave.c - no way to pick a tile format, always jpeg - no way to pick a container format, always zip - needs testing - examine memuse, how well do we work for large zips - can we make a zip over 4gb?
2014-04-19 16:15:25 +02:00
parent->children = g_slist_prepend( parent->children, dir );
return( dir );
}
/* Return a GsfOutput for writing to a path. Paths are object name first, then
* path components with least-specific first, NULL-terminated. For example:
*
* GsfOutput *obj = vips_gsf_path( tree, "fred.jpg", "a", "b", NULL );
*
* Returns an obj you can use to write to a/b/fred.jpg.
*
* You must write, close and unref obj.
*/
static GsfOutput *
vips_gsf_path( VipsGsfDirectory *tree, const char *name, ... )
{
va_list ap;
VipsGsfDirectory *dir;
VipsGsfDirectory *child;
char *dir_name;
GsfOutput *obj;
better >4gb detection for zip dzsave older libgsfs can't save zip64 and will fail silently for very large output trees ... improve the slightly sketchy >4gb detection in dzsave this a a version of the patch in https://github.com/jcupitt/libvips/pull/462
2016-11-11 17:24:18 +01:00
/* vips_gsf_path() always makes a new file, though it may add to an
* existing directory. Note the file, and note the length of the full
* path we are creating.
*/
tree->file_count += 1;
tiny formatting improvement
2021-04-22 16:23:20 +02:00
tree->filename_lengths +=
strlen( tree->out->name ) + strlen( name ) + 1;
better >4gb detection for zip dzsave older libgsfs can't save zip64 and will fail silently for very large output trees ... improve the slightly sketchy >4gb detection in dzsave this a a version of the patch in https://github.com/jcupitt/libvips/pull/462
2016-11-11 17:24:18 +01:00
dzsave will write a zip file but see the TODO in dzsave.c - no way to pick a tile format, always jpeg - no way to pick a container format, always zip - needs testing - examine memuse, how well do we work for large zips - can we make a zip over 4gb?
2014-04-19 16:15:25 +02:00
dir = tree;
va_start( ap, name );
better >4gb detection for zip dzsave older libgsfs can't save zip64 and will fail silently for very large output trees ... improve the slightly sketchy >4gb detection in dzsave this a a version of the patch in https://github.com/jcupitt/libvips/pull/462
2016-11-11 17:24:18 +01:00
while( (dir_name = va_arg( ap, char * )) ) {
dzsave will write a zip file but see the TODO in dzsave.c - no way to pick a tile format, always jpeg - no way to pick a container format, always zip - needs testing - examine memuse, how well do we work for large zips - can we make a zip over 4gb?
2014-04-19 16:15:25 +02:00
if( (child = vips_gsf_child_by_name( dir, dir_name )) )
dir = child;
else
dir = vips_gsf_dir_new( dir, dir_name );
better >4gb detection for zip dzsave older libgsfs can't save zip64 and will fail silently for very large output trees ... improve the slightly sketchy >4gb detection in dzsave this a a version of the patch in https://github.com/jcupitt/libvips/pull/462
2016-11-11 17:24:18 +01:00
tree->filename_lengths += strlen( dir_name ) + 1;
}
dzsave will write a zip file but see the TODO in dzsave.c - no way to pick a tile format, always jpeg - no way to pick a container format, always zip - needs testing - examine memuse, how well do we work for large zips - can we make a zip over 4gb?
2014-04-19 16:15:25 +02:00
va_end( ap );
remove is_zip from dzsave we can use GSF_IS_OUTFILE_ZIP() instead
2016-06-08 15:46:45 +02:00
if( GSF_IS_OUTFILE_ZIP( dir->out ) ) {
small dzsave cleanups - rename compression_level as deflate_level to match libgsf convention - use libgsf enums for GSF_ZIP_STORED, etc. - configure-time checks for zip64 and deflate-level - off-topic: rename NEED_TYPE_INIT as HAVE_TYPE_INIT for consistency
2016-06-09 10:36:37 +02:00
/* Confusingly, libgsf compression-level really means
* compression-method. They have a separate deflate-level
* property for the deflate compression level.
*/
if( dir->deflate_level == 0 )
Add dzsave zip compression support
2016-06-04 02:14:46 +02:00
obj = gsf_outfile_new_child_full(
(GsfOutfile *) dir->out,
name, FALSE,
small dzsave cleanups - rename compression_level as deflate_level to match libgsf convention - use libgsf enums for GSF_ZIP_STORED, etc. - configure-time checks for zip64 and deflate-level - off-topic: rename NEED_TYPE_INIT as HAVE_TYPE_INIT for consistency
2016-06-09 10:36:37 +02:00
"compression-level", GSF_ZIP_STORED,
Add dzsave zip compression support
2016-06-04 02:14:46 +02:00
NULL );
small dzsave cleanups - rename compression_level as deflate_level to match libgsf convention - use libgsf enums for GSF_ZIP_STORED, etc. - configure-time checks for zip64 and deflate-level - off-topic: rename NEED_TYPE_INIT as HAVE_TYPE_INIT for consistency
2016-06-09 10:36:37 +02:00
else if( dir->deflate_level == -1 )
Add dzsave zip compression support
2016-06-04 02:14:46 +02:00
obj = gsf_outfile_new_child_full(
(GsfOutfile *) dir->out,
name, FALSE,
small dzsave cleanups - rename compression_level as deflate_level to match libgsf convention - use libgsf enums for GSF_ZIP_STORED, etc. - configure-time checks for zip64 and deflate-level - off-topic: rename NEED_TYPE_INIT as HAVE_TYPE_INIT for consistency
2016-06-09 10:36:37 +02:00
"compression-level", GSF_ZIP_DEFLATED,
Add dzsave zip compression support
2016-06-04 02:14:46 +02:00
NULL );
else
obj = gsf_outfile_new_child_full(
(GsfOutfile *) dir->out,
name, FALSE,
small dzsave cleanups - rename compression_level as deflate_level to match libgsf convention - use libgsf enums for GSF_ZIP_STORED, etc. - configure-time checks for zip64 and deflate-level - off-topic: rename NEED_TYPE_INIT as HAVE_TYPE_INIT for consistency
2016-06-09 10:36:37 +02:00
"compression-level", GSF_ZIP_DEFLATED,
"deflate-level", dir->deflate_level,
Add dzsave zip compression support
2016-06-04 02:14:46 +02:00
NULL );
remove is_zip from dzsave we can use GSF_IS_OUTFILE_ZIP() instead
2016-06-08 15:46:45 +02:00
}
add @container option to dzsave
2014-04-25 11:26:23 +02:00
else
obj = gsf_outfile_new_child( (GsfOutfile *) dir->out,
name, FALSE );
dzsave will write a zip file but see the TODO in dzsave.c - no way to pick a tile format, always jpeg - no way to pick a container format, always zip - needs testing - examine memuse, how well do we work for large zips - can we make a zip over 4gb?
2014-04-19 16:15:25 +02:00
return( obj );
}
hackety hack
2012-03-23 15:57:41 +01:00
typedef struct _VipsForeignSaveDz VipsForeignSaveDz;
typedef struct _Layer Layer;
started hacking in dzsave
2012-03-21 15:53:53 +01:00
/* A layer in the pyramid.
*/
struct _Layer {
VipsForeignSaveDz *dz;
fixes to deep zoom write - shrink down to a 1x1 pixel tile, even for very long and thin images - round image size up on shrink - write a .dzi file with the pyramid params - default tile size and overlap now matches the openslide writer thanks to Benjamin Gilbert for pointing out the errors
2012-08-06 16:14:59 +02:00
/* The real size of the image. image->Xsize and image->Ysize are
* always even to make x2 shrink easy. The real image may be a
* smaller, odd size,
*/
int width;
int height;
dzsave can write zoomify and google maps use "layout" to pick the directory layout, deepzoom / zoomify / google all allowed
2012-10-03 15:26:59 +02:00
/* Number of tiles across and down in this layer. Zoomify needs this
* to calculate the directory to put each tile in.
*/
int tiles_across;
int tiles_down;
faster --centre logic for dzsave don't test pixel-by-pixel for background tiles, instead track the rect of real pixels and test for intersection this might help performance scaling on many-core CPUs since we are no longer spinning up a threadgroup for each output tile thanks Kacey
2013-06-19 15:03:39 +02:00
/* The rect within width/height that contains real image, as opposed
* to background. In centre mode we can have large image borders.
*/
VipsRect real_pixels;
/* The image we build.
*/
VipsImage *image;
hackety hack
2012-03-23 15:57:41 +01:00
almost done fixing dzsave see https://github.com/jcupitt/libvips/issues/326 see TODO, problem with bottom edge tiles
2015-09-09 15:53:10 +02:00
/* The top of this strip of tiles.
more dzsave overlap fixups works with odd-sized tiles, overlaps handled correctly
2012-07-07 14:58:50 +02:00
*/
int y;
/* The next line we write to in this strip.
*/
int write_y;
hackery
2012-03-23 10:21:28 +01:00
VipsRegion *strip; /* The current strip of pixels */
works!
2012-03-26 23:08:59 +02:00
VipsRegion *copy; /* Pixels we copy to the next strip */
hackety hack
2012-03-23 15:57:41 +01:00
started hacking in dzsave
2012-03-21 15:53:53 +01:00
int sub; /* Subsample factor for this layer */
int n; /* Layer number ... 0 for smallest */
Layer *below; /* Tiles go to here */
Layer *above; /* Tiles come from here */
};
struct _VipsForeignSaveDz {
VipsForeignSave parent_object;
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
/* The target we are writing to. This is set by our subclasses.
*/
VipsTarget *target;
/* Alternatively, the filename, for filesystem output.
*/
char *filename;
started hacking in dzsave
2012-03-21 15:53:53 +01:00
char *suffix;
int overlap;
more improvements to dzsave now writes x_files and x.dzi, as per spec (thanks Benjamin) deprecate tile_width and tile_height, just have tile_size now
2012-08-07 14:29:03 +02:00
int tile_size;
dzsave can write zoomify and google maps use "layout" to pick the directory layout, deepzoom / zoomify / google all allowed
2012-10-03 15:26:59 +02:00
VipsForeignDzLayout layout;
add @depth option to dzsave so you can ask for a 1 layer pyramid
2012-11-01 19:11:46 +01:00
VipsForeignDzDepth depth;
add --centre option to dzsave with --centre turned on, in google maps mode dzsave will centre tiles rather than placing them at the top-left also, make sure blank.png is set to @background
2013-01-21 15:48:14 +01:00
gboolean centre;
more tweaks to dzsave props add docs too
2014-06-30 13:58:57 +02:00
gboolean properties;
add --angle option to dzsave
2013-05-29 15:06:23 +02:00
VipsAngle angle;
add @container option to dzsave
2014-04-25 11:26:23 +02:00
VipsForeignDzContainer container;
Add dzsave zip compression support
2016-06-04 02:14:46 +02:00
int compression;
Adds VipsRegionShrink type
2018-03-27 22:35:49 +02:00
VipsRegionShrink region_shrink;
make skip_blanks an int so you can adjust the threshold
2018-12-20 18:07:07 +01:00
int skip_blanks;
add no_strip option to dzsave it's sometimes useful to have pyramid tiles with ICC profiles see https://github.com/libvips/libvips/issues/1451
2019-10-21 17:46:20 +02:00
gboolean no_strip;
add @id arg to dzsave So you can now set the base of the id property in info.json in IIIF output. Thanks regisrob. See https://github.com/libvips/libvips/issues/1531
2020-01-24 18:12:46 +01:00
char *id;
started hacking in dzsave
2012-03-21 15:53:53 +01:00
new overlap system for dzsave again ... sigh see https://github.com/jcupitt/libvips/issues/512
2016-09-05 19:17:30 +02:00
/* Tile and overlap geometry. The members above are the parameters we
dzsave outputs extra tiles along right and bottom the deepzoom spec wants tiles to be written along the right and bottom, even of those tiles contain no new pixels this patch seems to work, but needs testing see: https://github.com/jcupitt/libvips/issues/795
2017-11-24 16:54:22 +01:00
* accept, this next set are the derived values which are actually
new overlap system for dzsave again ... sigh see https://github.com/jcupitt/libvips/issues/512
2016-09-05 19:17:30 +02:00
* used in pyramid generation.
*
dzsave outputs extra tiles along right and bottom the deepzoom spec wants tiles to be written along the right and bottom, even of those tiles contain no new pixels this patch seems to work, but needs testing see: https://github.com/jcupitt/libvips/issues/795
2017-11-24 16:54:22 +01:00
* Tiles have a base tile_size. Imagine a square placed at the top left.
new overlap system for dzsave again ... sigh see https://github.com/jcupitt/libvips/issues/512
2016-09-05 19:17:30 +02:00
* This is the size of that square.
*
* Tiles have a margin. The square from tile_size is expanded outward
* up/down/left/right by this amount. Parts going off the image are
* clipped.
*
* Each time we write a new tile, we step the position by tile_step
* pixels.
*
* We need all three of tile_size, tile_margin and tile_step since
* deepzoom and google maps have different meanings for overlap and we
* want to be able to support both models.
*
* For deepzoom:
*
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
* tile_margin = overlap
* tile_step = tile_size
new overlap system for dzsave again ... sigh see https://github.com/jcupitt/libvips/issues/512
2016-09-05 19:17:30 +02:00
*
* For google maps:
*
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
* tile_margin = 0
new overlap system for dzsave again ... sigh see https://github.com/jcupitt/libvips/issues/512
2016-09-05 19:17:30 +02:00
* tile_step = tile_size - overlap
*/
int tile_margin;
int tile_step;
started hacking in dzsave
2012-03-21 15:53:53 +01:00
Layer *layer; /* x2 shrink pyr layer */
dzsave can write zoomify and google maps use "layout" to pick the directory layout, deepzoom / zoomify / google all allowed
2012-10-03 15:26:59 +02:00
/* Count zoomify tiles we write.
*/
int tile_count;
dzsave will write a zip file but see the TODO in dzsave.c - no way to pick a tile format, always jpeg - no way to pick a container format, always zip - needs testing - examine memuse, how well do we work for large zips - can we make a zip over 4gb?
2014-04-19 16:15:25 +02:00
/* The tree structure we are writing tiles to. Can be filesystem, a
* zipfile, etc.
*/
VipsGsfDirectory *tree;
add dzsave_buffer zip only ... see https://github.com/jcupitt/libvips/issues/415
2016-10-15 20:01:34 +02:00
/* The actual output object our zip (or whatever) is writing to.
*/
GsfOutput *out;
/* The name to save as, eg. deepzoom tiles go into ${basename}_files.
* No suffix, no path at the start.
dzsave will write a zip file but see the TODO in dzsave.c - no way to pick a tile format, always jpeg - no way to pick a container format, always zip - needs testing - examine memuse, how well do we work for large zips - can we make a zip over 4gb?
2014-04-19 16:15:25 +02:00
*/
add @container option to dzsave
2014-04-25 11:26:23 +02:00
char *basename;
add dzsave_buffer zip only ... see https://github.com/jcupitt/libvips/issues/415
2016-10-15 20:01:34 +02:00
/* The directory we write the output to, or NULL for memory output.
add @container option to dzsave
2014-04-25 11:26:23 +02:00
*/
char *dirname;
dzsave will write a zip file but see the TODO in dzsave.c - no way to pick a tile format, always jpeg - no way to pick a container format, always zip - needs testing - examine memuse, how well do we work for large zips - can we make a zip over 4gb?
2014-04-19 16:15:25 +02:00
sort out dzsave paths, auto-enable zip rationalised output paths for dzsave enable zip automatically if output name ends in .zip
2014-05-06 15:28:41 +02:00
/* The root directory name ... $basename with perhaps some extra
* stuff, eg. $(basename)_files, etc.
dzsave will write a zip file but see the TODO in dzsave.c - no way to pick a tile format, always jpeg - no way to pick a container format, always zip - needs testing - examine memuse, how well do we work for large zips - can we make a zip over 4gb?
2014-04-19 16:15:25 +02:00
*/
char *root_name;
remove suffix from tiles see https://github.com/jcupitt/libvips/issues/129
2014-06-10 15:09:35 +02:00
/* @suffix, but without any options. So @suffix == ".jpg[Q=90]"
* becomes ".jpg".
*/
char *file_suffix;
stop dzsave zip write at 4gb see https://github.com/jcupitt/libvips/issues/138 not great, but better than producing a corrupt file
2014-06-25 15:34:13 +02:00
allow dzsave zip output > 4gb thanks benjamin
2015-06-08 10:35:25 +02:00
/* libgsf before 1.14.31 can't write zip files larger than 4gb.
* Track bytes written here and try to guess when we'll go over.
stop dzsave zip write at 4gb see https://github.com/jcupitt/libvips/issues/138 not great, but better than producing a corrupt file
2014-06-25 15:34:13 +02:00
*/
size_t bytes_written;
try skipping blank tiles ... again, but maybe we can be quicker this time
2015-11-24 16:07:23 +01:00
/* save->background turned into a pixel that matches the image we are
* saving .. used to test for blank tiles.
*/
VipsPel *ink;
started hacking in dzsave
2012-03-21 15:53:53 +01:00
};
typedef VipsForeignSaveClass VipsForeignSaveDzClass;
add dzsave_buffer zip only ... see https://github.com/jcupitt/libvips/issues/415
2016-10-15 20:01:34 +02:00
G_DEFINE_ABSTRACT_TYPE( VipsForeignSaveDz, vips_foreign_save_dz,
started hacking in dzsave
2012-03-21 15:53:53 +01:00
VIPS_TYPE_FOREIGN_SAVE );
oops fix copy-paste error
2018-01-28 18:23:05 +01:00
#define VIPS_ZIP_FIXED_LH_SIZE (30 + 29)
#define VIPS_ZIP_FIXED_CD_SIZE (46 + 9)
#define VIPS_ZIP_EOCD_SIZE 22
#ifndef HAVE_GSF_ZIP64
fix build with older libgsf build on libgsf without zip64 support was broken, thanks @remicollet see https://github.com/libvips/libvips/issues/2878
2022-06-20 14:19:35 +02:00
/* ZIP and SZI are both written as zip files.
*/
static gboolean
iszip( VipsForeignDzContainer container )
{
switch( container ) {
case VIPS_FOREIGN_DZ_CONTAINER_ZIP:
case VIPS_FOREIGN_DZ_CONTAINER_SZI:
return( TRUE );
default:
return( FALSE );
}
}
oops fix copy-paste error
2018-01-28 18:23:05 +01:00
static size_t
estimate_zip_size( VipsForeignSaveDz *dz )
{
Merge branch 'master' into HEAD
2019-04-11 17:18:39 +02:00
size_t estimated_zip_size = dz->bytes_written +
dz->tree->file_count * VIPS_ZIP_FIXED_LH_SIZE +
dz->tree->filename_lengths +
dz->tree->file_count * VIPS_ZIP_FIXED_CD_SIZE +
dz->tree->filename_lengths +
oops fix copy-paste error
2018-01-28 18:23:05 +01:00
VIPS_ZIP_EOCD_SIZE;
#ifdef DEBUG_VERBOSE
printf( "estimate_zip_size: %zd\n", estimated_zip_size );
#endif /*DEBUG_VERBOSE*/
return( estimated_zip_size );
}
#endif /*HAVE_GSF_ZIP64*/
write associated images to szi
2018-01-28 16:56:08 +01:00
static int
Merge branch 'master' into HEAD
2019-04-11 17:18:39 +02:00
write_image( VipsForeignSaveDz *dz,
write associated images to szi
2018-01-28 16:56:08 +01:00
GsfOutput *out, VipsImage *image, const char *format )
{
Merge branch 'master' into HEAD
2019-04-11 17:18:39 +02:00
VipsObjectClass *class = VIPS_OBJECT_GET_CLASS( dz );
write associated images to szi
2018-01-28 16:56:08 +01:00
fix some more metadata updates some more minor cases picked up by the test suite
2019-11-30 13:11:23 +01:00
VipsImage *t;
write associated images to szi
2018-01-28 16:56:08 +01:00
void *buf;
size_t len;
fix some more metadata updates some more minor cases picked up by the test suite
2019-11-30 13:11:23 +01:00
/* We need to block progress signalling on individual image write, so
* we need a copy of the tile in case it's shared (eg. associated
* images).
*/
if( vips_copy( image, &t, NULL ) )
return( -1 );
fix icc-profiles and dzsave --no-strip We were not copying metadata down pyramid layers in dzsave, so --no-strip didn't allow icc profiles on tiles. Thanks altert See https://github.com/libvips/libvips/issues/1879
2020-11-11 11:02:46 +01:00
/* We default to stripping all metadata. "no_strip" turns this
* off. Most people don't want metadata on every tile.
write associated images to szi
2018-01-28 16:56:08 +01:00
*/
fix some more metadata updates some more minor cases picked up by the test suite
2019-11-30 13:11:23 +01:00
vips_image_set_int( t, "hide-progress", 1 );
if( vips_image_write_to_buffer( t, format, &buf, &len,
add no_strip option to dzsave it's sometimes useful to have pyramid tiles with ICC profiles see https://github.com/libvips/libvips/issues/1451
2019-10-21 17:46:20 +02:00
"strip", !dz->no_strip,
fix some more metadata updates some more minor cases picked up by the test suite
2019-11-30 13:11:23 +01:00
NULL ) ) {
VIPS_UNREF( t );
write associated images to szi
2018-01-28 16:56:08 +01:00
return( -1 );
fix some more metadata updates some more minor cases picked up by the test suite
2019-11-30 13:11:23 +01:00
}
VIPS_UNREF( t );
write associated images to szi
2018-01-28 16:56:08 +01:00
/* gsf doesn't like more than one write active at once.
*/
g_mutex_lock( vips__global_lock );
if( !gsf_output_write( out, len, buf ) ) {
gsf_output_close( out );
g_mutex_unlock( vips__global_lock );
g_free( buf );
vips_error( class->nickname,
Merge branch 'master' into HEAD
2019-04-11 17:18:39 +02:00
"%s", gsf_output_error( out )->message );
write associated images to szi
2018-01-28 16:56:08 +01:00
return( -1 );
}
dz->bytes_written += len;
gsf_output_close( out );
#ifndef HAVE_GSF_ZIP64
fix build with older libgsf build on libgsf without zip64 support was broken, thanks @remicollet see https://github.com/libvips/libvips/issues/2878
2022-06-20 14:19:35 +02:00
if( iszip( dz->container ) ) {
Merge branch 'master' into HEAD
2019-04-11 17:18:39 +02:00
/* Leave 3 entry headroom for blank.png and metadata files.
write associated images to szi
2018-01-28 16:56:08 +01:00
*/
if( dz->tree->file_count + 3 >= (unsigned int) USHRT_MAX ) {
g_mutex_unlock( vips__global_lock );
vips_error( class->nickname,
Merge branch 'master' into HEAD
2019-04-11 17:18:39 +02:00
"%s", _( "too many files in zip" ) );
write associated images to szi
2018-01-28 16:56:08 +01:00
return( -1 );
}
/* Leave 16k headroom for blank.png and metadata files.
*/
if( estimate_zip_size( dz ) > (size_t) UINT_MAX - 16384) {
g_mutex_unlock( vips__global_lock );
vips_error( class->nickname,
"%s", _( "output file too large" ) );
return( -1 );
}
}
#endif /*HAVE_GSF_ZIP64*/
g_mutex_unlock( vips__global_lock );
g_free( buf );
return( 0 );
}
started hacking in dzsave
2012-03-21 15:53:53 +01:00
/* Free a pyramid.
*/
static void
layer_free( Layer *layer )
{
hackery
2012-03-23 10:21:28 +01:00
VIPS_FREEF( g_object_unref, layer->strip );
works!
2012-03-26 23:08:59 +02:00
VIPS_FREEF( g_object_unref, layer->copy );
hackery
2012-03-23 10:21:28 +01:00
VIPS_FREEF( g_object_unref, layer->image );
started hacking in dzsave
2012-03-21 15:53:53 +01:00
hackety hack
2012-03-23 15:57:41 +01:00
VIPS_FREEF( layer_free, layer->below );
started hacking in dzsave
2012-03-21 15:53:53 +01:00
}
static void
vips_foreign_save_dz_dispose( GObject *gobject )
{
VipsForeignSaveDz *dz = (VipsForeignSaveDz *) gobject;
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
VIPS_UNREF( dz->target );
hackety hack
2012-03-23 15:57:41 +01:00
VIPS_FREEF( layer_free, dz->layer );
better dzsave with zip output to a file close down output earlier to help mark-sweep bindings
2017-08-18 13:31:24 +02:00
VIPS_FREEF( vips_gsf_tree_close, dz->tree );
add dzsave_buffer zip only ... see https://github.com/jcupitt/libvips/issues/415
2016-10-15 20:01:34 +02:00
VIPS_FREEF( g_object_unref, dz->out );
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
add @container option to dzsave
2014-04-25 11:26:23 +02:00
VIPS_FREE( dz->basename );
VIPS_FREE( dz->dirname );
dzsave will write a zip file but see the TODO in dzsave.c - no way to pick a tile format, always jpeg - no way to pick a container format, always zip - needs testing - examine memuse, how well do we work for large zips - can we make a zip over 4gb?
2014-04-19 16:15:25 +02:00
VIPS_FREE( dz->root_name );
remove suffix from tiles see https://github.com/jcupitt/libvips/issues/129
2014-06-10 15:09:35 +02:00
VIPS_FREE( dz->file_suffix );
started hacking in dzsave
2012-03-21 15:53:53 +01:00
almost done fixing dzsave see https://github.com/jcupitt/libvips/issues/326 see TODO, problem with bottom edge tiles
2015-09-09 15:53:10 +02:00
G_OBJECT_CLASS( vips_foreign_save_dz_parent_class )->
dispose( gobject );
started hacking in dzsave
2012-03-21 15:53:53 +01:00
}
/* Build a pyramid.
faster --centre logic for dzsave don't test pixel-by-pixel for background tiles, instead track the rect of real pixels and test for intersection this might help performance scaling on many-core CPUs since we are no longer spinning up a threadgroup for each output tile thanks Kacey
2013-06-19 15:03:39 +02:00
*
* width/height is the size of this layer, real_* the subsection of the layer
* which is real pixels (as opposed to background).
started hacking in dzsave
2012-03-21 15:53:53 +01:00
*/
static Layer *
faster --centre logic for dzsave don't test pixel-by-pixel for background tiles, instead track the rect of real pixels and test for intersection this might help performance scaling on many-core CPUs since we are no longer spinning up a threadgroup for each output tile thanks Kacey
2013-06-19 15:03:39 +02:00
pyramid_build( VipsForeignSaveDz *dz, Layer *above,
int width, int height, VipsRect *real_pixels )
started hacking in dzsave
2012-03-21 15:53:53 +01:00
{
hackety hack
2012-03-23 15:57:41 +01:00
VipsForeignSave *save = VIPS_FOREIGN_SAVE( dz );
started hacking in dzsave
2012-03-21 15:53:53 +01:00
Layer *layer = VIPS_NEW( dz, Layer );
hackety hack
2012-03-23 15:57:41 +01:00
VipsRect strip;
dzsave can write zoomify and google maps use "layout" to pick the directory layout, deepzoom / zoomify / google all allowed
2012-10-03 15:26:59 +02:00
int limit;
started hacking in dzsave
2012-03-21 15:53:53 +01:00
layer->dz = dz;
fixes to deep zoom write - shrink down to a 1x1 pixel tile, even for very long and thin images - round image size up on shrink - write a .dzi file with the pyramid params - default tile size and overlap now matches the openslide writer thanks to Benjamin Gilbert for pointing out the errors
2012-08-06 16:14:59 +02:00
layer->width = width;
layer->height = height;
dzsave outputs extra tiles along right and bottom the deepzoom spec wants tiles to be written along the right and bottom, even of those tiles contain no new pixels this patch seems to work, but needs testing see: https://github.com/jcupitt/libvips/issues/795
2017-11-24 16:54:22 +01:00
/* We need to output all possible tiles, even if they give no new pixels.
seems to work still needs tests for gm overlap mode
2016-09-07 11:12:04 +02:00
*/
dzsave outputs extra tiles along right and bottom the deepzoom spec wants tiles to be written along the right and bottom, even of those tiles contain no new pixels this patch seems to work, but needs testing see: https://github.com/jcupitt/libvips/issues/795
2017-11-24 16:54:22 +01:00
layer->tiles_across = VIPS_ROUND_UP( width, dz->tile_step ) /
dz->tile_step;
layer->tiles_down = VIPS_ROUND_UP( height, dz->tile_step ) /
dz->tile_step;
dzsave can write zoomify and google maps use "layout" to pick the directory layout, deepzoom / zoomify / google all allowed
2012-10-03 15:26:59 +02:00
faster --centre logic for dzsave don't test pixel-by-pixel for background tiles, instead track the rect of real pixels and test for intersection this might help performance scaling on many-core CPUs since we are no longer spinning up a threadgroup for each output tile thanks Kacey
2013-06-19 15:03:39 +02:00
layer->real_pixels = *real_pixels;
works!
2012-03-26 23:08:59 +02:00
layer->image = NULL;
layer->strip = NULL;
layer->copy = NULL;
started hacking in dzsave
2012-03-21 15:53:53 +01:00
if( !above )
/* Top of pyramid.
*/
layer->sub = 1;
else
layer->sub = above->sub * 2;
layer->below = NULL;
layer->above = above;
fixes to deep zoom write - shrink down to a 1x1 pixel tile, even for very long and thin images - round image size up on shrink - write a .dzi file with the pyramid params - default tile size and overlap now matches the openslide writer thanks to Benjamin Gilbert for pointing out the errors
2012-08-06 16:14:59 +02:00
/* We round the image size up to an even number to make x2 shrink
* easy.
*/
hackery
2012-03-23 10:21:28 +01:00
layer->image = vips_image_new();
combin copy_fields and demand_hint We had two API calls, vips_image_copy_fields() and vips_demand_hint(). They are now combined in the single vips_image_pipeline() call. All operations are now slightly smaller and simpler.
2013-10-22 10:29:40 +02:00
if( vips_image_pipelinev( layer->image,
VIPS_DEMAND_STYLE_ANY, save->ready, NULL ) ) {
hackery
2012-03-23 10:21:28 +01:00
layer_free( layer );
return( NULL );
}
fixes to deep zoom write - shrink down to a 1x1 pixel tile, even for very long and thin images - round image size up on shrink - write a .dzi file with the pyramid params - default tile size and overlap now matches the openslide writer thanks to Benjamin Gilbert for pointing out the errors
2012-08-06 16:14:59 +02:00
layer->image->Xsize = width + (width & 1);
layer->image->Ysize = height + (height & 1);
works!
2012-03-26 23:08:59 +02:00
layer->strip = vips_region_new( layer->image );
layer->copy = vips_region_new( layer->image );
/* The regions will get used in the bg thread callback, so make sure
* we don't own them.
*/
vips__region_no_ownership( layer->strip );
vips__region_no_ownership( layer->copy );
hackery
2012-03-23 10:21:28 +01:00
almost done fixing dzsave see https://github.com/jcupitt/libvips/issues/326 see TODO, problem with bottom edge tiles
2015-09-09 15:53:10 +02:00
/* Build a line of tiles here.
fix odd tile height + no overlap dzsave with zero overlap and odd height was broken also, fix cache init
2012-07-08 11:37:35 +02:00
*
fix another dzsave corner case better padding calculation fixes yet another corner case for dzsave, thanks Martin
2013-02-26 13:32:01 +01:00
* Expand the strip if necessary to make sure we have an even
* number of lines.
new overlap system for dzsave again ... sigh see https://github.com/jcupitt/libvips/issues/512
2016-09-05 19:17:30 +02:00
*
* This is just the height of the first row of tiles, so only add 1*
* tile_margin.
hackety hack
2012-03-23 15:57:41 +01:00
*/
more dzsave overlap fixups works with odd-sized tiles, overlaps handled correctly
2012-07-07 14:58:50 +02:00
layer->y = 0;
layer->write_y = 0;
hackety hack
2012-03-23 15:57:41 +01:00
strip.left = 0;
strip.top = 0;
fixes to deep zoom write - shrink down to a 1x1 pixel tile, even for very long and thin images - round image size up on shrink - write a .dzi file with the pyramid params - default tile size and overlap now matches the openslide writer thanks to Benjamin Gilbert for pointing out the errors
2012-08-06 16:14:59 +02:00
strip.width = layer->image->Xsize;
new overlap system for dzsave again ... sigh see https://github.com/jcupitt/libvips/issues/512
2016-09-05 19:17:30 +02:00
strip.height = dz->tile_size + dz->tile_margin;
fix another dzsave corner case better padding calculation fixes yet another corner case for dzsave, thanks Martin
2013-02-26 13:32:01 +01:00
if( (strip.height & 1) == 1 )
strip.height += 1;
hackety hack
2012-03-23 15:57:41 +01:00
if( vips_region_buffer( layer->strip, &strip ) ) {
layer_free( layer );
return( NULL );
}
add @depth option to dzsave so you can ask for a 1 layer pyramid
2012-11-01 19:11:46 +01:00
switch( dz->depth ) {
rename _!PIXEL etc. to help bindings ruby gobject-introspection hates numbers at the start of _ separated names
2015-01-07 15:22:13 +01:00
case VIPS_FOREIGN_DZ_DEPTH_ONEPIXEL:
dzsave can write zoomify and google maps use "layout" to pick the directory layout, deepzoom / zoomify / google all allowed
2012-10-03 15:26:59 +02:00
limit = 1;
add @depth option to dzsave so you can ask for a 1 layer pyramid
2012-11-01 19:11:46 +01:00
break;
rename _!PIXEL etc. to help bindings ruby gobject-introspection hates numbers at the start of _ separated names
2015-01-07 15:22:13 +01:00
case VIPS_FOREIGN_DZ_DEPTH_ONETILE:
dzsave can write zoomify and google maps use "layout" to pick the directory layout, deepzoom / zoomify / google all allowed
2012-10-03 15:26:59 +02:00
limit = dz->tile_size;
add @depth option to dzsave so you can ask for a 1 layer pyramid
2012-11-01 19:11:46 +01:00
break;
rename _!PIXEL etc. to help bindings ruby gobject-introspection hates numbers at the start of _ separated names
2015-01-07 15:22:13 +01:00
case VIPS_FOREIGN_DZ_DEPTH_ONE:
add @depth option to dzsave so you can ask for a 1 layer pyramid
2012-11-01 19:11:46 +01:00
limit = VIPS_MAX( width, height );
break;
default:
use g_assert_not_reached() we were using g_assert(0) ... g_assert_not_reached() makes the intention clearer
2016-01-24 14:21:53 +01:00
g_assert_not_reached();
/* Stop compiler warnings.
*/
limit = 1;
add @depth option to dzsave so you can ask for a 1 layer pyramid
2012-11-01 19:11:46 +01:00
}
dzsave can write zoomify and google maps use "layout" to pick the directory layout, deepzoom / zoomify / google all allowed
2012-10-03 15:26:59 +02:00
if( width > limit ||
height > limit ) {
fixes to deep zoom write - shrink down to a 1x1 pixel tile, even for very long and thin images - round image size up on shrink - write a .dzi file with the pyramid params - default tile size and overlap now matches the openslide writer thanks to Benjamin Gilbert for pointing out the errors
2012-08-06 16:14:59 +02:00
/* Round up, so eg. a 5 pixel wide image becomes 3 a layer
* down.
faster --centre logic for dzsave don't test pixel-by-pixel for background tiles, instead track the rect of real pixels and test for intersection this might help performance scaling on many-core CPUs since we are no longer spinning up a threadgroup for each output tile thanks Kacey
2013-06-19 15:03:39 +02:00
*
* For the rect, round left/top down, round bottom/right up,
* so we get all possible pixels.
fixes to deep zoom write - shrink down to a 1x1 pixel tile, even for very long and thin images - round image size up on shrink - write a .dzi file with the pyramid params - default tile size and overlap now matches the openslide writer thanks to Benjamin Gilbert for pointing out the errors
2012-08-06 16:14:59 +02:00
*/
faster --centre logic for dzsave don't test pixel-by-pixel for background tiles, instead track the rect of real pixels and test for intersection this might help performance scaling on many-core CPUs since we are no longer spinning up a threadgroup for each output tile thanks Kacey
2013-06-19 15:03:39 +02:00
VipsRect halfrect;
halfrect.left = real_pixels->left / 2;
halfrect.top = real_pixels->top / 2;
halfrect.width = (VIPS_RECT_RIGHT( real_pixels ) + 1) / 2 -
halfrect.left;
halfrect.height = (VIPS_RECT_BOTTOM( real_pixels ) + 1) / 2 -
halfrect.top;
if( !(layer->below = pyramid_build( dz, layer,
(width + 1) / 2, (height + 1) / 2,
&halfrect )) ) {
started hacking in dzsave
2012-03-21 15:53:53 +01:00
layer_free( layer );
hackety hack
2012-03-23 15:57:41 +01:00
return( NULL );
started hacking in dzsave
2012-03-21 15:53:53 +01:00
}
hackety hack
2012-03-23 15:57:41 +01:00
layer->n = layer->below->n + 1;
started hacking in dzsave
2012-03-21 15:53:53 +01:00
}
else
layer->n = 0;
sdzsave fix if dzsave was running a scanline at a time and height was odd, the final set of tiles were not being written -- thanks Martin
2012-10-01 15:08:59 +02:00
#ifdef DEBUG
printf( "pyramid_build:\n" );
printf( "\tn = %d\n", layer->n );
printf( "\twidth = %d, height = %d\n", width, height );
printf( "\tXsize = %d, Ysize = %d\n",
layer->image->Xsize, layer->image->Ysize );
new overlap system for dzsave again ... sigh see https://github.com/jcupitt/libvips/issues/512
2016-09-05 19:17:30 +02:00
printf( "\ttiles_across = %d, tiles_down = %d\n",
layer->tiles_across, layer->tiles_down );
faster --centre logic for dzsave don't test pixel-by-pixel for background tiles, instead track the rect of real pixels and test for intersection this might help performance scaling on many-core CPUs since we are no longer spinning up a threadgroup for each output tile thanks Kacey
2013-06-19 15:03:39 +02:00
printf( "\treal_pixels.left = %d, real_pixels.top = %d\n",
real_pixels->left, real_pixels->top );
printf( "\treal_pixels.width = %d, real_pixels.height = %d\n",
real_pixels->width, real_pixels->height );
formatting in dzsave.c
2019-11-19 09:54:23 +01:00
#endif /*DEBUG*/
sdzsave fix if dzsave was running a scanline at a time and height was odd, the final set of tiles were not being written -- thanks Martin
2012-10-01 15:08:59 +02:00
started hacking in dzsave
2012-03-21 15:53:53 +01:00
return( layer );
}
fixes to deep zoom write - shrink down to a 1x1 pixel tile, even for very long and thin images - round image size up on shrink - write a .dzi file with the pyramid params - default tile size and overlap now matches the openslide writer thanks to Benjamin Gilbert for pointing out the errors
2012-08-06 16:14:59 +02:00
static int
write_dzi( VipsForeignSaveDz *dz )
{
dzsave will write a zip file but see the TODO in dzsave.c - no way to pick a tile format, always jpeg - no way to pick a container format, always zip - needs testing - examine memuse, how well do we work for large zips - can we make a zip over 4gb?
2014-04-19 16:15:25 +02:00
GsfOutput *out;
use a define for PATH_MAX add VIPS_PATH_MAX, a long-enough-for-almost-any-path buffer size we ought to switch to dynamic paths really ... do this when we fix filename encoding
2013-12-01 13:18:16 +01:00
char buf[VIPS_PATH_MAX];
fix .dzi format string dzsave left the save options attached to the Format string in the .dzi file also, version bump
2012-10-02 14:11:43 +02:00
char *p;
fixes to deep zoom write - shrink down to a 1x1 pixel tile, even for very long and thin images - round image size up on shrink - write a .dzi file with the pyramid params - default tile size and overlap now matches the openslide writer thanks to Benjamin Gilbert for pointing out the errors
2012-08-06 16:14:59 +02:00
sort out dzsave paths, auto-enable zip rationalised output paths for dzsave enable zip automatically if output name ends in .zip
2014-05-06 15:28:41 +02:00
vips_snprintf( buf, VIPS_PATH_MAX, "%s.dzi", dz->basename );
dzsave will write a zip file but see the TODO in dzsave.c - no way to pick a tile format, always jpeg - no way to pick a container format, always zip - needs testing - examine memuse, how well do we work for large zips - can we make a zip over 4gb?
2014-04-19 16:15:25 +02:00
out = vips_gsf_path( dz->tree, buf, NULL );
fixes to deep zoom write - shrink down to a 1x1 pixel tile, even for very long and thin images - round image size up on shrink - write a .dzi file with the pyramid params - default tile size and overlap now matches the openslide writer thanks to Benjamin Gilbert for pointing out the errors
2012-08-06 16:14:59 +02:00
use a define for PATH_MAX add VIPS_PATH_MAX, a long-enough-for-almost-any-path buffer size we ought to switch to dynamic paths really ... do this when we fix filename encoding
2013-12-01 13:18:16 +01:00
vips_snprintf( buf, VIPS_PATH_MAX, "%s", dz->suffix + 1 );
Merge remote-tracking branch 'origin/7.30' Conflicts: ChangeLog libvips/foreign/dzsave.c
2012-10-03 15:45:53 +02:00
if( (p = (char *) vips__find_rightmost_brackets( buf )) )
fix .dzi format string dzsave left the save options attached to the Format string in the .dzi file also, version bump
2012-10-02 14:11:43 +02:00
*p = '\0';
almost done fixing dzsave see https://github.com/jcupitt/libvips/issues/326 see TODO, problem with bottom edge tiles
2015-09-09 15:53:10 +02:00
gsf_output_printf( out, "<?xml "
"version=\"1.0\" encoding=\"UTF-8\"?>\n" );
dzsave will write a zip file but see the TODO in dzsave.c - no way to pick a tile format, always jpeg - no way to pick a container format, always zip - needs testing - examine memuse, how well do we work for large zips - can we make a zip over 4gb?
2014-04-19 16:15:25 +02:00
gsf_output_printf( out, "<Image "
fixes to deep zoom write - shrink down to a 1x1 pixel tile, even for very long and thin images - round image size up on shrink - write a .dzi file with the pyramid params - default tile size and overlap now matches the openslide writer thanks to Benjamin Gilbert for pointing out the errors
2012-08-06 16:14:59 +02:00
"xmlns=\"http://schemas.microsoft.com/deepzoom/2008\"\n" );
dzsave will write a zip file but see the TODO in dzsave.c - no way to pick a tile format, always jpeg - no way to pick a container format, always zip - needs testing - examine memuse, how well do we work for large zips - can we make a zip over 4gb?
2014-04-19 16:15:25 +02:00
gsf_output_printf( out, " Format=\"%s\"\n", buf );
gsf_output_printf( out, " Overlap=\"%d\"\n", dz->overlap );
gsf_output_printf( out, " TileSize=\"%d\"\n", dz->tile_size );
gsf_output_printf( out, " >\n" );
gsf_output_printf( out, " <Size \n" );
gsf_output_printf( out, " Height=\"%d\"\n", dz->layer->height );
gsf_output_printf( out, " Width=\"%d\"\n", dz->layer->width );
gsf_output_printf( out, " />\n" );
gsf_output_printf( out, "</Image>\n" );
(void) gsf_output_close( out );
g_object_unref( out );
fixes to deep zoom write - shrink down to a 1x1 pixel tile, even for very long and thin images - round image size up on shrink - write a .dzi file with the pyramid params - default tile size and overlap now matches the openslide writer thanks to Benjamin Gilbert for pointing out the errors
2012-08-06 16:14:59 +02:00
return( 0 );
}
dzsave can write zoomify and google maps use "layout" to pick the directory layout, deepzoom / zoomify / google all allowed
2012-10-03 15:26:59 +02:00
static int
write_properties( VipsForeignSaveDz *dz )
{
dzsave will write a zip file but see the TODO in dzsave.c - no way to pick a tile format, always jpeg - no way to pick a container format, always zip - needs testing - examine memuse, how well do we work for large zips - can we make a zip over 4gb?
2014-04-19 16:15:25 +02:00
GsfOutput *out;
dzsave can write zoomify and google maps use "layout" to pick the directory layout, deepzoom / zoomify / google all allowed
2012-10-03 15:26:59 +02:00
move ImageProperties.xml it's now in x/, not x/x/, same as the official tool
2014-09-25 12:16:14 +02:00
out = vips_gsf_path( dz->tree, "ImageProperties.xml", NULL );
dzsave can write zoomify and google maps use "layout" to pick the directory layout, deepzoom / zoomify / google all allowed
2012-10-03 15:26:59 +02:00
dzsave will write a zip file but see the TODO in dzsave.c - no way to pick a tile format, always jpeg - no way to pick a container format, always zip - needs testing - examine memuse, how well do we work for large zips - can we make a zip over 4gb?
2014-04-19 16:15:25 +02:00
gsf_output_printf( out, "<IMAGE_PROPERTIES "
dzsave can write zoomify and google maps use "layout" to pick the directory layout, deepzoom / zoomify / google all allowed
2012-10-03 15:26:59 +02:00
"WIDTH=\"%d\" HEIGHT=\"%d\" NUMTILES=\"%d\" "
"NUMIMAGES=\"1\" VERSION=\"1.8\" TILESIZE=\"%d\" />\n",
dz->layer->width,
dz->layer->height,
dz->tile_count,
dz->tile_size );
dzsave will write a zip file but see the TODO in dzsave.c - no way to pick a tile format, always jpeg - no way to pick a container format, always zip - needs testing - examine memuse, how well do we work for large zips - can we make a zip over 4gb?
2014-04-19 16:15:25 +02:00
(void) gsf_output_close( out );
g_object_unref( out );
dzsave can write zoomify and google maps use "layout" to pick the directory layout, deepzoom / zoomify / google all allowed
2012-10-03 15:26:59 +02:00
return( 0 );
}
static int
write_blank( VipsForeignSaveDz *dz )
{
two small fixes for dzsave fix a tiny leak, fix coloured @background
2015-10-06 14:30:25 +02:00
VipsForeignSave *save = (VipsForeignSave *) dz;
add --centre option to dzsave with --centre turned on, in google maps mode dzsave will centre tiles rather than placing them at the top-left also, make sure blank.png is set to @background
2013-01-21 15:48:14 +01:00
VipsImage *x, *t;
int n;
VipsArea *ones;
double *d;
two small fixes for dzsave fix a tiny leak, fix coloured @background
2015-10-06 14:30:25 +02:00
double *bg;
add --centre option to dzsave with --centre turned on, in google maps mode dzsave will centre tiles rather than placing them at the top-left also, make sure blank.png is set to @background
2013-01-21 15:48:14 +01:00
int i;
dzsave will write a zip file but see the TODO in dzsave.c - no way to pick a tile format, always jpeg - no way to pick a container format, always zip - needs testing - examine memuse, how well do we work for large zips - can we make a zip over 4gb?
2014-04-19 16:15:25 +02:00
GsfOutput *out;
dzsave can write zoomify and google maps use "layout" to pick the directory layout, deepzoom / zoomify / google all allowed
2012-10-03 15:26:59 +02:00
two small fixes for dzsave fix a tiny leak, fix coloured @background
2015-10-06 14:30:25 +02:00
/* Number of bands we will end up making. We need to set this in
* vips_black() to make sure we set Type correctly, otherwise we can
* try saving a B_W image as PNG, with disasterous results.
*/
remove some casts and use VIPS_AREA instead
2017-07-26 15:57:27 +02:00
bg = (double *) vips_area_get_data( VIPS_AREA( save->background ),
two small fixes for dzsave fix a tiny leak, fix coloured @background
2015-10-06 14:30:25 +02:00
NULL, &n, NULL, NULL );
if( vips_black( &x, dz->tile_size, dz->tile_size, "bands", n, NULL ) )
add --centre option to dzsave with --centre turned on, in google maps mode dzsave will centre tiles rather than placing them at the top-left also, make sure blank.png is set to @background
2013-01-21 15:48:14 +01:00
return( -1 );
ones = vips_area_new_array( G_TYPE_DOUBLE, sizeof( double ), n );
d = (double *) vips_area_get_data( ones, NULL, NULL, NULL, NULL );
for( i = 0; i < n; i++ )
d[i] = 1.0;
two small fixes for dzsave fix a tiny leak, fix coloured @background
2015-10-06 14:30:25 +02:00
if( vips_linear( x, &t, d, bg, n, NULL ) ) {
add --centre option to dzsave with --centre turned on, in google maps mode dzsave will centre tiles rather than placing them at the top-left also, make sure blank.png is set to @background
2013-01-21 15:48:14 +01:00
vips_area_unref( ones );
g_object_unref( x );
dzsave can write zoomify and google maps use "layout" to pick the directory layout, deepzoom / zoomify / google all allowed
2012-10-03 15:26:59 +02:00
return( -1 );
add --centre option to dzsave with --centre turned on, in google maps mode dzsave will centre tiles rather than placing them at the top-left also, make sure blank.png is set to @background
2013-01-21 15:48:14 +01:00
}
vips_area_unref( ones );
g_object_unref( x );
x = t;
szi save will write associated images if they have been attached by openslide load
2018-01-28 17:39:13 +01:00
out = vips_gsf_path( dz->tree, "blank.png", NULL );
if( write_image( dz, out, x, ".png" ) ) {
g_object_unref( out );
add --centre option to dzsave with --centre turned on, in google maps mode dzsave will centre tiles rather than placing them at the top-left also, make sure blank.png is set to @background
2013-01-21 15:48:14 +01:00
g_object_unref( x );
szi save will write associated images if they have been attached by openslide load
2018-01-28 17:39:13 +01:00
dzsave can write zoomify and google maps use "layout" to pick the directory layout, deepzoom / zoomify / google all allowed
2012-10-03 15:26:59 +02:00
return( -1 );
}
dzsave will write a zip file but see the TODO in dzsave.c - no way to pick a tile format, always jpeg - no way to pick a container format, always zip - needs testing - examine memuse, how well do we work for large zips - can we make a zip over 4gb?
2014-04-19 16:15:25 +02:00
g_object_unref( out );
szi save will write associated images if they have been attached by openslide load
2018-01-28 17:39:13 +01:00
g_object_unref( x );
dzsave will write a zip file but see the TODO in dzsave.c - no way to pick a tile format, always jpeg - no way to pick a container format, always zip - needs testing - examine memuse, how well do we work for large zips - can we make a zip over 4gb?
2014-04-19 16:15:25 +02:00
dzsave can write zoomify and google maps use "layout" to pick the directory layout, deepzoom / zoomify / google all allowed
2012-10-03 15:26:59 +02:00
return( 0 );
}
Add iiif3 (#2483) * add iiif3 support to dzsave some small things have changed from v2
2021-10-19 15:18:40 +02:00
/* Write IIIF/IIF3 JSON metadata.
add iiif layout to dzsave
2019-11-09 21:40:39 +01:00
*/
static int
write_json( VipsForeignSaveDz *dz )
{
set name in info.json
2019-11-11 16:05:51 +01:00
/* Can be NULL for memory output.
*/
const char *name = dz->basename ? dz->basename : "untitled";
clean up iiif output a bit better info.json, but still not support for the full/ directory see https://github.com/libvips/libvips/issues/1465
2019-11-11 14:38:58 +01:00
/* dz->file_suffix has a leading "." character.
*/
const char *suffix = dz->file_suffix[0] == '.' ?
dz->file_suffix + 1 : dz->file_suffix;
add iiif layout to dzsave
2019-11-09 21:40:39 +01:00
GsfOutput *out;
int i;
out = vips_gsf_path( dz->tree, "info.json", NULL );
Add iiif3 (#2483) * add iiif3 support to dzsave some small things have changed from v2
2021-10-19 15:18:40 +02:00
if( dz->layout == VIPS_FOREIGN_DZ_LAYOUT_IIIF3 )
gsf_output_printf( out,
"{\n"
" \"@context\": "
"\"http://iiif.io/api/image/3/context.json\",\n"
" \"id\": \"%s/%s\",\n"
" \"type\": \"ImageService3\",\n"
" \"profile\": \"level0\",\n"
" \"protocol\": \"http://iiif.io/api/image\",\n",
dz->id ? dz->id : "https://example.com/iiif",
name );
else
gsf_output_printf( out,
"{\n"
" \"@context\": "
"\"http://iiif.io/api/image/2/context.json\",\n"
" \"@id\": \"%s/%s\",\n"
" \"profile\": [\n"
" \"http://iiif.io/api/image/2/level0.json\",\n"
" {\n"
" \"formats\": [\n"
" \"%s\"\n"
" ],\n"
" \"qualities\": [\n"
" \"default\"\n"
" ]\n"
" }\n"
" ],\n"
" \"protocol\": \"http://iiif.io/api/image\",\n",
dz->id ? dz->id : "https://example.com/iiif",
name,
suffix );
clean up iiif output a bit better info.json, but still not support for the full/ directory see https://github.com/libvips/libvips/issues/1465
2019-11-11 14:38:58 +01:00
/* "sizes" is needed for the full/ set of untiled images, which we
* don't yet support. Leave this commented out for now.
gsf_output_printf( out,
add iiif layout to dzsave
2019-11-09 21:40:39 +01:00
" \"sizes\": [\n" );
for( i = 0; i < dz->layer->n + 5; i++ ) {
gsf_output_printf( out,
" {\n"
" \"width\": %d,\n"
" \"height\": \"full\"\n"
" }",
1 << (i + 4) );
if( i != dz->layer->n - 4 )
gsf_output_printf( out, "," );
gsf_output_printf( out, "\n" );
}
gsf_output_printf( out,
" ],\n" );
clean up iiif output a bit better info.json, but still not support for the full/ directory see https://github.com/libvips/libvips/issues/1465
2019-11-11 14:38:58 +01:00
*/
/* The set of pyramid layers we have written.
*/
add iiif layout to dzsave
2019-11-09 21:40:39 +01:00
gsf_output_printf( out,
" \"tiles\": [\n"
" {\n"
better iiif json Change "scalefactor" to "scaleFactor", thanks regisrob. See https://github.com/libvips/libvips/issues/1530
2020-01-24 17:28:52 +01:00
" \"scaleFactors\": [\n" );
add iiif layout to dzsave
2019-11-09 21:40:39 +01:00
for( i = 0; i < dz->layer->n; i++ ) {
gsf_output_printf( out,
" %d",
1 << i );
if( i != dz->layer->n - 1 )
gsf_output_printf( out, "," );
gsf_output_printf( out, "\n" );
}
gsf_output_printf( out,
" ],\n"
" \"width\": %d\n"
" }\n"
" ],\n", dz->tile_size );
gsf_output_printf( out,
" \"width\": %d,\n"
" \"height\": %d\n",
fix dzsave iiif dimensions dzsave in iiif mode could set info.json dimensions off by one thanks Linden6 see https://github.com/libvips/libvips/issues/1818
2020-09-17 18:07:08 +02:00
dz->layer->width,
dz->layer->height );
add iiif layout to dzsave
2019-11-09 21:40:39 +01:00
gsf_output_printf( out,
"}\n" );
(void) gsf_output_close( out );
g_object_unref( out );
return( 0 );
}
use libxml2 to write props don't write xml ourselves, get libxml2 to do the encoding
2014-07-01 13:07:31 +02:00
static int
write_vips_meta( VipsForeignSaveDz *dz )
{
add @properties flag to tiffsave writes all vips metadata to imagedescription tag
2015-12-21 15:43:59 +01:00
VipsForeignSave *save = (VipsForeignSave *) dz;
use libxml2 to write props don't write xml ourselves, get libxml2 to do the encoding
2014-07-01 13:07:31 +02:00
char *dump;
write props to a separate file and dump all vips props
2014-06-26 10:30:36 +02:00
GsfOutput *out;
remove libxml from xml save just some printfs now
2017-02-25 19:10:42 +01:00
if( !(dump = vips__xml_properties( save->ready )) )
use libxml2 to write props don't write xml ourselves, get libxml2 to do the encoding
2014-07-01 13:07:31 +02:00
return( -1 );
notes on vips-properties changes
2016-09-08 13:45:50 +02:00
/* For deepzom the props must go inside the ${name}_files subdir, for
* gm and zoomify it can sit in the main folder.
*/
Fix dzsave vips-properties path for non-dz layout This fixes a duplicate root path for the vips-properties.xml, when either the zoomify or goolgle layouts are used. Example: foo/foo/image-properties.xml -> foo/image-properties.xml
2016-06-06 01:27:16 +02:00
if( dz->layout == VIPS_FOREIGN_DZ_LAYOUT_DZ )
out = vips_gsf_path( dz->tree,
"vips-properties.xml", dz->root_name, NULL );
else
out = vips_gsf_path( dz->tree, "vips-properties.xml", NULL );
add @properties flag to tiffsave writes all vips metadata to imagedescription tag
2015-12-21 15:43:59 +01:00
gsf_output_write( out, strlen( dump ), (guchar *) dump );
write props to a separate file and dump all vips props
2014-06-26 10:30:36 +02:00
(void) gsf_output_close( out );
g_object_unref( out );
remove libxml from xml save just some printfs now
2017-02-25 19:10:42 +01:00
g_free( dump );
use libxml2 to write props don't write xml ourselves, get libxml2 to do the encoding
2014-07-01 13:07:31 +02:00
write props to a separate file and dump all vips props
2014-06-26 10:30:36 +02:00
return( 0 );
}
write scan-properties.xml to szi see https://github.com/jcupitt/libvips/issues/853
2018-01-26 18:20:58 +01:00
static void
build_scan_property( VipsDbuf *dbuf, VipsImage *image,
const char *vips_name, const char *szi_name )
{
const char *str;
add SZI container type and improve the scan-properties file a bit
2018-01-27 13:52:22 +01:00
GValue value = { 0 };
GValue save_value = { 0 };
GType type;
if( !vips_image_get_typeof( image, vips_name ) )
return;
if( vips_image_get( image, vips_name, &value ) )
return;
type = G_VALUE_TYPE( &value );
if( !g_value_type_transformable( type, VIPS_TYPE_SAVE_STRING ) ) {
g_value_unset( &value );
return;
}
g_value_init( &save_value, VIPS_TYPE_SAVE_STRING );
if( !g_value_transform( &value, &save_value ) ) {
g_value_unset( &value );
return;
}
g_value_unset( &value );
if( !(str = vips_value_get_save_string( &save_value )) ) {
g_value_unset( &save_value );
return;
}
write scan-properties.xml to szi see https://github.com/jcupitt/libvips/issues/853
2018-01-26 18:20:58 +01:00
Merge branch 'master' into HEAD
2019-04-11 17:18:39 +02:00
if( !g_utf8_validate( str, -1, NULL ) ) {
add SZI container type and improve the scan-properties file a bit
2018-01-27 13:52:22 +01:00
g_value_unset( &save_value );
return;
write scan-properties.xml to szi see https://github.com/jcupitt/libvips/issues/853
2018-01-26 18:20:58 +01:00
}
add SZI container type and improve the scan-properties file a bit
2018-01-27 13:52:22 +01:00
Merge branch 'master' into HEAD
2019-04-11 17:18:39 +02:00
vips_dbuf_writef( dbuf, " <property>\n" );
vips_dbuf_writef( dbuf, " <name>" );
add SZI container type and improve the scan-properties file a bit
2018-01-27 13:52:22 +01:00
vips_dbuf_write_amp( dbuf, szi_name );
Merge branch 'master' into HEAD
2019-04-11 17:18:39 +02:00
vips_dbuf_writef( dbuf, "</name>\n" );
vips_dbuf_writef( dbuf, " <value type=\"%s\">",
g_type_name( type ) );
add SZI container type and improve the scan-properties file a bit
2018-01-27 13:52:22 +01:00
vips_dbuf_write_amp( dbuf, str );
Merge branch 'master' into HEAD
2019-04-11 17:18:39 +02:00
vips_dbuf_writef( dbuf, "</value>\n" );
vips_dbuf_writef( dbuf, " </property>\n" );
add SZI container type and improve the scan-properties file a bit
2018-01-27 13:52:22 +01:00
g_value_unset( &save_value );
write scan-properties.xml to szi see https://github.com/jcupitt/libvips/issues/853
2018-01-26 18:20:58 +01:00
}
add SZI container type and improve the scan-properties file a bit
2018-01-27 13:52:22 +01:00
static char *scan_property_names[][2] = {
Merge branch 'master' into HEAD
2019-04-11 17:18:39 +02:00
{ "openslide.vendor", "Vendor" },
{ "openslide.objective-power", "ObjectiveMagnification" },
{ "openslide.mpp-x", "MicronsPerPixelX" },
{ "openslide.mpp-y", "MicronsPerPixelY" },
{ "width", "ImageWidth" },
add SZI container type and improve the scan-properties file a bit
2018-01-27 13:52:22 +01:00
{ "height", "ImageHeight" }
};
Merge branch 'master' into HEAD
2019-04-11 17:18:39 +02:00
/* Make the xml we write to scan-properties.xml in szi write.
write scan-properties.xml to szi see https://github.com/jcupitt/libvips/issues/853
2018-01-26 18:20:58 +01:00
* Free with g_free().
*/
Mark some functions and variables as static (#2691) These are only used in the same file, and not exported.
2022-02-27 14:41:39 +01:00
static char *
write scan-properties.xml to szi see https://github.com/jcupitt/libvips/issues/853
2018-01-26 18:20:58 +01:00
build_scan_properties( VipsImage *image )
{
VipsDbuf dbuf;
char *date;
add SZI container type and improve the scan-properties file a bit
2018-01-27 13:52:22 +01:00
int i;
write scan-properties.xml to szi see https://github.com/jcupitt/libvips/issues/853
2018-01-26 18:20:58 +01:00
move ppm save into the class no separate save function now
2019-11-15 11:23:45 +01:00
date = vips__get_iso8601();
revise decode/unminimise use
2019-10-28 06:04:04 +01:00
Merge branch 'master' into HEAD
2019-04-11 17:18:39 +02:00
vips_dbuf_init( &dbuf );
write scan-properties.xml to szi see https://github.com/jcupitt/libvips/issues/853
2018-01-26 18:20:58 +01:00
vips_dbuf_writef( &dbuf, "<?xml version=\"1.0\"?>\n" );
vips_dbuf_writef( &dbuf, "<image xmlns=\"http://www.pathozoom.com/szi\""
Merge branch 'master' into HEAD
2019-04-11 17:18:39 +02:00
" date=\"%s\" version=\"1.0\">\n", date );
write scan-properties.xml to szi see https://github.com/jcupitt/libvips/issues/853
2018-01-26 18:20:58 +01:00
vips_dbuf_writef( &dbuf, " <properties>\n" );
revise decode/unminimise use
2019-10-28 06:04:04 +01:00
g_free( date );
add SZI container type and improve the scan-properties file a bit
2018-01-27 13:52:22 +01:00
for( i = 0; i < VIPS_NUMBER( scan_property_names ); i++ )
Merge branch 'master' into HEAD
2019-04-11 17:18:39 +02:00
build_scan_property( &dbuf, image,
scan_property_names[i][0],
scan_property_names[i][1] );
write scan-properties.xml to szi see https://github.com/jcupitt/libvips/issues/853
2018-01-26 18:20:58 +01:00
Merge branch 'master' into HEAD
2019-04-11 17:18:39 +02:00
vips_dbuf_writef( &dbuf, " </properties>\n" );
vips_dbuf_writef( &dbuf, "</image>\n" );
write scan-properties.xml to szi see https://github.com/jcupitt/libvips/issues/853
2018-01-26 18:20:58 +01:00
return( (char *) vips_dbuf_steal( &dbuf, NULL ) );
}
static int
write_scan_properties( VipsForeignSaveDz *dz )
{
VipsForeignSave *save = (VipsForeignSave *) dz;
char *dump;
GsfOutput *out;
if( !(dump = build_scan_properties( save->ready )) )
return( -1 );
out = vips_gsf_path( dz->tree, "scan-properties.xml", NULL );
Merge branch 'master' into HEAD
2019-04-11 17:18:39 +02:00
gsf_output_write( out, strlen( dump ), (guchar *) dump );
write scan-properties.xml to szi see https://github.com/jcupitt/libvips/issues/853
2018-01-26 18:20:58 +01:00
(void) gsf_output_close( out );
g_object_unref( out );
g_free( dump );
return( 0 );
}
write associated images to szi
2018-01-28 16:56:08 +01:00
static void *
Merge branch 'master' into HEAD
2019-04-11 17:18:39 +02:00
write_associated_images( VipsImage *image,
write associated images to szi
2018-01-28 16:56:08 +01:00
const char *field, GValue *value, void *a )
{
VipsForeignSaveDz *dz = (VipsForeignSaveDz *) a;
if( vips_isprefix( "openslide.associated.", field ) ) {
VipsImage *associated;
const char *p;
const char *q;
GsfOutput *out;
char buf[VIPS_PATH_MAX];
p = field + strlen( "openslide.associated." );
/* Make sure there are no '/' in the filename.
*/
Merge branch 'master' into HEAD
2019-04-11 17:18:39 +02:00
if( (q = strrchr( p, '/' )) )
write associated images to szi
2018-01-28 16:56:08 +01:00
p = q + 1;
if( vips_image_get_image( image, field, &associated ) )
return( image );
vips_snprintf( buf, VIPS_PATH_MAX, "%s.jpg", p );
out = vips_gsf_path( dz->tree, buf, "associated_images", NULL );
if( write_image( dz, out, associated, ".jpg" ) ) {
g_object_unref( out );
g_object_unref( associated );
return( image );
}
g_object_unref( out );
g_object_unref( associated );
}
return( NULL );
}
static int
write_associated( VipsForeignSaveDz *dz )
{
VipsForeignSave *save = (VipsForeignSave *) dz;
szi save will write associated images if they have been attached by openslide load
2018-01-28 17:39:13 +01:00
if( vips_image_map( save->ready, write_associated_images, dz ) )
write associated images to szi
2018-01-28 16:56:08 +01:00
return( -1 );
return( 0 );
}
add threaded write of deepzoom tiles
2012-07-07 19:42:12 +02:00
/* Our state during a threaded write of a strip.
started hacking in dzsave
2012-03-21 15:53:53 +01:00
*/
add threaded write of deepzoom tiles
2012-07-07 19:42:12 +02:00
typedef struct _Strip {
Layer *layer;
fixes to deep zoom write - shrink down to a 1x1 pixel tile, even for very long and thin images - round image size up on shrink - write a .dzi file with the pyramid params - default tile size and overlap now matches the openslide writer thanks to Benjamin Gilbert for pointing out the errors
2012-08-06 16:14:59 +02:00
add threaded write of deepzoom tiles
2012-07-07 19:42:12 +02:00
VipsImage *image;
/* Allocate the next tile on this boundary.
*/
int x;
} Strip;
static void
strip_free( Strip *strip )
{
g_object_unref( strip->image );
}
static void
strip_init( Strip *strip, Layer *layer )
started hacking in dzsave
2012-03-21 15:53:53 +01:00
{
VipsForeignSaveDz *dz = layer->dz;
add threaded write of deepzoom tiles
2012-07-07 19:42:12 +02:00
VipsRect line, image;
started hacking in dzsave
2012-03-21 15:53:53 +01:00
add threaded write of deepzoom tiles
2012-07-07 19:42:12 +02:00
strip->layer = layer;
strip->image = NULL;
strip->x = 0;
fixes to deep zoom write - shrink down to a 1x1 pixel tile, even for very long and thin images - round image size up on shrink - write a .dzi file with the pyramid params - default tile size and overlap now matches the openslide writer thanks to Benjamin Gilbert for pointing out the errors
2012-08-06 16:14:59 +02:00
/* The image we wrap around our pixel buffer must be the full width,
* including any rounding up, since we must have contiguous pixels.
* We can trim the height down though.
*
* When we loop across the strip writing tiles we have to look out for
* the smaller width.
*/
more dzsave overlap fixups works with odd-sized tiles, overlaps handled correctly
2012-07-07 14:58:50 +02:00
image.left = 0;
image.top = 0;
image.width = layer->image->Xsize;
fixes to deep zoom write - shrink down to a 1x1 pixel tile, even for very long and thin images - round image size up on shrink - write a .dzi file with the pyramid params - default tile size and overlap now matches the openslide writer thanks to Benjamin Gilbert for pointing out the errors
2012-08-06 16:14:59 +02:00
image.height = layer->height;
line.left = 0;
almost done fixing dzsave see https://github.com/jcupitt/libvips/issues/326 see TODO, problem with bottom edge tiles
2015-09-09 15:53:10 +02:00
line.top = layer->y;
fixes to deep zoom write - shrink down to a 1x1 pixel tile, even for very long and thin images - round image size up on shrink - write a .dzi file with the pyramid params - default tile size and overlap now matches the openslide writer thanks to Benjamin Gilbert for pointing out the errors
2012-08-06 16:14:59 +02:00
line.width = image.width;
almost done fixing dzsave see https://github.com/jcupitt/libvips/issues/326 see TODO, problem with bottom edge tiles
2015-09-09 15:53:10 +02:00
line.height = dz->tile_size;
new overlap system for dzsave again ... sigh see https://github.com/jcupitt/libvips/issues/512
2016-09-05 19:17:30 +02:00
vips_rect_marginadjust( &line, dz->tile_margin );
more dzsave overlap fixups works with odd-sized tiles, overlaps handled correctly
2012-07-07 14:58:50 +02:00
add threaded write of deepzoom tiles
2012-07-07 19:42:12 +02:00
vips_rect_intersectrect( &image, &line, &line );
more dzsave overlap fixups works with odd-sized tiles, overlaps handled correctly
2012-07-07 14:58:50 +02:00
add threaded write of deepzoom tiles
2012-07-07 19:42:12 +02:00
if( !(strip->image = vips_image_new_from_memory(
VIPS_REGION_ADDR( layer->strip, 0, line.top ),
python new_from_memory()/write_to_memory() done
2014-09-05 13:54:07 +02:00
VIPS_IMAGE_SIZEOF_LINE( layer->image ) * line.height,
add threaded write of deepzoom tiles
2012-07-07 19:42:12 +02:00
line.width, line.height,
layer->image->Bands, layer->image->BandFmt )) ) {
strip_free( strip );
return;
}
set Type on memory strip we need to set Type on memory strips so we can convert to the output format correctly, thanks philipgiuliani see https://github.com/jcupitt/libvips/issues/120
2014-05-08 15:24:22 +02:00
fix icc-profiles and dzsave --no-strip We were not copying metadata down pyramid layers in dzsave, so --no-strip didn't allow icc profiles on tiles. Thanks altert See https://github.com/libvips/libvips/issues/1879
2020-11-11 11:02:46 +01:00
/* The strip needs to inherit the layer's metadata.
*/
if( vips__image_meta_copy( strip->image, layer->image ) ) {
strip_free( strip );
return;
}
set Type on memory strip we need to set Type on memory strips so we can convert to the output format correctly, thanks philipgiuliani see https://github.com/jcupitt/libvips/issues/120
2014-05-08 15:24:22 +02:00
/* Type needs to be set so we know how to convert for save correctly.
*/
strip->image->Type = layer->image->Type;
add threaded write of deepzoom tiles
2012-07-07 19:42:12 +02:00
}
started hacking in dzsave
2012-03-21 15:53:53 +01:00
add threaded write of deepzoom tiles
2012-07-07 19:42:12 +02:00
static int
strip_allocate( VipsThreadState *state, void *a, gboolean *stop )
{
Strip *strip = (Strip *) a;
Layer *layer = strip->layer;
VipsForeignSaveDz *dz = layer->dz;
hackety hack
2012-03-23 15:57:41 +01:00
add threaded write of deepzoom tiles
2012-07-07 19:42:12 +02:00
VipsRect image;
hackety hack
2012-03-23 15:57:41 +01:00
hacking
2013-02-25 15:54:39 +01:00
#ifdef DEBUG_VERBOSE
easier array arg creation added stdarg array double creator, used the _new as well, small other fixes
2012-11-02 22:45:13 +01:00
printf( "strip_allocate\n" );
hacking
2013-02-25 15:54:39 +01:00
#endif /*DEBUG_VERBOSE*/
easier array arg creation added stdarg array double creator, used the _new as well, small other fixes
2012-11-02 22:45:13 +01:00
fix up dzsave overlap handling There was a mixup with the previous fix to dzsave overlap handling, correct it and update the test suite. In the previous revision, dzsave overlapped tiles by overlap and sized them by tile_size. In fact, tiles should be sized as (tile_size + overlap * 2), ie. tile_size refers to the number of unique pixels per tile. See https://github.com/jcupitt/libvips/issues/357
2015-12-17 12:54:38 +01:00
/* We can't test for allocated area empty, since it might just have
* bits of the left-hand overlap in and no new pixels. Safest to count
* tiles across.
*/
new overlap system for dzsave again ... sigh see https://github.com/jcupitt/libvips/issues/512
2016-09-05 19:17:30 +02:00
if( strip->x / dz->tile_step >= layer->tiles_across ) {
fix up dzsave overlap handling There was a mixup with the previous fix to dzsave overlap handling, correct it and update the test suite. In the previous revision, dzsave overlapped tiles by overlap and sized them by tile_size. In fact, tiles should be sized as (tile_size + overlap * 2), ie. tile_size refers to the number of unique pixels per tile. See https://github.com/jcupitt/libvips/issues/357
2015-12-17 12:54:38 +01:00
*stop = TRUE;
#ifdef DEBUG_VERBOSE
printf( "strip_allocate: done\n" );
#endif /*DEBUG_VERBOSE*/
return( 0 );
}
fixes to deep zoom write - shrink down to a 1x1 pixel tile, even for very long and thin images - round image size up on shrink - write a .dzi file with the pyramid params - default tile size and overlap now matches the openslide writer thanks to Benjamin Gilbert for pointing out the errors
2012-08-06 16:14:59 +02:00
image.left = 0;
image.top = 0;
image.width = layer->width;
image.height = layer->height;
add threaded write of deepzoom tiles
2012-07-07 19:42:12 +02:00
/* Position this tile.
*/
almost done fixing dzsave see https://github.com/jcupitt/libvips/issues/326 see TODO, problem with bottom edge tiles
2015-09-09 15:53:10 +02:00
state->pos.left = strip->x;
working, at least with bug case working, at least with robclouth test case see https://github.com/jcupitt/libvips/issues/326#issuecomment-138888981
2015-09-09 18:27:19 +02:00
state->pos.top = layer->y;
almost done fixing dzsave see https://github.com/jcupitt/libvips/issues/326 see TODO, problem with bottom edge tiles
2015-09-09 15:53:10 +02:00
state->pos.width = dz->tile_size;
state->pos.height = dz->tile_size;
new overlap system for dzsave again ... sigh see https://github.com/jcupitt/libvips/issues/512
2016-09-05 19:17:30 +02:00
vips_rect_marginadjust( &state->pos, dz->tile_margin );
started hacking in dzsave
2012-03-21 15:53:53 +01:00
add threaded write of deepzoom tiles
2012-07-07 19:42:12 +02:00
vips_rect_intersectrect( &image, &state->pos, &state->pos );
state->x = strip->x;
state->y = layer->y;
more dzsave overlap fixups works with odd-sized tiles, overlaps handled correctly
2012-07-07 14:58:50 +02:00
new overlap system for dzsave again ... sigh see https://github.com/jcupitt/libvips/issues/512
2016-09-05 19:17:30 +02:00
strip->x += dz->tile_step;
add threaded write of deepzoom tiles
2012-07-07 19:42:12 +02:00
return( 0 );
}
dzsave will write a zip file but see the TODO in dzsave.c - no way to pick a tile format, always jpeg - no way to pick a container format, always zip - needs testing - examine memuse, how well do we work for large zips - can we make a zip over 4gb?
2014-04-19 16:15:25 +02:00
/* Make an output object for a tile in the current layout.
dzsave can write zoomify and google maps use "layout" to pick the directory layout, deepzoom / zoomify / google all allowed
2012-10-03 15:26:59 +02:00
*/
dzsave will write a zip file but see the TODO in dzsave.c - no way to pick a tile format, always jpeg - no way to pick a container format, always zip - needs testing - examine memuse, how well do we work for large zips - can we make a zip over 4gb?
2014-04-19 16:15:25 +02:00
static GsfOutput *
tile_name( Layer *layer, int x, int y )
dzsave can write zoomify and google maps use "layout" to pick the directory layout, deepzoom / zoomify / google all allowed
2012-10-03 15:26:59 +02:00
{
VipsForeignSaveDz *dz = layer->dz;
better iiif tile naming We were clipping edge tiles against the size of the scaled up layer. Instead, clip against the size of the top level image. See https://github.com/libvips/libvips/issues/1621
2020-04-24 20:29:35 +02:00
VipsForeignSave *save = (VipsForeignSave *) dz;
dzsave can write zoomify and google maps use "layout" to pick the directory layout, deepzoom / zoomify / google all allowed
2012-10-03 15:26:59 +02:00
dzsave will write a zip file but see the TODO in dzsave.c - no way to pick a tile format, always jpeg - no way to pick a container format, always zip - needs testing - examine memuse, how well do we work for large zips - can we make a zip over 4gb?
2014-04-19 16:15:25 +02:00
GsfOutput *out;
char name[VIPS_PATH_MAX];
use a define for PATH_MAX add VIPS_PATH_MAX, a long-enough-for-almost-any-path buffer size we ought to switch to dynamic paths really ... do this when we fix filename encoding
2013-12-01 13:18:16 +01:00
char dirname[VIPS_PATH_MAX];
char dirname2[VIPS_PATH_MAX];
dzsave can write zoomify and google maps use "layout" to pick the directory layout, deepzoom / zoomify / google all allowed
2012-10-03 15:26:59 +02:00
Layer *p;
int n;
switch( dz->layout ) {
case VIPS_FOREIGN_DZ_LAYOUT_DZ:
dzsave will write a zip file but see the TODO in dzsave.c - no way to pick a tile format, always jpeg - no way to pick a container format, always zip - needs testing - examine memuse, how well do we work for large zips - can we make a zip over 4gb?
2014-04-19 16:15:25 +02:00
vips_snprintf( dirname, VIPS_PATH_MAX, "%d", layer->n );
vips_snprintf( name, VIPS_PATH_MAX,
remove suffix from tiles see https://github.com/jcupitt/libvips/issues/129
2014-06-10 15:09:35 +02:00
"%d_%d%s", x, y, dz->file_suffix );
dzsave will write a zip file but see the TODO in dzsave.c - no way to pick a tile format, always jpeg - no way to pick a container format, always zip - needs testing - examine memuse, how well do we work for large zips - can we make a zip over 4gb?
2014-04-19 16:15:25 +02:00
out = vips_gsf_path( dz->tree, name,
dz->root_name, dirname, NULL );
dzsave can write zoomify and google maps use "layout" to pick the directory layout, deepzoom / zoomify / google all allowed
2012-10-03 15:26:59 +02:00
break;
case VIPS_FOREIGN_DZ_LAYOUT_ZOOMIFY:
/* We need to work out the tile number so we can calculate the
* directory to put this tile in.
*
* Tiles are numbered from 0 for the most-zoomed-out tile.
*/
n = 0;
/* Count all tiles in layers below this one.
*/
for( p = layer->below; p; p = p->below )
n += p->tiles_across * p->tiles_down;
/* And count tiles so far in this layer.
*/
n += y * layer->tiles_across + x;
dzsave will write a zip file but see the TODO in dzsave.c - no way to pick a tile format, always jpeg - no way to pick a container format, always zip - needs testing - examine memuse, how well do we work for large zips - can we make a zip over 4gb?
2014-04-19 16:15:25 +02:00
vips_snprintf( dirname, VIPS_PATH_MAX, "TileGroup%d", n / 256 );
vips_snprintf( name, VIPS_PATH_MAX,
remove suffix from tiles see https://github.com/jcupitt/libvips/issues/129
2014-06-10 15:09:35 +02:00
"%d-%d-%d%s", layer->n, x, y, dz->file_suffix );
dzsave can write zoomify and google maps use "layout" to pick the directory layout, deepzoom / zoomify / google all allowed
2012-10-03 15:26:59 +02:00
/* Used at the end in ImageProperties.xml
*/
dz->tile_count += 1;
sort out dzsave paths, auto-enable zip rationalised output paths for dzsave enable zip automatically if output name ends in .zip
2014-05-06 15:28:41 +02:00
out = vips_gsf_path( dz->tree, name, dirname, NULL );
dzsave can write zoomify and google maps use "layout" to pick the directory layout, deepzoom / zoomify / google all allowed
2012-10-03 15:26:59 +02:00
break;
case VIPS_FOREIGN_DZ_LAYOUT_GOOGLE:
dzsave will write a zip file but see the TODO in dzsave.c - no way to pick a tile format, always jpeg - no way to pick a container format, always zip - needs testing - examine memuse, how well do we work for large zips - can we make a zip over 4gb?
2014-04-19 16:15:25 +02:00
vips_snprintf( dirname, VIPS_PATH_MAX, "%d", layer->n );
vips_snprintf( dirname2, VIPS_PATH_MAX, "%d", y );
remove suffix from tiles see https://github.com/jcupitt/libvips/issues/129
2014-06-10 15:09:35 +02:00
vips_snprintf( name, VIPS_PATH_MAX,
"%d%s", x, dz->file_suffix );
dzsave can write zoomify and google maps use "layout" to pick the directory layout, deepzoom / zoomify / google all allowed
2012-10-03 15:26:59 +02:00
sort out dzsave paths, auto-enable zip rationalised output paths for dzsave enable zip automatically if output name ends in .zip
2014-05-06 15:28:41 +02:00
out = vips_gsf_path( dz->tree, name, dirname, dirname2, NULL );
dzsave can write zoomify and google maps use "layout" to pick the directory layout, deepzoom / zoomify / google all allowed
2012-10-03 15:26:59 +02:00
break;
add iiif layout to dzsave
2019-11-09 21:40:39 +01:00
case VIPS_FOREIGN_DZ_LAYOUT_IIIF:
Add iiif3 (#2483) * add iiif3 support to dzsave some small things have changed from v2
2021-10-19 15:18:40 +02:00
case VIPS_FOREIGN_DZ_LAYOUT_IIIF3:
add iiif layout to dzsave
2019-11-09 21:40:39 +01:00
{
/* Tiles are addressed in full resolution coordinates, so
* scale up by layer->sub and dz->tile_size
better iiif tile naming We were clipping edge tiles against the size of the scaled up layer. Instead, clip against the size of the top level image. See https://github.com/libvips/libvips/issues/1621
2020-04-24 20:29:35 +02:00
*
* We always clip against the full-sized image, not the scaled
* up layer.
Add iiif3 (#2483) * add iiif3 support to dzsave some small things have changed from v2
2021-10-19 15:18:40 +02:00
*
* This will break for overlap != 0, but hopefully no one will
* ever use that.
add iiif layout to dzsave
2019-11-09 21:40:39 +01:00
*/
int left = x * dz->tile_size * layer->sub;
int top = y * dz->tile_size * layer->sub;
int width = VIPS_MIN( dz->tile_size * layer->sub,
better iiif tile naming We were clipping edge tiles against the size of the scaled up layer. Instead, clip against the size of the top level image. See https://github.com/libvips/libvips/issues/1621
2020-04-24 20:29:35 +02:00
save->ready->Xsize - left );
add iiif layout to dzsave
2019-11-09 21:40:39 +01:00
int height = VIPS_MIN( dz->tile_size * layer->sub,
better iiif tile naming We were clipping edge tiles against the size of the scaled up layer. Instead, clip against the size of the top level image. See https://github.com/libvips/libvips/issues/1621
2020-04-24 20:29:35 +02:00
save->ready->Ysize - top );
add iiif layout to dzsave
2019-11-09 21:40:39 +01:00
vips_snprintf( dirname, VIPS_PATH_MAX, "%d,%d,%d,%d",
left, top, width, height );
Add iiif3 (#2483) * add iiif3 support to dzsave some small things have changed from v2
2021-10-19 15:18:40 +02:00
if( dz->layout == VIPS_FOREIGN_DZ_LAYOUT_IIIF3 ) {
int xsize = VIPS_MIN( dz->tile_size,
layer->width - x * dz->tile_size );
int ysize = VIPS_MIN( dz->tile_size,
layer->height - y * dz->tile_size );
dzsave.c: Remove trailing comma from IIIFv3 folder names (#2590) Fixes #2589. Signed-off-by: Aaron <admin@datahoarder.dev>
2021-12-21 10:43:48 +01:00
vips_snprintf( dirname2, VIPS_PATH_MAX, "%d,%d",
Add iiif3 (#2483) * add iiif3 support to dzsave some small things have changed from v2
2021-10-19 15:18:40 +02:00
xsize, ysize );
}
else {
/* IIIF2 "size" is just real tile width, I think.
*/
int size = VIPS_MIN( dz->tile_size,
layer->width - x * dz->tile_size );
vips_snprintf( dirname2, VIPS_PATH_MAX, "%d,", size );
}
add iiif layout to dzsave
2019-11-09 21:40:39 +01:00
vips_snprintf( name, VIPS_PATH_MAX, "default%s",
dz->file_suffix );
/* "0" is rotation and is always 0.
*/
out = vips_gsf_path( dz->tree,
name, dirname, dirname2, "0", NULL );
}
break;
dzsave can write zoomify and google maps use "layout" to pick the directory layout, deepzoom / zoomify / google all allowed
2012-10-03 15:26:59 +02:00
default:
use g_assert_not_reached() we were using g_assert(0) ... g_assert_not_reached() makes the intention clearer
2016-01-24 14:21:53 +01:00
g_assert_not_reached();
/* Stop compiler warnings.
*/
out = NULL;
dzsave can write zoomify and google maps use "layout" to pick the directory layout, deepzoom / zoomify / google all allowed
2012-10-03 15:26:59 +02:00
}
almost done fixing dzsave see https://github.com/jcupitt/libvips/issues/326 see TODO, problem with bottom edge tiles
2015-09-09 15:53:10 +02:00
#ifdef DEBUG_VERBOSE
printf( "tile_name: writing to %s\n", name );
#endif /*DEBUG_VERBOSE*/
dzsave will write a zip file but see the TODO in dzsave.c - no way to pick a tile format, always jpeg - no way to pick a container format, always zip - needs testing - examine memuse, how well do we work for large zips - can we make a zip over 4gb?
2014-04-19 16:15:25 +02:00
return( out );
dzsave can write zoomify and google maps use "layout" to pick the directory layout, deepzoom / zoomify / google all allowed
2012-10-03 15:26:59 +02:00
}
make skip_blanks an int so you can adjust the threshold
2018-12-20 18:07:07 +01:00
/* Test for tile nearly equal to background colour. In google maps mode, we
* skip blank background tiles.
only save tiles more than 5 from blank looks for max absolute difference
2016-04-26 11:06:59 +02:00
*
make skip_blanks an int so you can adjust the threshold
2018-12-20 18:07:07 +01:00
* Don't use exactly equality since compression artefacts or noise can upset
only save tiles more than 5 from blank looks for max absolute difference
2016-04-26 11:06:59 +02:00
* this.
*/
try skipping blank tiles ... again, but maybe we can be quicker this time
2015-11-24 16:07:23 +01:00
static gboolean
make skip_blanks an int so you can adjust the threshold
2018-12-20 18:07:07 +01:00
tile_equal( VipsImage *image, int threshold, VipsPel * restrict ink )
try skipping blank tiles ... again, but maybe we can be quicker this time
2015-11-24 16:07:23 +01:00
{
const int bytes = VIPS_IMAGE_SIZEOF_PEL( image );
VipsRect rect;
VipsRegion *region;
int x, y, b;
region = vips_region_new( image );
/* We know @image is part of a memory buffer, so this will be quick.
*/
rect.left = 0;
rect.top = 0;
rect.width = image->Xsize;
rect.height = image->Ysize;
if( vips_region_prepare( region, &rect ) ) {
g_object_unref( region );
return( FALSE );
}
for( y = 0; y < image->Ysize; y++ ) {
VipsPel * restrict p = VIPS_REGION_ADDR( region, 0, y );
for( x = 0; x < image->Xsize; x++ ) {
for( b = 0; b < bytes; b++ )
make skip_blanks an int so you can adjust the threshold
2018-12-20 18:07:07 +01:00
if( VIPS_ABS( p[b] - ink[b] ) > threshold ) {
try skipping blank tiles ... again, but maybe we can be quicker this time
2015-11-24 16:07:23 +01:00
g_object_unref( region );
return( FALSE );
}
p += bytes;
}
}
g_object_unref( region );
return( TRUE );
}
add threaded write of deepzoom tiles
2012-07-07 19:42:12 +02:00
static int
strip_work( VipsThreadState *state, void *a )
{
Strip *strip = (Strip *) a;
Layer *layer = strip->layer;
VipsForeignSaveDz *dz = layer->dz;
two small fixes for dzsave fix a tiny leak, fix coloured @background
2015-10-06 14:30:25 +02:00
VipsForeignSave *save = (VipsForeignSave *) dz;
add threaded write of deepzoom tiles
2012-07-07 19:42:12 +02:00
add --centre option to dzsave with --centre turned on, in google maps mode dzsave will centre tiles rather than placing them at the top-left also, make sure blank.png is set to @background
2013-01-21 15:48:14 +01:00
VipsImage *x;
VipsImage *t;
dzsave will write a zip file but see the TODO in dzsave.c - no way to pick a tile format, always jpeg - no way to pick a container format, always zip - needs testing - examine memuse, how well do we work for large zips - can we make a zip over 4gb?
2014-04-19 16:15:25 +02:00
GsfOutput *out;
dzsave can write zoomify and google maps use "layout" to pick the directory layout, deepzoom / zoomify / google all allowed
2012-10-03 15:26:59 +02:00
hacking
2013-02-25 15:54:39 +01:00
#ifdef DEBUG_VERBOSE
easier array arg creation added stdarg array double creator, used the _new as well, small other fixes
2012-11-02 22:45:13 +01:00
printf( "strip_work\n" );
hacking
2013-02-25 15:54:39 +01:00
#endif /*DEBUG_VERBOSE*/
easier array arg creation added stdarg array double creator, used the _new as well, small other fixes
2012-11-02 22:45:13 +01:00
earlier quit of dzsave on kill We were only testing "killed" once for each strip of tiles. Check on each tile as well to make set_kill() faster. see https://github.com/kleisauke/net-vips/issues/179
2022-09-21 12:48:48 +02:00
/* killed is checked by sink_disc, but that's only once per strip, and
* they can be huge. Check per output tile as well.
*/
if( vips_image_iskilled( save->in ) )
return( -1 );
try skipping blank tiles ... again, but maybe we can be quicker this time
2015-11-24 16:07:23 +01:00
/* If we are centering we may be outside the real pixels. Skip in
faster --centre logic for dzsave don't test pixel-by-pixel for background tiles, instead track the rect of real pixels and test for intersection this might help performance scaling on many-core CPUs since we are no longer spinning up a threadgroup for each output tile thanks Kacey
2013-06-19 15:03:39 +02:00
* this case, and the viewer will display blank.png for us.
*/
if( dz->centre ) {
VipsRect tile;
tile.left = state->x;
tile.top = state->y;
tile.width = dz->tile_size;
tile.height = dz->tile_size;
new subimage code for composite done! seems to work, passes all tests
2018-12-07 17:11:58 +01:00
if( !vips_rect_overlapsrect( &tile, &layer->real_pixels ) ) {
faster --centre logic for dzsave don't test pixel-by-pixel for background tiles, instead track the rect of real pixels and test for intersection this might help performance scaling on many-core CPUs since we are no longer spinning up a threadgroup for each output tile thanks Kacey
2013-06-19 15:03:39 +02:00
#ifdef DEBUG_VERBOSE
printf( "strip_work: skipping tile %d x %d\n",
state->x / dz->tile_size,
state->y / dz->tile_size );
#endif /*DEBUG_VERBOSE*/
return( 0 );
}
}
only enable sanity checks in debug mode We had a couple of sanity checks always on. As a result, libvips could get slow for large pipelines. Move them into g_assert().
2019-03-01 14:01:08 +01:00
g_assert( vips_object_sanity( VIPS_OBJECT( strip->image ) ) );
Merge remote-tracking branch 'origin/master' into parallel-read Conflicts: libvips/foreign/openslide2vips.c
2012-10-17 00:46:01 +02:00
add threaded write of deepzoom tiles
2012-07-07 19:42:12 +02:00
/* Extract relative to the strip top-left corner.
*/
Merge remote-tracking branch 'origin/master' into parallel-read Conflicts: libvips/foreign/openslide2vips.c
2012-10-17 00:46:01 +02:00
if( vips_extract_area( strip->image, &x,
add threaded write of deepzoom tiles
2012-07-07 19:42:12 +02:00
state->pos.left, 0,
state->pos.width, state->pos.height, NULL ) )
return( -1 );
use -1 for do-nothing so that users can have skip-only-exact-blanks see https://github.com/libvips/libvips/issues/1185#issuecomment-449115143
2018-12-21 15:08:54 +01:00
if( dz->skip_blanks >= 0 &&
make skip_blanks an int so you can adjust the threshold
2018-12-20 18:07:07 +01:00
tile_equal( x, dz->skip_blanks, dz->ink ) ) {
working! doesn't seem to slow us down significantly ... before: $ time vips dzsave CMU-1.svs x --layout google --background "243 243 243 255" real 1m1.940s user 2m15.004s sys 0m37.092s after: $ time vips dzsave CMU-1.svs x --layout google --background "243 243 243 255" strip_work: skipping blank tile 176 x 67 strip_work: skipping blank tile 21 x 112 real 1m3.503s user 2m16.012s sys 0m40.328s
2015-11-24 18:08:56 +01:00
g_object_unref( x );
try skipping blank tiles ... again, but maybe we can be quicker this time
2015-11-24 16:07:23 +01:00
#ifdef DEBUG_VERBOSE
printf( "strip_work: skipping blank tile %d x %d\n",
state->x / dz->tile_size,
state->y / dz->tile_size );
only save tiles more than 5 from blank looks for max absolute difference
2016-04-26 11:06:59 +02:00
#endif /*DEBUG_VERBOSE*/
try skipping blank tiles ... again, but maybe we can be quicker this time
2015-11-24 16:07:23 +01:00
return( 0 );
}
dzsave can write zoomify and google maps use "layout" to pick the directory layout, deepzoom / zoomify / google all allowed
2012-10-03 15:26:59 +02:00
/* Google tiles need to be padded up to tilesize.
*/
if( dz->layout == VIPS_FOREIGN_DZ_LAYOUT_GOOGLE ) {
add --centre option to dzsave with --centre turned on, in google maps mode dzsave will centre tiles rather than placing them at the top-left also, make sure blank.png is set to @background
2013-01-21 15:48:14 +01:00
if( vips_embed( x, &t, 0, 0, dz->tile_size, dz->tile_size,
two small fixes for dzsave fix a tiny leak, fix coloured @background
2015-10-06 14:30:25 +02:00
"background", save->background,
dzsave can write zoomify and google maps use "layout" to pick the directory layout, deepzoom / zoomify / google all allowed
2012-10-03 15:26:59 +02:00
NULL ) ) {
g_object_unref( x );
return( -1 );
}
g_object_unref( x );
add --centre option to dzsave with --centre turned on, in google maps mode dzsave will centre tiles rather than placing them at the top-left also, make sure blank.png is set to @background
2013-01-21 15:48:14 +01:00
x = t;
}
szi save will write associated images if they have been attached by openslide load
2018-01-28 17:39:13 +01:00
/* we need to single-thread around calls to gsf.
dzsave will write a zip file but see the TODO in dzsave.c - no way to pick a tile format, always jpeg - no way to pick a container format, always zip - needs testing - examine memuse, how well do we work for large zips - can we make a zip over 4gb?
2014-04-19 16:15:25 +02:00
*/
g_mutex_lock( vips__global_lock );
out = tile_name( layer,
new overlap system for dzsave again ... sigh see https://github.com/jcupitt/libvips/issues/512
2016-09-05 19:17:30 +02:00
state->x / dz->tile_step, state->y / dz->tile_step );
dzsave will write a zip file but see the TODO in dzsave.c - no way to pick a tile format, always jpeg - no way to pick a container format, always zip - needs testing - examine memuse, how well do we work for large zips - can we make a zip over 4gb?
2014-04-19 16:15:25 +02:00
szi save will write associated images if they have been attached by openslide load
2018-01-28 17:39:13 +01:00
g_mutex_unlock( vips__global_lock );
better >4gb detection for zip dzsave older libgsfs can't save zip64 and will fail silently for very large output trees ... improve the slightly sketchy >4gb detection in dzsave this a a version of the patch in https://github.com/jcupitt/libvips/pull/462
2016-11-11 17:24:18 +01:00
szi save will write associated images if they have been attached by openslide load
2018-01-28 17:39:13 +01:00
if( write_image( dz, out, x, dz->suffix ) ) {
fix refleak in dzsave
2019-12-21 05:46:16 +01:00
g_object_unref( out );
szi save will write associated images if they have been attached by openslide load
2018-01-28 17:39:13 +01:00
g_object_unref( x );
better >4gb detection for zip dzsave older libgsfs can't save zip64 and will fail silently for very large output trees ... improve the slightly sketchy >4gb detection in dzsave this a a version of the patch in https://github.com/jcupitt/libvips/pull/462
2016-11-11 17:24:18 +01:00
szi save will write associated images if they have been attached by openslide load
2018-01-28 17:39:13 +01:00
return( -1 );
stop dzsave zip write at 4gb see https://github.com/jcupitt/libvips/issues/138 not great, but better than producing a corrupt file
2014-06-25 15:34:13 +02:00
}
szi save will write associated images if they have been attached by openslide load
2018-01-28 17:39:13 +01:00
g_object_unref( out );
fix refleak in dzsave
2019-12-21 05:46:16 +01:00
g_object_unref( x );
stop dzsave zip write at 4gb see https://github.com/jcupitt/libvips/issues/138 not great, but better than producing a corrupt file
2014-06-25 15:34:13 +02:00
hacking
2013-02-25 15:54:39 +01:00
#ifdef DEBUG_VERBOSE
easier array arg creation added stdarg array double creator, used the _new as well, small other fixes
2012-11-02 22:45:13 +01:00
printf( "strip_work: success\n" );
hacking
2013-02-25 15:54:39 +01:00
#endif /*DEBUG_VERBOSE*/
easier array arg creation added stdarg array double creator, used the _new as well, small other fixes
2012-11-02 22:45:13 +01:00
add threaded write of deepzoom tiles
2012-07-07 19:42:12 +02:00
return( 0 );
}
/* Write a line of tiles with a threadpool.
*/
static int
strip_save( Layer *layer )
{
Strip strip;
sdzsave fix if dzsave was running a scanline at a time and height was odd, the final set of tiles were not being written -- thanks Martin
2012-10-01 15:08:59 +02:00
#ifdef DEBUG
printf( "strip_save: n = %d, y = %d\n", layer->n, layer->y );
#endif /*DEBUG*/
add threaded write of deepzoom tiles
2012-07-07 19:42:12 +02:00
strip_init( &strip, layer );
if( vips_threadpool_run( strip.image,
vips_thread_state_new, strip_allocate, strip_work, NULL,
&strip ) ) {
strip_free( &strip );
return( -1 );
}
strip_free( &strip );
started hacking in dzsave
2012-03-21 15:53:53 +01:00
almost done fixing dzsave see https://github.com/jcupitt/libvips/issues/326 see TODO, problem with bottom edge tiles
2015-09-09 15:53:10 +02:00
#ifdef DEBUG
printf( "strip_save: success\n" );
#endif /*DEBUG*/
started hacking in dzsave
2012-03-21 15:53:53 +01:00
return( 0 );
}
fixes to deep zoom write - shrink down to a 1x1 pixel tile, even for very long and thin images - round image size up on shrink - write a .dzi file with the pyramid params - default tile size and overlap now matches the openslide writer thanks to Benjamin Gilbert for pointing out the errors
2012-08-06 16:14:59 +02:00
/* A strip has filled, but the rightmost column and the bottom-most row may
sdzsave fix if dzsave was running a scanline at a time and height was odd, the final set of tiles were not being written -- thanks Martin
2012-10-01 15:08:59 +02:00
* not have been if we've rounded the size up.
fixes to deep zoom write - shrink down to a 1x1 pixel tile, even for very long and thin images - round image size up on shrink - write a .dzi file with the pyramid params - default tile size and overlap now matches the openslide writer thanks to Benjamin Gilbert for pointing out the errors
2012-08-06 16:14:59 +02:00
*
sdzsave fix if dzsave was running a scanline at a time and height was odd, the final set of tiles were not being written -- thanks Martin
2012-10-01 15:08:59 +02:00
* Fill them, if necessary, by copying the previous row/column.
fixes to deep zoom write - shrink down to a 1x1 pixel tile, even for very long and thin images - round image size up on shrink - write a .dzi file with the pyramid params - default tile size and overlap now matches the openslide writer thanks to Benjamin Gilbert for pointing out the errors
2012-08-06 16:14:59 +02:00
*/
static void
layer_generate_extras( Layer *layer )
{
VipsRegion *strip = layer->strip;
/* We only work for full-width strips.
*/
g_assert( strip->valid.width == layer->image->Xsize );
if( layer->width < layer->image->Xsize ) {
int ps = VIPS_IMAGE_SIZEOF_PEL( strip->im );
int b, y;
/* Need to add a right-most column.
*/
for( y = 0; y < strip->valid.height; y++ ) {
VipsPel *p = VIPS_REGION_ADDR( strip,
layer->width - 1, strip->valid.top + y );
VipsPel *q = p + ps;
for( b = 0; b < ps; b++ )
q[b] = p[b];
}
}
if( layer->height < layer->image->Ysize ) {
VipsRect last;
/* The last two lines of the image.
*/
last.left = 0;
last.top = layer->image->Ysize - 2;
last.width = layer->image->Xsize;
last.height = 2;
/* Do we have them both? Fill the last with the next-to-last.
*/
vips_rect_intersectrect( &last, &strip->valid, &last );
if( last.height == 2 ) {
last.height = 1;
vips_region_copy( strip, strip, &last,
0, last.top + 1 );
}
}
}
more dzsave overlap fixups works with odd-sized tiles, overlaps handled correctly
2012-07-07 14:58:50 +02:00
static int strip_arrived( Layer *layer );
fix another dzsave corner case better padding calculation fixes yet another corner case for dzsave, thanks Martin
2013-02-26 13:32:01 +01:00
/* Shrink what pixels we can from this strip into the layer below. If the
* strip below fills, recurse.
started hacking in dzsave
2012-03-21 15:53:53 +01:00
*/
static int
more dzsave overlap fixups works with odd-sized tiles, overlaps handled correctly
2012-07-07 14:58:50 +02:00
strip_shrink( Layer *layer )
started hacking in dzsave
2012-03-21 15:53:53 +01:00
{
hackety hack
2012-03-23 15:57:41 +01:00
Layer *below = layer->below;
more dzsave overlap fixups works with odd-sized tiles, overlaps handled correctly
2012-07-07 14:58:50 +02:00
VipsRegion *from = layer->strip;
VipsRegion *to = below->strip;
dzsave now uses the specified region shrink method
2018-03-28 04:43:54 +02:00
VipsForeignSaveDz *dz = layer->dz;
VipsRegionShrink region_shrink = dz->region_shrink;
started hacking in dzsave
2012-03-21 15:53:53 +01:00
hackety hack
2012-03-23 15:57:41 +01:00
VipsRect target;
more dzsave overlap fixups works with odd-sized tiles, overlaps handled correctly
2012-07-07 14:58:50 +02:00
VipsRect source;
started hacking in dzsave
2012-03-21 15:53:53 +01:00
almost done fixing dzsave see https://github.com/jcupitt/libvips/issues/326 see TODO, problem with bottom edge tiles
2015-09-09 15:53:10 +02:00
#ifdef DEBUG
printf( "strip_shrink: %d lines in layer %d to layer %d\n",
from->valid.height, layer->n, below->n );
formatting in dzsave.c
2019-11-19 09:54:23 +01:00
#endif /*DEBUG*/
almost done fixing dzsave see https://github.com/jcupitt/libvips/issues/326 see TODO, problem with bottom edge tiles
2015-09-09 15:53:10 +02:00
fixes to deep zoom write - shrink down to a 1x1 pixel tile, even for very long and thin images - round image size up on shrink - write a .dzi file with the pyramid params - default tile size and overlap now matches the openslide writer thanks to Benjamin Gilbert for pointing out the errors
2012-08-06 16:14:59 +02:00
/* We may have an extra column of pixels on the right or
* bottom that need filling: generate them.
*/
layer_generate_extras( layer );
more dzsave overlap fixups works with odd-sized tiles, overlaps handled correctly
2012-07-07 14:58:50 +02:00
/* Our pixels might cross a strip boundary in the layer below, so we
* have to write repeatedly until we run out of pixels.
started hacking in dzsave
2012-03-21 15:53:53 +01:00
*/
more dzsave overlap fixups works with odd-sized tiles, overlaps handled correctly
2012-07-07 14:58:50 +02:00
for(;;) {
/* The pixels the layer below needs.
*/
target.left = 0;
target.top = below->write_y;
target.width = below->image->Xsize;
target.height = to->valid.height;
vips_rect_intersectrect( &target, &to->valid, &target );
started hacking in dzsave
2012-03-21 15:53:53 +01:00
more dzsave overlap fixups works with odd-sized tiles, overlaps handled correctly
2012-07-07 14:58:50 +02:00
/* Those pixels need this area of this layer.
*/
source.left = target.left * 2;
source.top = target.top * 2;
source.width = target.width * 2;
source.height = target.height * 2;
fix odd tile height + no overlap dzsave with zero overlap and odd height was broken also, fix cache init
2012-07-08 11:37:35 +02:00
/* Of which we have these available.
*/
more dzsave overlap fixups works with odd-sized tiles, overlaps handled correctly
2012-07-07 14:58:50 +02:00
vips_rect_intersectrect( &source, &from->valid, &source );
/* So these are the pixels in the layer below we can provide.
started hacking in dzsave
2012-03-21 15:53:53 +01:00
*/
more dzsave overlap fixups works with odd-sized tiles, overlaps handled correctly
2012-07-07 14:58:50 +02:00
target.left = source.left / 2;
target.top = source.top / 2;
target.width = source.width / 2;
target.height = source.height / 2;
fix odd tile height + no overlap dzsave with zero overlap and odd height was broken also, fix cache init
2012-07-08 11:37:35 +02:00
/* None? All done.
more dzsave overlap fixups works with odd-sized tiles, overlaps handled correctly
2012-07-07 14:58:50 +02:00
*/
Adds MEAN and MEDIAN methods for region shrink
2018-03-27 12:15:48 +02:00
if( vips_rect_isempty( &target ) )
more dzsave overlap fixups works with odd-sized tiles, overlaps handled correctly
2012-07-07 14:58:50 +02:00
break;
add a compat stub for vips_region_shrink() to keep ABI compat with 8.6
2018-08-29 10:50:35 +02:00
(void) vips_region_shrink_method( from, to,
&target, region_shrink );
more dzsave overlap fixups works with odd-sized tiles, overlaps handled correctly
2012-07-07 14:58:50 +02:00
below->write_y += target.height;
started hacking in dzsave
2012-03-21 15:53:53 +01:00
fix another dzsave corner case better padding calculation fixes yet another corner case for dzsave, thanks Martin
2013-02-26 13:32:01 +01:00
/* If we've filled the strip below, let it know.
fixes to deep zoom write - shrink down to a 1x1 pixel tile, even for very long and thin images - round image size up on shrink - write a .dzi file with the pyramid params - default tile size and overlap now matches the openslide writer thanks to Benjamin Gilbert for pointing out the errors
2012-08-06 16:14:59 +02:00
* We can either fill the region, if it's somewhere half-way
* down the image, or, if it's at the bottom, get to the last
fix another dzsave corner case better padding calculation fixes yet another corner case for dzsave, thanks Martin
2013-02-26 13:32:01 +01:00
* real line of pixels.
started hacking in dzsave
2012-03-21 15:53:53 +01:00
*/
fixes to deep zoom write - shrink down to a 1x1 pixel tile, even for very long and thin images - round image size up on shrink - write a .dzi file with the pyramid params - default tile size and overlap now matches the openslide writer thanks to Benjamin Gilbert for pointing out the errors
2012-08-06 16:14:59 +02:00
if( below->write_y == VIPS_RECT_BOTTOM( &to->valid ) ||
below->write_y == below->height ) {
if( strip_arrived( below ) )
return( -1 );
}
started hacking in dzsave
2012-03-21 15:53:53 +01:00
}
more dzsave overlap fixups works with odd-sized tiles, overlaps handled correctly
2012-07-07 14:58:50 +02:00
return( 0 );
}
fix another dzsave corner case better padding calculation fixes yet another corner case for dzsave, thanks Martin
2013-02-26 13:32:01 +01:00
/* A new strip has arrived! The strip has enough pixels in to write a line of
* tiles.
more dzsave overlap fixups works with odd-sized tiles, overlaps handled correctly
2012-07-07 14:58:50 +02:00
*
* - write a line of tiles
* - shrink what we can to the layer below
almost done fixing dzsave see https://github.com/jcupitt/libvips/issues/326 see TODO, problem with bottom edge tiles
2015-09-09 15:53:10 +02:00
* - move our strip down by the tile step
fix another dzsave corner case better padding calculation fixes yet another corner case for dzsave, thanks Martin
2013-02-26 13:32:01 +01:00
* - copy the overlap with the previous strip
more dzsave overlap fixups works with odd-sized tiles, overlaps handled correctly
2012-07-07 14:58:50 +02:00
*/
static int
strip_arrived( Layer *layer )
{
VipsForeignSaveDz *dz = layer->dz;
fix dzsave/vips2tiff for some image sizes we were not clipping new strip size against image size, see https://github.com/jcupitt/libvips/issues/300
2015-05-24 18:53:05 +02:00
more dzsave overlap fixups works with odd-sized tiles, overlaps handled correctly
2012-07-07 14:58:50 +02:00
VipsRect new_strip;
VipsRect overlap;
fix dzsave/vips2tiff for some image sizes we were not clipping new strip size against image size, see https://github.com/jcupitt/libvips/issues/300
2015-05-24 18:53:05 +02:00
VipsRect image_area;
more dzsave overlap fixups works with odd-sized tiles, overlaps handled correctly
2012-07-07 14:58:50 +02:00
almost done fixing dzsave see https://github.com/jcupitt/libvips/issues/326 see TODO, problem with bottom edge tiles
2015-09-09 15:53:10 +02:00
#ifdef DEBUG
printf( "strip_arrived: layer %d, strip at %d, height %d\n",
layer->n, layer->y, layer->strip->valid.height );
formatting in dzsave.c
2019-11-19 09:54:23 +01:00
#endif /*DEBUG*/
almost done fixing dzsave see https://github.com/jcupitt/libvips/issues/326 see TODO, problem with bottom edge tiles
2015-09-09 15:53:10 +02:00
more dzsave overlap fixups works with odd-sized tiles, overlaps handled correctly
2012-07-07 14:58:50 +02:00
if( strip_save( layer ) )
return( -1 );
if( layer->below &&
strip_shrink( layer ) )
return( -1 );
fix another dzsave corner case better padding calculation fixes yet another corner case for dzsave, thanks Martin
2013-02-26 13:32:01 +01:00
/* Position our strip down the image.
*
* Expand the strip if necessary to make sure we have an even
* number of lines.
hackety hack
2012-03-23 15:57:41 +01:00
*/
new overlap system for dzsave again ... sigh see https://github.com/jcupitt/libvips/issues/512
2016-09-05 19:17:30 +02:00
layer->y += dz->tile_step;
hackety hack
2012-03-23 15:57:41 +01:00
new_strip.left = 0;
new overlap system for dzsave again ... sigh see https://github.com/jcupitt/libvips/issues/512
2016-09-05 19:17:30 +02:00
new_strip.top = layer->y - dz->tile_margin;
hackety hack
2012-03-23 15:57:41 +01:00
new_strip.width = layer->image->Xsize;
new overlap system for dzsave again ... sigh see https://github.com/jcupitt/libvips/issues/512
2016-09-05 19:17:30 +02:00
new_strip.height = dz->tile_size + 2 * dz->tile_margin;
fix dzsave/vips2tiff for some image sizes we were not clipping new strip size against image size, see https://github.com/jcupitt/libvips/issues/300
2015-05-24 18:53:05 +02:00
image_area.left = 0;
image_area.top = 0;
image_area.width = layer->image->Xsize;
image_area.height = layer->image->Ysize;
vips_rect_intersectrect( &new_strip, &image_area, &new_strip );
fix another dzsave corner case better padding calculation fixes yet another corner case for dzsave, thanks Martin
2013-02-26 13:32:01 +01:00
if( (new_strip.height & 1) == 1 )
new_strip.height += 1;
works!
2012-03-26 23:08:59 +02:00
fix an obscure dzsave bug in some cases dzsave could fail to write all the pyramid layers out the problem was if a layer was odd-sized and needed rounding up so it could be shrunk from, and if the tile size was such that the final strip was sized exactly to fit the last line of pixels that would come into this layer, the final line of pixels could never be interpolated thanks to Martin for the report
2013-01-17 14:14:47 +01:00
/* We may exactly hit the bottom of the real image (ie. before borders
* have been possibly expanded by 1 pixel). In this case, we'll not
* be able to do the expansion in layer_generate_extras(), since the
* region won't be large enough, and we'll not get another chance
* since this is the bottom.
*
* Add another scanline if this has happened.
*/
if( VIPS_RECT_BOTTOM( &new_strip ) == layer->height )
new_strip.height = layer->image->Ysize - new_strip.top;
more dzsave overlap fixups works with odd-sized tiles, overlaps handled correctly
2012-07-07 14:58:50 +02:00
/* What pixels that we will need do we already have? Save them in
* overlap.
*/
works!
2012-03-26 23:08:59 +02:00
vips_rect_intersectrect( &new_strip, &layer->strip->valid, &overlap );
if( !vips_rect_isempty( &overlap ) ) {
if( vips_region_buffer( layer->copy, &overlap ) )
return( -1 );
vips_region_copy( layer->strip, layer->copy,
&overlap, overlap.left, overlap.top );
}
fix dzsave/vips2tiff for some image sizes we were not clipping new strip size against image size, see https://github.com/jcupitt/libvips/issues/300
2015-05-24 18:53:05 +02:00
if( !vips_rect_isempty( &new_strip ) ) {
if( vips_region_buffer( layer->strip, &new_strip ) )
return( -1 );
hackety hack
2012-03-23 15:57:41 +01:00
fix dzsave/vips2tiff for some image sizes we were not clipping new strip size against image size, see https://github.com/jcupitt/libvips/issues/300
2015-05-24 18:53:05 +02:00
/* And copy back again.
*/
if( !vips_rect_isempty( &overlap ) )
vips_region_copy( layer->copy, layer->strip,
&overlap, overlap.left, overlap.top );
}
works!
2012-03-26 23:08:59 +02:00
started hacking in dzsave
2012-03-21 15:53:53 +01:00
return( 0 );
}
dzsave outputs extra tiles along right and bottom the deepzoom spec wants tiles to be written along the right and bottom, even of those tiles contain no new pixels this patch seems to work, but needs testing see: https://github.com/jcupitt/libvips/issues/795
2017-11-24 16:54:22 +01:00
/* The image has been completely written. Flush any strips which might have
* overlaps in.
*/
static int
strip_flush( Layer *layer )
{
if( layer->y < layer->height )
if( strip_save( layer ) )
return( -1 );
if( layer->below )
if( strip_flush( layer->below ) )
return( -1 );
return( 0 );
}
more dzsave overlap fixups works with odd-sized tiles, overlaps handled correctly
2012-07-07 14:58:50 +02:00
/* Another strip of image pixels from vips_sink_disc(). Write into the top
* pyramid layer.
started hacking in dzsave
2012-03-21 15:53:53 +01:00
*/
static int
hackery
2012-03-23 10:21:28 +01:00
pyramid_strip( VipsRegion *region, VipsRect *area, void *a )
started hacking in dzsave
2012-03-21 15:53:53 +01:00
{
VipsForeignSaveDz *dz = (VipsForeignSaveDz *) a;
more dzsave overlap fixups works with odd-sized tiles, overlaps handled correctly
2012-07-07 14:58:50 +02:00
Layer *layer = dz->layer;
hackery
2012-03-23 10:21:28 +01:00
sdzsave fix if dzsave was running a scanline at a time and height was odd, the final set of tiles were not being written -- thanks Martin
2012-10-01 15:08:59 +02:00
#ifdef DEBUG
printf( "pyramid_strip: strip at %d, height %d\n",
area->top, area->height );
formatting in dzsave.c
2019-11-19 09:54:23 +01:00
#endif /*DEBUG*/
sdzsave fix if dzsave was running a scanline at a time and height was odd, the final set of tiles were not being written -- thanks Martin
2012-10-01 15:08:59 +02:00
more dzsave overlap fixups works with odd-sized tiles, overlaps handled correctly
2012-07-07 14:58:50 +02:00
for(;;) {
sdzsave fix if dzsave was running a scanline at a time and height was odd, the final set of tiles were not being written -- thanks Martin
2012-10-01 15:08:59 +02:00
VipsRect *to = &layer->strip->valid;
more dzsave overlap fixups works with odd-sized tiles, overlaps handled correctly
2012-07-07 14:58:50 +02:00
VipsRect target;
hackery
2012-03-23 10:21:28 +01:00
more dzsave overlap fixups works with odd-sized tiles, overlaps handled correctly
2012-07-07 14:58:50 +02:00
/* The bit of strip that needs filling.
hackety hack
2012-03-23 15:57:41 +01:00
*/
more dzsave overlap fixups works with odd-sized tiles, overlaps handled correctly
2012-07-07 14:58:50 +02:00
target.left = 0;
target.top = layer->write_y;
target.width = layer->image->Xsize;
sdzsave fix if dzsave was running a scanline at a time and height was odd, the final set of tiles were not being written -- thanks Martin
2012-10-01 15:08:59 +02:00
target.height = to->height;
vips_rect_intersectrect( &target, to, &target );
more dzsave overlap fixups works with odd-sized tiles, overlaps handled correctly
2012-07-07 14:58:50 +02:00
/* Clip against what we have available.
*/
vips_rect_intersectrect( &target, area, &target );
started hacking in dzsave
2012-03-21 15:53:53 +01:00
dzsave outputs extra tiles along right and bottom the deepzoom spec wants tiles to be written along the right and bottom, even of those tiles contain no new pixels this patch seems to work, but needs testing see: https://github.com/jcupitt/libvips/issues/795
2017-11-24 16:54:22 +01:00
/* Have we written all the pixels we were given? We are done.
hackery
2012-03-23 10:21:28 +01:00
*/
more dzsave overlap fixups works with odd-sized tiles, overlaps handled correctly
2012-07-07 14:58:50 +02:00
if( vips_rect_isempty( &target ) )
break;
hackery
2012-03-23 10:21:28 +01:00
more dzsave overlap fixups works with odd-sized tiles, overlaps handled correctly
2012-07-07 14:58:50 +02:00
/* And copy those pixels in.
*
* FIXME: If the strip fits inside the region we've just
fix odd tile height + no overlap dzsave with zero overlap and odd height was broken also, fix cache init
2012-07-08 11:37:35 +02:00
* received, we could skip the copy. Will this happen very
* often? Unclear.
more dzsave overlap fixups works with odd-sized tiles, overlaps handled correctly
2012-07-07 14:58:50 +02:00
*/
vips_region_copy( region, layer->strip,
&target, target.left, target.top );
fixes to deep zoom write - shrink down to a 1x1 pixel tile, even for very long and thin images - round image size up on shrink - write a .dzi file with the pyramid params - default tile size and overlap now matches the openslide writer thanks to Benjamin Gilbert for pointing out the errors
2012-08-06 16:14:59 +02:00
more dzsave overlap fixups works with odd-sized tiles, overlaps handled correctly
2012-07-07 14:58:50 +02:00
layer->write_y += target.height;
fix another dzsave corner case better padding calculation fixes yet another corner case for dzsave, thanks Martin
2013-02-26 13:32:01 +01:00
/* We can either fill the strip, if it's somewhere half-way
sdzsave fix if dzsave was running a scanline at a time and height was odd, the final set of tiles were not being written -- thanks Martin
2012-10-01 15:08:59 +02:00
* down the image, or, if it's at the bottom, get to the last
fix another dzsave corner case better padding calculation fixes yet another corner case for dzsave, thanks Martin
2013-02-26 13:32:01 +01:00
* real line of pixels.
more dzsave overlap fixups works with odd-sized tiles, overlaps handled correctly
2012-07-07 14:58:50 +02:00
*/
sdzsave fix if dzsave was running a scanline at a time and height was odd, the final set of tiles were not being written -- thanks Martin
2012-10-01 15:08:59 +02:00
if( layer->write_y == VIPS_RECT_BOTTOM( to ) ||
layer->write_y == layer->height ) {
stop dzsave zip write at 4gb see https://github.com/jcupitt/libvips/issues/138 not great, but better than producing a corrupt file
2014-06-25 15:34:13 +02:00
if( strip_arrived( layer ) )
sdzsave fix if dzsave was running a scanline at a time and height was odd, the final set of tiles were not being written -- thanks Martin
2012-10-01 15:08:59 +02:00
return( -1 );
}
hackery
2012-03-23 10:21:28 +01:00
}
started hacking in dzsave
2012-03-21 15:53:53 +01:00
dzsave outputs extra tiles along right and bottom the deepzoom spec wants tiles to be written along the right and bottom, even of those tiles contain no new pixels this patch seems to work, but needs testing see: https://github.com/jcupitt/libvips/issues/795
2017-11-24 16:54:22 +01:00
/* If we've reached the bottom of the image, we won't get called again.
*
* However, there may be some unwritten pixels in the pyramid still!
* Suppose a layer is exactly a multiple of tile_step in height.
* When we finished that last strip, we will have copied the last few
* lines of overlap over into the top of the next row. Deepzoom says we
* must flush these half-written strips to the output.
*/
if( layer->write_y == layer->height ) {
#ifdef DEBUG
printf( "pyramid_strip: flushing ..\n" );
formatting in dzsave.c
2019-11-19 09:54:23 +01:00
#endif /*DEBUG*/
dzsave outputs extra tiles along right and bottom the deepzoom spec wants tiles to be written along the right and bottom, even of those tiles contain no new pixels this patch seems to work, but needs testing see: https://github.com/jcupitt/libvips/issues/795
2017-11-24 16:54:22 +01:00
if( strip_flush( layer ) )
return( -1 );
}
hackery
2012-03-23 10:21:28 +01:00
return( 0 );
started hacking in dzsave
2012-03-21 15:53:53 +01:00
}
static int
vips_foreign_save_dz_build( VipsObject *object )
{
VipsForeignSave *save = (VipsForeignSave *) object;
VipsForeignSaveDz *dz = (VipsForeignSaveDz *) object;
use a better name for the dzsave temp dir see https://github.com/jcupitt/libvips/issues/242 dzsave with deepzoom layout and fs container writes a temp dir called dzsave-temp-324857 or similar, and makes sure that this directory does not exist after write, the foo_files and foo.dzi files are picked out to the current dir and the temp removed
2015-02-22 16:59:31 +01:00
VipsObjectClass *class = VIPS_OBJECT_GET_CLASS( dz );
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
faster --centre logic for dzsave don't test pixel-by-pixel for background tiles, instead track the rect of real pixels and test for intersection this might help performance scaling on many-core CPUs since we are no longer spinning up a threadgroup for each output tile thanks Kacey
2013-06-19 15:03:39 +02:00
VipsRect real_pixels;
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
char *p;
started hacking in dzsave
2012-03-21 15:53:53 +01:00
add iiif layout to dzsave
2019-11-09 21:40:39 +01:00
/* Google, zoomify and iiif default to zero overlap, ".jpg".
dzsave can write zoomify and google maps use "layout" to pick the directory layout, deepzoom / zoomify / google all allowed
2012-10-03 15:26:59 +02:00
*/
if( dz->layout == VIPS_FOREIGN_DZ_LAYOUT_ZOOMIFY ||
add iiif layout to dzsave
2019-11-09 21:40:39 +01:00
dz->layout == VIPS_FOREIGN_DZ_LAYOUT_GOOGLE ||
Add iiif3 (#2483) * add iiif3 support to dzsave some small things have changed from v2
2021-10-19 15:18:40 +02:00
dz->layout == VIPS_FOREIGN_DZ_LAYOUT_IIIF ||
dz->layout == VIPS_FOREIGN_DZ_LAYOUT_IIIF3 ) {
rename some _get funcs all the _get_ funcs that returned a bool have become _isthing() functions, eg. vips_object_argument_get_assigned() -> vips_object_argument_isset()
2012-11-09 14:08:59 +01:00
if( !vips_object_argument_isset( object, "overlap" ) )
dzsave can write zoomify and google maps use "layout" to pick the directory layout, deepzoom / zoomify / google all allowed
2012-10-03 15:26:59 +02:00
dz->overlap = 0;
rename some _get funcs all the _get_ funcs that returned a bool have become _isthing() functions, eg. vips_object_argument_get_assigned() -> vips_object_argument_isset()
2012-11-09 14:08:59 +01:00
if( !vips_object_argument_isset( object, "suffix" ) )
dzsave can write zoomify and google maps use "layout" to pick the directory layout, deepzoom / zoomify / google all allowed
2012-10-03 15:26:59 +02:00
VIPS_SETSTR( dz->suffix, ".jpg" );
}
fix dzsave tile-width default was wrong for google / zoomify layouts revise tests too
2015-12-30 12:23:33 +01:00
/* Google and zoomify default to 256 pixel tiles.
*/
if( dz->layout == VIPS_FOREIGN_DZ_LAYOUT_ZOOMIFY ||
dz->layout == VIPS_FOREIGN_DZ_LAYOUT_GOOGLE ) {
if( !vips_object_argument_isset( object, "tile_size" ) )
dz->tile_size = 256;
}
Add iiif3 (#2483) * add iiif3 support to dzsave some small things have changed from v2
2021-10-19 15:18:40 +02:00
/* Some iiif writers default to 256, some to 512. We pick 512.
add iiif layout to dzsave
2019-11-09 21:40:39 +01:00
*/
Add iiif3 (#2483) * add iiif3 support to dzsave some small things have changed from v2
2021-10-19 15:18:40 +02:00
if( dz->layout == VIPS_FOREIGN_DZ_LAYOUT_IIIF ||
dz->layout == VIPS_FOREIGN_DZ_LAYOUT_IIIF3 ) {
add iiif layout to dzsave
2019-11-09 21:40:39 +01:00
if( !vips_object_argument_isset( object, "tile_size" ) )
dz->tile_size = 512;
}
make skip_blanks an int so you can adjust the threshold
2018-12-20 18:07:07 +01:00
/* skip_blanks defaults to 5 in google mode.
add skip-blanks option to dzsave Skip-blanks was a feature of google layout. This patch makes it into an option you can control which defaults on in gm mode. See https://github.com/libvips/libvips/issues/1185
2018-12-19 15:52:33 +01:00
*/
if( dz->layout == VIPS_FOREIGN_DZ_LAYOUT_GOOGLE &&
!vips_object_argument_isset( object, "skip_blanks" ) )
make skip_blanks an int so you can adjust the threshold
2018-12-20 18:07:07 +01:00
dz->skip_blanks = 5;
add skip-blanks option to dzsave Skip-blanks was a feature of google layout. This patch makes it into an option you can control which defaults on in gm mode. See https://github.com/libvips/libvips/issues/1185
2018-12-19 15:52:33 +01:00
new overlap system for dzsave again ... sigh see https://github.com/jcupitt/libvips/issues/512
2016-09-05 19:17:30 +02:00
/* Our tile layout.
*/
if( dz->layout == VIPS_FOREIGN_DZ_LAYOUT_DZ ) {
dz->tile_margin = dz->overlap;
dz->tile_step = dz->tile_size;
}
else {
dz->tile_margin = 0;
dz->tile_step = dz->tile_size - dz->overlap;
}
if( dz->tile_step <= 0 ) {
vips_error( "dzsave", "%s", _( "overlap too large" ) );
return( -1 );
}
two small fixes for dzsave fix a tiny leak, fix coloured @background
2015-10-06 14:30:25 +02:00
/* Default to white background. vips_foreign_save_init() defaults to
* black.
make VipsArea threadsafe google maps output needs threaded create / destroy of VipsArea (used to hold the background colour). We need a lock on VipsArea::count. Fixes https://github.com/jcupitt/libvips/issues/46 Thanks fuho
2012-12-19 10:58:35 +01:00
*/
add skip-blanks option to dzsave Skip-blanks was a feature of google layout. This patch makes it into an option you can control which defaults on in gm mode. See https://github.com/libvips/libvips/issues/1185
2018-12-19 15:52:33 +01:00
if( !vips_object_argument_isset( object, "background" ) ) {
make VipsArea threadsafe google maps output needs threaded create / destroy of VipsArea (used to hold the background colour). We need a lock on VipsArea::count. Fixes https://github.com/jcupitt/libvips/issues/46 Thanks fuho
2012-12-19 10:58:35 +01:00
VipsArrayDouble *background;
fix a few more problems with input args a few other operators were modifying input args and messing up caching
2017-05-29 19:58:24 +02:00
/* Using g_object_set() to set an input param in build will
* change the hash and confuse caching, but we don't cache
* savers, so it's fine.
*/
make VipsArea threadsafe google maps output needs threaded create / destroy of VipsArea (used to hold the background colour). We need a lock on VipsArea::count. Fixes https://github.com/jcupitt/libvips/issues/46 Thanks fuho
2012-12-19 10:58:35 +01:00
background = vips_array_double_newv( 1, 255.0 );
g_object_set( object, "background", background, NULL );
vips array types are more binding-friendly this all works now: ```python from gi.repository import Vips a = Vips.ArrayDouble.new([1,2,3]) a.get() a = Vips.ArrayInt.new([1,2,3]) a.get() a = Vips.ArrayImage.new([c, d, e]) a.get() ```
2014-08-31 11:41:53 +02:00
vips_area_unref( VIPS_AREA( background ) );
make VipsArea threadsafe google maps output needs threaded create / destroy of VipsArea (used to hold the background colour). We need a lock on VipsArea::count. Fixes https://github.com/jcupitt/libvips/issues/46 Thanks fuho
2012-12-19 10:58:35 +01:00
}
add @depth option to dzsave so you can ask for a 1 layer pyramid
2012-11-01 19:11:46 +01:00
/* DeepZoom stops at 1x1 pixels, others when the image fits within a
* tile.
*/
if( dz->layout == VIPS_FOREIGN_DZ_LAYOUT_DZ ) {
rename some _get funcs all the _get_ funcs that returned a bool have become _isthing() functions, eg. vips_object_argument_get_assigned() -> vips_object_argument_isset()
2012-11-09 14:08:59 +01:00
if( !vips_object_argument_isset( object, "depth" ) )
rename _!PIXEL etc. to help bindings ruby gobject-introspection hates numbers at the start of _ separated names
2015-01-07 15:22:13 +01:00
dz->depth = VIPS_FOREIGN_DZ_DEPTH_ONEPIXEL;
add @depth option to dzsave so you can ask for a 1 layer pyramid
2012-11-01 19:11:46 +01:00
}
else
rename some _get funcs all the _get_ funcs that returned a bool have become _isthing() functions, eg. vips_object_argument_get_assigned() -> vips_object_argument_isset()
2012-11-09 14:08:59 +01:00
if( !vips_object_argument_isset( object, "depth" ) )
rename _!PIXEL etc. to help bindings ruby gobject-introspection hates numbers at the start of _ separated names
2015-01-07 15:22:13 +01:00
dz->depth = VIPS_FOREIGN_DZ_DEPTH_ONETILE;
add @depth option to dzsave so you can ask for a 1 layer pyramid
2012-11-01 19:11:46 +01:00
if( VIPS_OBJECT_CLASS( vips_foreign_save_dz_parent_class )->
build( object ) )
return( -1 );
add --angle option to dzsave
2013-05-29 15:06:23 +02:00
/* Optional rotate.
*/
{
VipsImage *z;
if( vips_rot( save->ready, &z, dz->angle, NULL ) )
return( -1 );
VIPS_UNREF( save->ready );
save->ready = z;
}
add skip-blanks option to dzsave Skip-blanks was a feature of google layout. This patch makes it into an option you can control which defaults on in gm mode. See https://github.com/libvips/libvips/issues/1185
2018-12-19 15:52:33 +01:00
/* We use ink to check for blank tiles.
working! doesn't seem to slow us down significantly ... before: $ time vips dzsave CMU-1.svs x --layout google --background "243 243 243 255" real 1m1.940s user 2m15.004s sys 0m37.092s after: $ time vips dzsave CMU-1.svs x --layout google --background "243 243 243 255" strip_work: skipping blank tile 176 x 67 strip_work: skipping blank tile 21 x 112 real 1m3.503s user 2m16.012s sys 0m40.328s
2015-11-24 18:08:56 +01:00
*/
use -1 for do-nothing so that users can have skip-only-exact-blanks see https://github.com/libvips/libvips/issues/1185#issuecomment-449115143
2018-12-21 15:08:54 +01:00
if( dz->skip_blanks >= 0 ) {
working! doesn't seem to slow us down significantly ... before: $ time vips dzsave CMU-1.svs x --layout google --background "243 243 243 255" real 1m1.940s user 2m15.004s sys 0m37.092s after: $ time vips dzsave CMU-1.svs x --layout google --background "243 243 243 255" strip_work: skipping blank tile 176 x 67 strip_work: skipping blank tile 21 x 112 real 1m3.503s user 2m16.012s sys 0m40.328s
2015-11-24 18:08:56 +01:00
if( !(dz->ink = vips__vector_to_ink(
class->nickname, save->ready,
VIPS_AREA( save->background )->data, NULL,
VIPS_AREA( save->background )->n )) )
return( -1 );
}
faster --centre logic for dzsave don't test pixel-by-pixel for background tiles, instead track the rect of real pixels and test for intersection this might help performance scaling on many-core CPUs since we are no longer spinning up a threadgroup for each output tile thanks Kacey
2013-06-19 15:03:39 +02:00
/* The real pixels we have from our input. This is about to get
* expanded with background.
*/
real_pixels.left = 0;
real_pixels.top = 0;
real_pixels.width = save->ready->Xsize;
real_pixels.height = save->ready->Ysize;
add --centre option to dzsave with --centre turned on, in google maps mode dzsave will centre tiles rather than placing them at the top-left also, make sure blank.png is set to @background
2013-01-21 15:48:14 +01:00
/* For centred images, imagine shrinking so that the image fits in a
* single tile, centering in that tile, then expanding back again.
*/
if( dz->layout == VIPS_FOREIGN_DZ_LAYOUT_GOOGLE &&
dz->centre ) {
VipsImage *z;
Layer *layer;
int n_layers;
int size;
faster --centre logic for dzsave don't test pixel-by-pixel for background tiles, instead track the rect of real pixels and test for intersection this might help performance scaling on many-core CPUs since we are no longer spinning up a threadgroup for each output tile thanks Kacey
2013-06-19 15:03:39 +02:00
if( !(layer = pyramid_build( dz, NULL,
save->ready->Xsize, save->ready->Ysize,
&real_pixels )) )
add --centre option to dzsave with --centre turned on, in google maps mode dzsave will centre tiles rather than placing them at the top-left also, make sure blank.png is set to @background
2013-01-21 15:48:14 +01:00
return( -1 );
n_layers = layer->n;
/* This would cause interesting problems.
*/
g_assert( n_layers < 30 );
layer_free( layer );
size = dz->tile_size * (1 << n_layers);
faster --centre logic for dzsave don't test pixel-by-pixel for background tiles, instead track the rect of real pixels and test for intersection this might help performance scaling on many-core CPUs since we are no longer spinning up a threadgroup for each output tile thanks Kacey
2013-06-19 15:03:39 +02:00
real_pixels.left = (size - save->ready->Xsize) / 2;
real_pixels.top = (size - save->ready->Ysize) / 2;
add --centre option to dzsave with --centre turned on, in google maps mode dzsave will centre tiles rather than placing them at the top-left also, make sure blank.png is set to @background
2013-01-21 15:48:14 +01:00
if( vips_embed( save->ready, &z,
faster --centre logic for dzsave don't test pixel-by-pixel for background tiles, instead track the rect of real pixels and test for intersection this might help performance scaling on many-core CPUs since we are no longer spinning up a threadgroup for each output tile thanks Kacey
2013-06-19 15:03:39 +02:00
real_pixels.left, real_pixels.top,
add --centre option to dzsave with --centre turned on, in google maps mode dzsave will centre tiles rather than placing them at the top-left also, make sure blank.png is set to @background
2013-01-21 15:48:14 +01:00
size, size,
two small fixes for dzsave fix a tiny leak, fix coloured @background
2015-10-06 14:30:25 +02:00
"background", save->background,
add --centre option to dzsave with --centre turned on, in google maps mode dzsave will centre tiles rather than placing them at the top-left also, make sure blank.png is set to @background
2013-01-21 15:48:14 +01:00
NULL ) )
return( -1 );
VIPS_UNREF( save->ready );
save->ready = z;
#ifdef DEBUG
tiny polish
2013-06-20 11:05:46 +02:00
printf( "centre: centring within a %d x %d image\n",
add --centre option to dzsave with --centre turned on, in google maps mode dzsave will centre tiles rather than placing them at the top-left also, make sure blank.png is set to @background
2013-01-21 15:48:14 +01:00
size, size );
formatting in dzsave.c
2019-11-19 09:54:23 +01:00
#endif /*DEBUG*/
add --centre option to dzsave with --centre turned on, in google maps mode dzsave will centre tiles rather than placing them at the top-left also, make sure blank.png is set to @background
2013-01-21 15:48:14 +01:00
}
hacking
2013-02-25 15:54:39 +01:00
#ifdef DEBUG
printf( "vips_foreign_save_dz_build: tile_size == %d\n",
dz->tile_size );
printf( "vips_foreign_save_dz_build: overlap == %d\n",
dz->overlap );
new overlap system for dzsave again ... sigh see https://github.com/jcupitt/libvips/issues/512
2016-09-05 19:17:30 +02:00
printf( "vips_foreign_save_dz_build: tile_margin == %d\n",
dz->tile_margin );
printf( "vips_foreign_save_dz_build: tile_step == %d\n",
dz->tile_step );
formatting in dzsave.c
2019-11-19 09:54:23 +01:00
#endif /*DEBUG*/
hacking
2013-02-25 15:54:39 +01:00
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
/* Init basename and dirname from the associated filesystem names, if
* we can.
*/
{
const char *filename = dz->filename ?
dz->filename :
vips_connection_filename( VIPS_CONNECTION( dz->target ) );
if( !vips_object_argument_isset( object, "basename" ) ) {
if( filename )
dz->basename = g_path_get_basename( filename );
else
dz->basename = g_strdup( "untitled" );
}
if( !vips_object_argument_isset( object, "dirname" ) ) {
if( filename )
dz->dirname = g_path_get_dirname( filename );
}
}
/* Remove any [options] from basename.
*/
if( (p = (char *) vips__find_rightmost_brackets( dz->basename )) )
*p = '\0';
/* If we're writing thing.zip or thing.szi, default to zip
* container.
*/
if( (p = strrchr( dz->basename, '.' )) ) {
if( !vips_object_argument_isset( object, "container" ) ) {
if( g_ascii_strcasecmp( p + 1, "zip" ) == 0 )
dz->container = VIPS_FOREIGN_DZ_CONTAINER_ZIP;
if( g_ascii_strcasecmp( p + 1, "szi" ) == 0 )
dz->container = VIPS_FOREIGN_DZ_CONTAINER_SZI;
}
/* Remove any legal suffix. We don't remove all suffixes
* since we might be writing to a dirname with a dot in.
*/
if( g_ascii_strcasecmp( p + 1, "zip" ) == 0 ||
g_ascii_strcasecmp( p + 1, "szi" ) == 0 ||
g_ascii_strcasecmp( p + 1, "dz" ) == 0 )
*p = '\0';
}
started hacking in dzsave
2012-03-21 15:53:53 +01:00
/* Build the skeleton of the image pyramid.
*/
faster --centre logic for dzsave don't test pixel-by-pixel for background tiles, instead track the rect of real pixels and test for intersection this might help performance scaling on many-core CPUs since we are no longer spinning up a threadgroup for each output tile thanks Kacey
2013-06-19 15:03:39 +02:00
if( !(dz->layer = pyramid_build( dz, NULL,
save->ready->Xsize, save->ready->Ysize, &real_pixels )) )
started hacking in dzsave
2012-03-21 15:53:53 +01:00
return( -1 );
dzsave will write a zip file but see the TODO in dzsave.c - no way to pick a tile format, always jpeg - no way to pick a container format, always zip - needs testing - examine memuse, how well do we work for large zips - can we make a zip over 4gb?
2014-04-19 16:15:25 +02:00
if( dz->layout == VIPS_FOREIGN_DZ_LAYOUT_DZ )
add @container option to dzsave
2014-04-25 11:26:23 +02:00
dz->root_name = g_strdup_printf( "%s_files", dz->basename );
dzsave will write a zip file but see the TODO in dzsave.c - no way to pick a tile format, always jpeg - no way to pick a container format, always zip - needs testing - examine memuse, how well do we work for large zips - can we make a zip over 4gb?
2014-04-19 16:15:25 +02:00
else
add @container option to dzsave
2014-04-25 11:26:23 +02:00
dz->root_name = g_strdup( dz->basename );
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
/* Drop any [options] from @suffix.
stop dzsave early if the output dir exists see https://github.com/jcupitt/libvips/issues/312
2015-07-08 15:25:14 +02:00
*/
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
dz->file_suffix = g_strdup( dz->suffix );
if( (p = (char *) vips__find_rightmost_brackets( dz->file_suffix )) )
*p = '\0';
stop dzsave early if the output dir exists see https://github.com/jcupitt/libvips/issues/312
2015-07-08 15:25:14 +02:00
add @container option to dzsave
2014-04-25 11:26:23 +02:00
/* Make the thing we write the tiles into.
*/
switch( dz->container ) {
case VIPS_FOREIGN_DZ_CONTAINER_FS:
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
{
GsfOutput *out;
char name[VIPS_PATH_MAX];
/* For filesystem output of deepzoom, we write
* dirname/basename_files/ and dirname/basename.dzi, ie. the
* output does not go into a subdirectory.
*/
if( dz->layout == VIPS_FOREIGN_DZ_LAYOUT_DZ )
vips_snprintf( name, VIPS_PATH_MAX,
"%s", dz->dirname );
else
vips_snprintf( name, VIPS_PATH_MAX,
"%s/%s", dz->dirname, dz->basename );
if( !(out = (GsfOutput *)
dzsave: improve compatibility with libgsf < 1.14.29 (#3017) * Implement `GsfOutfileStdio` by ourselves * Add missing `->root` init * Fix styling * Update ChangeLog
2022-08-30 18:10:38 +02:00
gsf_output_dir_new( name, NULL )) ) {
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
return( -1 );
add @container option to dzsave
2014-04-25 11:26:23 +02:00
}
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
dz->tree = vips_gsf_tree_new( out, 0 );
}
add @container option to dzsave
2014-04-25 11:26:23 +02:00
break;
case VIPS_FOREIGN_DZ_CONTAINER_ZIP:
add SZI container type and improve the scan-properties file a bit
2018-01-27 13:52:22 +01:00
case VIPS_FOREIGN_DZ_CONTAINER_SZI:
add @container option to dzsave
2014-04-25 11:26:23 +02:00
{
GsfOutput *zip;
sort out dzsave paths, auto-enable zip rationalised output paths for dzsave enable zip automatically if output name ends in .zip
2014-05-06 15:28:41 +02:00
GsfOutput *out2;
add @container option to dzsave
2014-04-25 11:26:23 +02:00
GError *error = NULL;
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
/* We can have dzsave("x.zip", container="fs"), ie. zip output
* from write to file. Make a target if we need one.
add @container option to dzsave
2014-04-25 11:26:23 +02:00
*/
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
if( !dz->target ) {
if( !(dz->target =
vips_target_new_to_file( dz->filename )) )
add dzsave_buffer zip only ... see https://github.com/jcupitt/libvips/issues/415
2016-10-15 20:01:34 +02:00
return( -1 );
add @container option to dzsave
2014-04-25 11:26:23 +02:00
}
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
/* Can be memory, a file (not a directory tree), pipe, etc.
*/
dz->out = gsf_output_target_new( dz->target );
add @container option to dzsave
2014-04-25 11:26:23 +02:00
if( !(zip = (GsfOutput *)
add dzsave_buffer zip only ... see https://github.com/jcupitt/libvips/issues/415
2016-10-15 20:01:34 +02:00
gsf_outfile_zip_new( dz->out, &error )) ) {
add @container option to dzsave
2014-04-25 11:26:23 +02:00
vips_g_error( &error );
return( -1 );
}
sort out dzsave paths, auto-enable zip rationalised output paths for dzsave enable zip automatically if output name ends in .zip
2014-05-06 15:28:41 +02:00
/* Make the base directory inside the zip. All stuff goes into
* this.
*/
out2 = gsf_outfile_new_child_full( (GsfOutfile *) zip,
dz->basename, TRUE,
small dzsave cleanups - rename compression_level as deflate_level to match libgsf convention - use libgsf enums for GSF_ZIP_STORED, etc. - configure-time checks for zip64 and deflate-level - off-topic: rename NEED_TYPE_INIT as HAVE_TYPE_INIT for consistency
2016-06-09 10:36:37 +02:00
"compression-level", GSF_ZIP_STORED,
sort out dzsave paths, auto-enable zip rationalised output paths for dzsave enable zip automatically if output name ends in .zip
2014-05-06 15:28:41 +02:00
NULL );
small dzsave cleanups - rename compression_level as deflate_level to match libgsf convention - use libgsf enums for GSF_ZIP_STORED, etc. - configure-time checks for zip64 and deflate-level - off-topic: rename NEED_TYPE_INIT as HAVE_TYPE_INIT for consistency
2016-06-09 10:36:37 +02:00
#ifndef HAVE_GSF_DEFLATE_LEVEL
if( dz->compression > 0 ) {
more travis poking
2017-01-04 17:39:58 +01:00
g_warning( "%s",
_( "deflate-level not supported by libgsf, "
better dsave zip compression on older libgsfs if you ask for compression on an older libgsf, you now get zlib default, rather than no compression
2016-06-09 12:58:01 +02:00
"using default compression" ) );
dz->compression = -1;
Add dzsave zip compression support
2016-06-04 02:14:46 +02:00
}
formatting in dzsave.c
2019-11-19 09:54:23 +01:00
#endif /*HAVE_GSF_DEFLATE_LEVEL*/
small dzsave cleanups - rename compression_level as deflate_level to match libgsf convention - use libgsf enums for GSF_ZIP_STORED, etc. - configure-time checks for zip64 and deflate-level - off-topic: rename NEED_TYPE_INIT as HAVE_TYPE_INIT for consistency
2016-06-09 10:36:37 +02:00
remove is_zip from dzsave we can use GSF_IS_OUTFILE_ZIP() instead
2016-06-08 15:46:45 +02:00
dz->tree = vips_gsf_tree_new( out2, dz->compression );
sort out dzsave paths, auto-enable zip rationalised output paths for dzsave enable zip automatically if output name ends in .zip
2014-05-06 15:28:41 +02:00
/* Note the thing that will need closing up on exit.
*/
dz->tree->container = zip;
add @container option to dzsave
2014-04-25 11:26:23 +02:00
}
break;
default:
use g_assert_not_reached() we were using g_assert(0) ... g_assert_not_reached() makes the intention clearer
2016-01-24 14:21:53 +01:00
g_assert_not_reached();
add @container option to dzsave
2014-04-25 11:26:23 +02:00
}
started hacking in dzsave
2012-03-21 15:53:53 +01:00
works!
2012-03-26 23:08:59 +02:00
if( vips_sink_disc( save->ready, pyramid_strip, dz ) )
started hacking in dzsave
2012-03-21 15:53:53 +01:00
return( -1 );
dzsave can write zoomify and google maps use "layout" to pick the directory layout, deepzoom / zoomify / google all allowed
2012-10-03 15:26:59 +02:00
switch( dz->layout ) {
case VIPS_FOREIGN_DZ_LAYOUT_DZ:
if( write_dzi( dz ) )
return( -1 );
break;
case VIPS_FOREIGN_DZ_LAYOUT_ZOOMIFY:
if( write_properties( dz ) )
return( -1 );
break;
case VIPS_FOREIGN_DZ_LAYOUT_GOOGLE:
if( write_blank( dz ) )
return( -1 );
break;
add iiif layout to dzsave
2019-11-09 21:40:39 +01:00
case VIPS_FOREIGN_DZ_LAYOUT_IIIF:
Add iiif3 (#2483) * add iiif3 support to dzsave some small things have changed from v2
2021-10-19 15:18:40 +02:00
case VIPS_FOREIGN_DZ_LAYOUT_IIIF3:
add iiif layout to dzsave
2019-11-09 21:40:39 +01:00
if( write_json( dz ) )
return( -1 );
break;
dzsave can write zoomify and google maps use "layout" to pick the directory layout, deepzoom / zoomify / google all allowed
2012-10-03 15:26:59 +02:00
default:
use g_assert_not_reached() we were using g_assert(0) ... g_assert_not_reached() makes the intention clearer
2016-01-24 14:21:53 +01:00
g_assert_not_reached();
dzsave can write zoomify and google maps use "layout" to pick the directory layout, deepzoom / zoomify / google all allowed
2012-10-03 15:26:59 +02:00
}
deprecate "properties", register ".szi" as a suff
2022-03-29 18:50:48 +02:00
if( write_vips_meta( dz ) )
write props to a separate file and dump all vips props
2014-06-26 10:30:36 +02:00
return( -1 );
add SZI container type and improve the scan-properties file a bit
2018-01-27 13:52:22 +01:00
if( dz->container == VIPS_FOREIGN_DZ_CONTAINER_SZI &&
write scan-properties.xml to szi see https://github.com/jcupitt/libvips/issues/853
2018-01-26 18:20:58 +01:00
write_scan_properties( dz ) )
return( -1 );
write associated images to szi
2018-01-28 16:56:08 +01:00
if( dz->container == VIPS_FOREIGN_DZ_CONTAINER_SZI &&
write_associated( dz ) )
return( -1 );
better dzsave with zip output to a file close down output earlier to help mark-sweep bindings
2017-08-18 13:31:24 +02:00
/* Shut down the output to flush everything.
*/
if( vips_gsf_tree_close( dz->tree ) )
return( -1 );
dz->tree = NULL;
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
/* unref out to force flush in gsf_output_target_close().
better dzsave with zip output to a file close down output earlier to help mark-sweep bindings
2017-08-18 13:31:24 +02:00
*/
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
VIPS_UNREF( dz->out );
better dzsave with zip output to a file close down output earlier to help mark-sweep bindings
2017-08-18 13:31:24 +02:00
started hacking in dzsave
2012-03-21 15:53:53 +01:00
return( 0 );
}
/* Save a bit of typing.
*/
#define UC VIPS_FORMAT_UCHAR
#define C VIPS_FORMAT_CHAR
#define US VIPS_FORMAT_USHORT
#define S VIPS_FORMAT_SHORT
#define UI VIPS_FORMAT_UINT
#define I VIPS_FORMAT_INT
#define F VIPS_FORMAT_FLOAT
#define X VIPS_FORMAT_COMPLEX
#define D VIPS_FORMAT_DOUBLE
#define DX VIPS_FORMAT_DPCOMPLEX
static int bandfmt_dz[10] = {
/* UC C US S UI I F X D DX */
UC, C, US, S, UI, I, F, F, D, D
};
deprecate "properties", register ".szi" as a suff
2022-03-29 18:50:48 +02:00
static const char *dz_suffs[] = { ".dz", ".szi", NULL };
works!
2012-03-26 23:08:59 +02:00
started hacking in dzsave
2012-03-21 15:53:53 +01:00
static void
vips_foreign_save_dz_class_init( VipsForeignSaveDzClass *class )
{
GObjectClass *gobject_class = G_OBJECT_CLASS( class );
VipsObjectClass *object_class = (VipsObjectClass *) class;
VipsForeignClass *foreign_class = (VipsForeignClass *) class;
VipsForeignSaveClass *save_class = (VipsForeignSaveClass *) class;
gobject_class->dispose = vips_foreign_save_dz_dispose;
gobject_class->set_property = vips_object_set_property;
gobject_class->get_property = vips_object_get_property;
add dzsave_buffer zip only ... see https://github.com/jcupitt/libvips/issues/415
2016-10-15 20:01:34 +02:00
object_class->nickname = "dzsave_base";
started hacking in dzsave
2012-03-21 15:53:53 +01:00
object_class->description = _( "save image to deep zoom format" );
object_class->build = vips_foreign_save_dz_build;
works!
2012-03-26 23:08:59 +02:00
foreign_class->suffs = dz_suffs;
started hacking in dzsave
2012-03-21 15:53:53 +01:00
save_class->saveable = VIPS_SAVEABLE_ANY;
save_class->format_table = bandfmt_dz;
save_class->coding[VIPS_CODING_LABQ] = TRUE;
add dzsave_buffer zip only ... see https://github.com/jcupitt/libvips/issues/415
2016-10-15 20:01:34 +02:00
VIPS_ARG_STRING( class, "basename", 2,
_( "Base name" ),
_( "Base name to save to" ),
VIPS_ARGUMENT_OPTIONAL_INPUT,
G_STRUCT_OFFSET( VipsForeignSaveDz, basename ),
started hacking in dzsave
2012-03-21 15:53:53 +01:00
NULL );
dzsave can write zoomify and google maps use "layout" to pick the directory layout, deepzoom / zoomify / google all allowed
2012-10-03 15:26:59 +02:00
VIPS_ARG_ENUM( class, "layout", 8,
_( "Layout" ),
_( "Directory layout" ),
VIPS_ARGUMENT_OPTIONAL_INPUT,
G_STRUCT_OFFSET( VipsForeignSaveDz, layout ),
tiffsave has region_shrink param for new pyr shrink stuff
2018-06-13 19:16:03 +02:00
VIPS_TYPE_FOREIGN_DZ_LAYOUT, VIPS_FOREIGN_DZ_LAYOUT_DZ );
dzsave can write zoomify and google maps use "layout" to pick the directory layout, deepzoom / zoomify / google all allowed
2012-10-03 15:26:59 +02:00
started hacking in dzsave
2012-03-21 15:53:53 +01:00
VIPS_ARG_STRING( class, "suffix", 9,
Update C++ binding and function list (#2851) * Update C++ binding and function list Also, start with a capital letter in descriptions whenever possible. * Ensure `dzsave_target` is generated
2022-06-10 15:43:44 +02:00
_( "Suffix" ),
started hacking in dzsave
2012-03-21 15:53:53 +01:00
_( "Filename suffix for tiles" ),
VIPS_ARGUMENT_OPTIONAL_INPUT,
G_STRUCT_OFFSET( VipsForeignSaveDz, suffix ),
add a vips7 interface to the deepzoom writer
2013-06-11 13:32:46 +02:00
".jpeg" );
started hacking in dzsave
2012-03-21 15:53:53 +01:00
VIPS_ARG_INT( class, "overlap", 10,
_( "Overlap" ),
_( "Tile overlap in pixels" ),
VIPS_ARGUMENT_OPTIONAL_INPUT,
G_STRUCT_OFFSET( VipsForeignSaveDz, overlap ),
correct default overlap value should be 1 for default layouot mode see https://github.com/jcupitt/libvips/issues/357
2015-12-16 17:26:30 +01:00
0, 8192, 1 );
started hacking in dzsave
2012-03-21 15:53:53 +01:00
more improvements to dzsave now writes x_files and x.dzi, as per spec (thanks Benjamin) deprecate tile_width and tile_height, just have tile_size now
2012-08-07 14:29:03 +02:00
VIPS_ARG_INT( class, "tile_size", 11,
_( "Tile size" ),
_( "Tile size in pixels" ),
VIPS_ARGUMENT_OPTIONAL_INPUT,
G_STRUCT_OFFSET( VipsForeignSaveDz, tile_size ),
Use 254 pixels as default tile size I remember @benvanik, a friend and colleague, who worked on our DZI renderer (Seadragon) told us that, I believe, it’s better for GPUs to have power of two sized tiles, so 254px + 2 x 1px tile overlap will give you 256 pixel.s > To keep my tiles nicely sized, I shrink in instead of expand out – that means my tiles are really 254×254 with a 1px border, so the images are 256x256px with 254×254 of useful imagery. If you wanted larger tiles, you’d go 510×510 with 1px border making 512×512 tiles. > — Source: http://www.noxa.org/blog/2009/11/29/megatextures-in-webgl-2/
2015-12-21 06:43:41 +01:00
1, 8192, 254 );
more improvements to dzsave now writes x_files and x.dzi, as per spec (thanks Benjamin) deprecate tile_width and tile_height, just have tile_size now
2012-08-07 14:29:03 +02:00
add @depth option to dzsave so you can ask for a 1 layer pyramid
2012-11-01 19:11:46 +01:00
VIPS_ARG_ENUM( class, "depth", 13,
_( "Depth" ),
_( "Pyramid depth" ),
VIPS_ARGUMENT_OPTIONAL_INPUT,
G_STRUCT_OFFSET( VipsForeignSaveDz, depth ),
tiffsave has region_shrink param for new pyr shrink stuff
2018-06-13 19:16:03 +02:00
VIPS_TYPE_FOREIGN_DZ_DEPTH, VIPS_FOREIGN_DZ_DEPTH_ONEPIXEL );
add @depth option to dzsave so you can ask for a 1 layer pyramid
2012-11-01 19:11:46 +01:00
add --centre option to dzsave with --centre turned on, in google maps mode dzsave will centre tiles rather than placing them at the top-left also, make sure blank.png is set to @background
2013-01-21 15:48:14 +01:00
VIPS_ARG_BOOL( class, "centre", 13,
_( "Center" ),
_( "Center image in tile" ),
VIPS_ARGUMENT_OPTIONAL_INPUT,
G_STRUCT_OFFSET( VipsForeignSaveDz, centre ),
FALSE );
add --angle option to dzsave
2013-05-29 15:06:23 +02:00
VIPS_ARG_ENUM( class, "angle", 14,
_( "Angle" ),
_( "Rotate image during save" ),
VIPS_ARGUMENT_OPTIONAL_INPUT,
G_STRUCT_OFFSET( VipsForeignSaveDz, angle ),
rename VIPS_ANGLE_0 as VIPS_ANGLE_D0 and other similar cases VIPS_ANGLE_0 becomes Vips.Angle.0 in Python, which is very inconvenient, so we rename ... the D stands for degrees, I think
2014-09-29 14:56:55 +02:00
VIPS_TYPE_ANGLE, VIPS_ANGLE_D0 );
add --angle option to dzsave
2013-05-29 15:06:23 +02:00
add @container option to dzsave
2014-04-25 11:26:23 +02:00
VIPS_ARG_ENUM( class, "container", 15,
_( "Container" ),
_( "Pyramid container type" ),
VIPS_ARGUMENT_OPTIONAL_INPUT,
G_STRUCT_OFFSET( VipsForeignSaveDz, container ),
tiffsave has region_shrink param for new pyr shrink stuff
2018-06-13 19:16:03 +02:00
VIPS_TYPE_FOREIGN_DZ_CONTAINER, VIPS_FOREIGN_DZ_CONTAINER_FS );
add @container option to dzsave
2014-04-25 11:26:23 +02:00
Add dzsave zip compression support
2016-06-04 02:14:46 +02:00
VIPS_ARG_INT( class, "compression", 17,
_( "Compression" ),
_( "ZIP deflate compression level" ),
VIPS_ARGUMENT_OPTIONAL_INPUT,
G_STRUCT_OFFSET( VipsForeignSaveDz, compression ),
-1, 9, 0 );
Adds VipsRegionShrink type
2018-03-27 22:35:49 +02:00
VIPS_ARG_ENUM( class, "region_shrink", 18,
_( "Region shrink" ),
tiffsave has region_shrink param for new pyr shrink stuff
2018-06-13 19:16:03 +02:00
_( "Method to shrink regions" ),
Adds VipsRegionShrink type
2018-03-27 22:35:49 +02:00
VIPS_ARGUMENT_OPTIONAL_INPUT,
G_STRUCT_OFFSET( VipsForeignSaveDz, region_shrink ),
tiffsave has region_shrink param for new pyr shrink stuff
2018-06-13 19:16:03 +02:00
VIPS_TYPE_REGION_SHRINK, VIPS_REGION_SHRINK_MEAN );
Adds VipsRegionShrink type
2018-03-27 22:35:49 +02:00
make skip_blanks an int so you can adjust the threshold
2018-12-20 18:07:07 +01:00
VIPS_ARG_INT( class, "skip_blanks", 19,
add skip-blanks option to dzsave Skip-blanks was a feature of google layout. This patch makes it into an option you can control which defaults on in gm mode. See https://github.com/libvips/libvips/issues/1185
2018-12-19 15:52:33 +01:00
_( "Skip blanks" ),
_( "Skip tiles which are nearly equal to the background" ),
VIPS_ARGUMENT_OPTIONAL_INPUT,
G_STRUCT_OFFSET( VipsForeignSaveDz, skip_blanks ),
fix dzsave skip_blanks default the default value was not being declared correctly [felixbuenemann] see https://github.com/libvips/libvips/issues/1185#issuecomment-450427889
2018-12-28 23:23:15 +01:00
-1, 65535, -1 );
add skip-blanks option to dzsave Skip-blanks was a feature of google layout. This patch makes it into an option you can control which defaults on in gm mode. See https://github.com/libvips/libvips/issues/1185
2018-12-19 15:52:33 +01:00
add no_strip option to dzsave it's sometimes useful to have pyramid tiles with ICC profiles see https://github.com/libvips/libvips/issues/1451
2019-10-21 17:46:20 +02:00
VIPS_ARG_BOOL( class, "no_strip", 20,
_( "No strip" ),
_( "Don't strip tile metadata" ),
VIPS_ARGUMENT_OPTIONAL_INPUT,
G_STRUCT_OFFSET( VipsForeignSaveDz, no_strip ),
FALSE );
add @id arg to dzsave So you can now set the base of the id property in info.json in IIIF output. Thanks regisrob. See https://github.com/libvips/libvips/issues/1531
2020-01-24 18:12:46 +01:00
VIPS_ARG_STRING( class, "id", 21,
_( "id" ),
_( "Resource ID" ),
VIPS_ARGUMENT_OPTIONAL_INPUT,
G_STRUCT_OFFSET( VipsForeignSaveDz, id ),
"https://example.com/iiif" );
more improvements to dzsave now writes x_files and x.dzi, as per spec (thanks Benjamin) deprecate tile_width and tile_height, just have tile_size now
2012-08-07 14:29:03 +02:00
/* How annoying. We stupidly had these in earlier versions.
*/
VIPS_ARG_STRING( class, "dirname", 1,
add dzsave_buffer zip only ... see https://github.com/jcupitt/libvips/issues/415
2016-10-15 20:01:34 +02:00
_( "Directory name" ),
_( "Directory name to save to" ),
VIPS_ARGUMENT_OPTIONAL_INPUT | VIPS_ARGUMENT_DEPRECATED,
G_STRUCT_OFFSET( VipsForeignSaveDz, dirname ),
allow .dz as a destination so this now works: $ vips extract_area 50310.svs x.dz 10 10 100 100 and writes a deepzoom pyramid
2013-09-24 10:00:55 +02:00
NULL );
more improvements to dzsave now writes x_files and x.dzi, as per spec (thanks Benjamin) deprecate tile_width and tile_height, just have tile_size now
2012-08-07 14:29:03 +02:00
VIPS_ARG_INT( class, "tile_width", 12,
started hacking in dzsave
2012-03-21 15:53:53 +01:00
_( "Tile width" ),
_( "Tile width in pixels" ),
more improvements to dzsave now writes x_files and x.dzi, as per spec (thanks Benjamin) deprecate tile_width and tile_height, just have tile_size now
2012-08-07 14:29:03 +02:00
VIPS_ARGUMENT_OPTIONAL_INPUT | VIPS_ARGUMENT_DEPRECATED,
G_STRUCT_OFFSET( VipsForeignSaveDz, tile_size ),
Use 254 pixels as default tile size I remember @benvanik, a friend and colleague, who worked on our DZI renderer (Seadragon) told us that, I believe, it’s better for GPUs to have power of two sized tiles, so 254px + 2 x 1px tile overlap will give you 256 pixel.s > To keep my tiles nicely sized, I shrink in instead of expand out – that means my tiles are really 254×254 with a 1px border, so the images are 256x256px with 254×254 of useful imagery. If you wanted larger tiles, you’d go 510×510 with 1px border making 512×512 tiles. > — Source: http://www.noxa.org/blog/2009/11/29/megatextures-in-webgl-2/
2015-12-21 06:43:41 +01:00
1, 8192, 254 );
started hacking in dzsave
2012-03-21 15:53:53 +01:00
VIPS_ARG_INT( class, "tile_height", 12,
_( "Tile height" ),
_( "Tile height in pixels" ),
more improvements to dzsave now writes x_files and x.dzi, as per spec (thanks Benjamin) deprecate tile_width and tile_height, just have tile_size now
2012-08-07 14:29:03 +02:00
VIPS_ARGUMENT_OPTIONAL_INPUT | VIPS_ARGUMENT_DEPRECATED,
G_STRUCT_OFFSET( VipsForeignSaveDz, tile_size ),
Use 254 pixels as default tile size I remember @benvanik, a friend and colleague, who worked on our DZI renderer (Seadragon) told us that, I believe, it’s better for GPUs to have power of two sized tiles, so 254px + 2 x 1px tile overlap will give you 256 pixel.s > To keep my tiles nicely sized, I shrink in instead of expand out – that means my tiles are really 254×254 with a 1px border, so the images are 256x256px with 254×254 of useful imagery. If you wanted larger tiles, you’d go 510×510 with 1px border making 512×512 tiles. > — Source: http://www.noxa.org/blog/2009/11/29/megatextures-in-webgl-2/
2015-12-21 06:43:41 +01:00
1, 8192, 254 );
started hacking in dzsave
2012-03-21 15:53:53 +01:00
deprecate "properties", register ".szi" as a suff
2022-03-29 18:50:48 +02:00
VIPS_ARG_BOOL( class, "properties", 16,
_( "Properties" ),
_( "Write a properties file to the output directory" ),
VIPS_ARGUMENT_OPTIONAL_INPUT | VIPS_ARGUMENT_DEPRECATED,
G_STRUCT_OFFSET( VipsForeignSaveDz, properties ),
FALSE );
started hacking in dzsave
2012-03-21 15:53:53 +01:00
}
static void
vips_foreign_save_dz_init( VipsForeignSaveDz *dz )
{
fixes to deep zoom write - shrink down to a 1x1 pixel tile, even for very long and thin images - round image size up on shrink - write a .dzi file with the pyramid params - default tile size and overlap now matches the openslide writer thanks to Benjamin Gilbert for pointing out the errors
2012-08-06 16:14:59 +02:00
VIPS_SETSTR( dz->suffix, ".jpeg" );
dzsave can write zoomify and google maps use "layout" to pick the directory layout, deepzoom / zoomify / google all allowed
2012-10-03 15:26:59 +02:00
dz->layout = VIPS_FOREIGN_DZ_LAYOUT_DZ;
fixes to deep zoom write - shrink down to a 1x1 pixel tile, even for very long and thin images - round image size up on shrink - write a .dzi file with the pyramid params - default tile size and overlap now matches the openslide writer thanks to Benjamin Gilbert for pointing out the errors
2012-08-06 16:14:59 +02:00
dz->overlap = 1;
Use 254 pixels as default tile size I remember @benvanik, a friend and colleague, who worked on our DZI renderer (Seadragon) told us that, I believe, it’s better for GPUs to have power of two sized tiles, so 254px + 2 x 1px tile overlap will give you 256 pixel.s > To keep my tiles nicely sized, I shrink in instead of expand out – that means my tiles are really 254×254 with a 1px border, so the images are 256x256px with 254×254 of useful imagery. If you wanted larger tiles, you’d go 510×510 with 1px border making 512×512 tiles. > — Source: http://www.noxa.org/blog/2009/11/29/megatextures-in-webgl-2/
2015-12-21 06:43:41 +01:00
dz->tile_size = 254;
dzsave can write zoomify and google maps use "layout" to pick the directory layout, deepzoom / zoomify / google all allowed
2012-10-03 15:26:59 +02:00
dz->tile_count = 0;
rename _!PIXEL etc. to help bindings ruby gobject-introspection hates numbers at the start of _ separated names
2015-01-07 15:22:13 +01:00
dz->depth = VIPS_FOREIGN_DZ_DEPTH_ONEPIXEL;
rename VIPS_ANGLE_0 as VIPS_ANGLE_D0 and other similar cases VIPS_ANGLE_0 becomes Vips.Angle.0 in Python, which is very inconvenient, so we rename ... the D stands for degrees, I think
2014-09-29 14:56:55 +02:00
dz->angle = VIPS_ANGLE_D0;
add @container option to dzsave
2014-04-25 11:26:23 +02:00
dz->container = VIPS_FOREIGN_DZ_CONTAINER_FS;
Add dzsave zip compression support
2016-06-04 02:14:46 +02:00
dz->compression = 0;
Adds VipsRegionShrink type
2018-03-27 22:35:49 +02:00
dz->region_shrink = VIPS_REGION_SHRINK_MEAN;
use -1 for do-nothing so that users can have skip-only-exact-blanks see https://github.com/libvips/libvips/issues/1185#issuecomment-449115143
2018-12-21 15:08:54 +01:00
dz->skip_blanks = -1;
started hacking in dzsave
2012-03-21 15:53:53 +01:00
}
more improvements to dzsave now writes x_files and x.dzi, as per spec (thanks Benjamin) deprecate tile_width and tile_height, just have tile_size now
2012-08-07 14:29:03 +02:00
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
typedef struct _VipsForeignSaveDzTarget {
VipsForeignSaveDz parent_object;
VipsTarget *target;
} VipsForeignSaveDzTarget;
typedef VipsForeignSaveDzClass VipsForeignSaveDzTargetClass;
G_DEFINE_TYPE( VipsForeignSaveDzTarget, vips_foreign_save_dz_target,
vips_foreign_save_dz_get_type() );
static int
vips_foreign_save_dz_target_build( VipsObject *object )
{
VipsForeignSaveDz *dz = (VipsForeignSaveDz *) object;
VipsForeignSaveDzTarget *target = (VipsForeignSaveDzTarget *) object;
dz->target = target->target;
Update C++ binding and function list (#2851) * Update C++ binding and function list Also, start with a capital letter in descriptions whenever possible. * Ensure `dzsave_target` is generated
2022-06-10 15:43:44 +02:00
g_object_ref( dz->target );
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
if( VIPS_OBJECT_CLASS( vips_foreign_save_dz_target_parent_class )->
build( object ) )
return( -1 );
return( 0 );
}
static void
vips_foreign_save_dz_target_class_init( VipsForeignSaveDzTargetClass *class )
{
GObjectClass *gobject_class = G_OBJECT_CLASS( class );
VipsObjectClass *object_class = (VipsObjectClass *) class;
gobject_class->set_property = vips_object_set_property;
gobject_class->get_property = vips_object_get_property;
Update C++ binding and function list (#2851) * Update C++ binding and function list Also, start with a capital letter in descriptions whenever possible. * Ensure `dzsave_target` is generated
2022-06-10 15:43:44 +02:00
object_class->nickname = "dzsave_target";
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
object_class->description = _( "save image to deepzoom target" );
object_class->build = vips_foreign_save_dz_target_build;
VIPS_ARG_OBJECT( class, "target", 1,
_( "Target" ),
_( "Target to save to" ),
VIPS_ARGUMENT_REQUIRED_INPUT,
G_STRUCT_OFFSET( VipsForeignSaveDzTarget, target ),
VIPS_TYPE_TARGET );
}
static void
vips_foreign_save_dz_target_init( VipsForeignSaveDzTarget *target )
{
VipsForeignSaveDz *dz = (VipsForeignSaveDz *) target;
/* zip default for target output.
*/
dz->container = VIPS_FOREIGN_DZ_CONTAINER_ZIP;
}
add dzsave_buffer zip only ... see https://github.com/jcupitt/libvips/issues/415
2016-10-15 20:01:34 +02:00
typedef struct _VipsForeignSaveDzFile {
VipsForeignSaveDz parent_object;
/* Filename for save.
*/
char *filename;
} VipsForeignSaveDzFile;
typedef VipsForeignSaveDzClass VipsForeignSaveDzFileClass;
G_DEFINE_TYPE( VipsForeignSaveDzFile, vips_foreign_save_dz_file,
vips_foreign_save_dz_get_type() );
static int
vips_foreign_save_dz_file_build( VipsObject *object )
{
VipsForeignSaveDz *dz = (VipsForeignSaveDz *) object;
VipsForeignSaveDzFile *file = (VipsForeignSaveDzFile *) object;
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
dz->filename = file->filename;
add dzsave_buffer zip only ... see https://github.com/jcupitt/libvips/issues/415
2016-10-15 20:01:34 +02:00
if( VIPS_OBJECT_CLASS( vips_foreign_save_dz_file_parent_class )->
build( object ) )
return( -1 );
return( 0 );
}
static void
vips_foreign_save_dz_file_class_init( VipsForeignSaveDzFileClass *class )
{
GObjectClass *gobject_class = G_OBJECT_CLASS( class );
VipsObjectClass *object_class = (VipsObjectClass *) class;
gobject_class->set_property = vips_object_set_property;
gobject_class->get_property = vips_object_get_property;
object_class->nickname = "dzsave";
object_class->description = _( "save image to deepzoom file" );
object_class->build = vips_foreign_save_dz_file_build;
VIPS_ARG_STRING( class, "filename", 1,
_( "Filename" ),
_( "Filename to save to" ),
VIPS_ARGUMENT_REQUIRED_INPUT,
G_STRUCT_OFFSET( VipsForeignSaveDzFile, filename ),
NULL );
}
static void
vips_foreign_save_dz_file_init( VipsForeignSaveDzFile *file )
{
}
typedef struct _VipsForeignSaveDzBuffer {
VipsForeignSaveDz parent_object;
VipsArea *buf;
} VipsForeignSaveDzBuffer;
typedef VipsForeignSaveDzClass VipsForeignSaveDzBufferClass;
G_DEFINE_TYPE( VipsForeignSaveDzBuffer, vips_foreign_save_dz_buffer,
vips_foreign_save_dz_get_type() );
static int
vips_foreign_save_dz_buffer_build( VipsObject *object )
{
VipsForeignSaveDz *dz = (VipsForeignSaveDz *) object;
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
VipsForeignSaveDzBuffer *buffer = (VipsForeignSaveDzBuffer *) object;
add dzsave_buffer zip only ... see https://github.com/jcupitt/libvips/issues/415
2016-10-15 20:01:34 +02:00
VipsBlob *blob;
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
if( !(dz->target = vips_target_new_to_memory()) )
return( -1 );
add dzsave_buffer zip only ... see https://github.com/jcupitt/libvips/issues/415
2016-10-15 20:01:34 +02:00
if( VIPS_OBJECT_CLASS( vips_foreign_save_dz_buffer_parent_class )->
build( object ) )
return( -1 );
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
g_object_get( dz->target, "blob", &blob, NULL );
g_object_set( buffer, "buffer", blob, NULL );
add dzsave_buffer zip only ... see https://github.com/jcupitt/libvips/issues/415
2016-10-15 20:01:34 +02:00
vips_area_unref( VIPS_AREA( blob ) );
return( 0 );
}
static void
vips_foreign_save_dz_buffer_class_init( VipsForeignSaveDzBufferClass *class )
{
GObjectClass *gobject_class = G_OBJECT_CLASS( class );
VipsObjectClass *object_class = (VipsObjectClass *) class;
gobject_class->set_property = vips_object_set_property;
gobject_class->get_property = vips_object_get_property;
object_class->nickname = "dzsave_buffer";
object_class->description = _( "save image to dz buffer" );
object_class->build = vips_foreign_save_dz_buffer_build;
VIPS_ARG_BOXED( class, "buffer", 1,
_( "Buffer" ),
_( "Buffer to save to" ),
VIPS_ARGUMENT_REQUIRED_OUTPUT,
G_STRUCT_OFFSET( VipsForeignSaveDzBuffer, buf ),
VIPS_TYPE_BLOB );
}
static void
vips_foreign_save_dz_buffer_init( VipsForeignSaveDzBuffer *buffer )
{
add SZI container type and improve the scan-properties file a bit
2018-01-27 13:52:22 +01:00
VipsForeignSaveDz *dz = (VipsForeignSaveDz *) buffer;
/* zip default for memory output.
*/
dz->container = VIPS_FOREIGN_DZ_CONTAINER_ZIP;
add dzsave_buffer zip only ... see https://github.com/jcupitt/libvips/issues/415
2016-10-15 20:01:34 +02:00
}
make libgsf-1 an optional dependency dzsave vanishes if you don't have it, but you can still build vips
2014-05-09 14:13:18 +02:00
#endif /*HAVE_GSF*/
more improvements to dzsave now writes x_files and x.dzi, as per spec (thanks Benjamin) deprecate tile_width and tile_height, just have tile_size now
2012-08-07 14:29:03 +02:00
/**
Update GObject introspection annotations for libvips/foreign directory
2017-09-27 00:39:06 +02:00
* vips_dzsave: (method)
more improvements to dzsave now writes x_files and x.dzi, as per spec (thanks Benjamin) deprecate tile_width and tile_height, just have tile_size now
2012-08-07 14:29:03 +02:00
* @in: image to save
add @container option to dzsave
2014-04-25 11:26:23 +02:00
* @name: name to save to
more improvements to dzsave now writes x_files and x.dzi, as per spec (thanks Benjamin) deprecate tile_width and tile_height, just have tile_size now
2012-08-07 14:29:03 +02:00
* @...: %NULL-terminated list of optional named arguments
*
* Optional arguments:
*
add @id arg to dzsave So you can now set the base of the id property in info.json in IIIF output. Thanks regisrob. See https://github.com/libvips/libvips/issues/1531
2020-01-24 18:12:46 +01:00
* * @basename: %gchar base part of name
better dsave zip compression on older libgsfs if you ask for compression on an older libgsf, you now get zlib default, rather than no compression
2016-06-09 12:58:01 +02:00
* * @layout: #VipsForeignDzLayout directory layout convention
add @id arg to dzsave So you can now set the base of the id property in info.json in IIIF output. Thanks regisrob. See https://github.com/libvips/libvips/issues/1531
2020-01-24 18:12:46 +01:00
* * @suffix: %gchar suffix for tiles
better dsave zip compression on older libgsfs if you ask for compression on an older libgsf, you now get zlib default, rather than no compression
2016-06-09 12:58:01 +02:00
* * @overlap: %gint set tile overlap
* * @tile_size: %gint set tile size
* * @background: #VipsArrayDouble background colour
* * @depth: #VipsForeignDzDepth how deep to make the pyramid
* * @centre: %gboolean centre the tiles
* * @angle: #VipsAngle rotate the image by this much
* * @container: #VipsForeignDzContainer set container type
* * @compression: %gint zip deflate compression level
use -1 for do-nothing so that users can have skip-only-exact-blanks see https://github.com/libvips/libvips/issues/1185#issuecomment-449115143
2018-12-21 15:08:54 +01:00
* * @region_shrink: #VipsRegionShrink how to shrink each 2x2 region
* * @skip_blanks: %gint skip tiles which are nearly equal to the background
add no_strip option to dzsave it's sometimes useful to have pyramid tiles with ICC profiles see https://github.com/libvips/libvips/issues/1451
2019-10-21 17:46:20 +02:00
* * @no_strip: %gboolean don't strip tiles
add @id arg to dzsave So you can now set the base of the id property in info.json in IIIF output. Thanks regisrob. See https://github.com/libvips/libvips/issues/1531
2020-01-24 18:12:46 +01:00
* * @id: %gchar id for IIIF properties
more improvements to dzsave now writes x_files and x.dzi, as per spec (thanks Benjamin) deprecate tile_width and tile_height, just have tile_size now
2012-08-07 14:29:03 +02:00
*
dzsave can write zoomify and google maps use "layout" to pick the directory layout, deepzoom / zoomify / google all allowed
2012-10-03 15:26:59 +02:00
* Save an image as a set of tiles at various resolutions. By default dzsave
* uses DeepZoom layout -- use @layout to pick other conventions.
more improvements to dzsave now writes x_files and x.dzi, as per spec (thanks Benjamin) deprecate tile_width and tile_height, just have tile_size now
2012-08-07 14:29:03 +02:00
*
update dzsave docs
2014-05-06 15:38:24 +02:00
* vips_dzsave() creates a directory called @name to hold the tiles. If @name
more tweaks to dzsave props add docs too
2014-06-30 13:58:57 +02:00
* ends `.zip`, vips_dzsave() will create a zip file called @name to hold the
add dzsave_buffer zip only ... see https://github.com/jcupitt/libvips/issues/415
2016-10-15 20:01:34 +02:00
* tiles. You can use @container to force zip file output.
*
* Use @basename to set the name of the directory tree we are creating. The
* default value is set from @name.
more improvements to dzsave now writes x_files and x.dzi, as per spec (thanks Benjamin) deprecate tile_width and tile_height, just have tile_size now
2012-08-07 14:29:03 +02:00
*
more tweaks to dzsave props add docs too
2014-06-30 13:58:57 +02:00
* You can set @suffix to something like `".jpg[Q=85]"` to control the tile
* write options.
add a @background option to embed and dzsave you can now set the background colour for edge tiles in dzsave and embed, eg.: vips dzsave k2.jpg x --background "1 2 3"
2012-10-10 14:22:30 +02:00
*
* In Google layout mode, edge tiles are expanded to @tile_size by @tile_size
* pixels. Normally they are filled with white, but you can set another colour
add --centre option to dzsave with --centre turned on, in google maps mode dzsave will centre tiles rather than placing them at the top-left also, make sure blank.png is set to @background
2013-01-21 15:48:14 +01:00
* with @background. Images are usually placed at the top-left of the tile,
* but you can have them centred by turning on @centre.
more improvements to dzsave now writes x_files and x.dzi, as per spec (thanks Benjamin) deprecate tile_width and tile_height, just have tile_size now
2012-08-07 14:29:03 +02:00
*
dzsave can write zoomify and google maps use "layout" to pick the directory layout, deepzoom / zoomify / google all allowed
2012-10-03 15:26:59 +02:00
* You can set the size and overlap of tiles with @tile_size and @overlap.
notes on dzsave change see https://github.com/jcupitt/libvips/pull/359
2015-12-23 10:16:58 +01:00
* They default to the correct settings for the selected @layout. The deepzoom
* defaults produce 256x256 jpeg files for centre tiles, the most efficient
* size.
dzsave can write zoomify and google maps use "layout" to pick the directory layout, deepzoom / zoomify / google all allowed
2012-10-03 15:26:59 +02:00
*
add @depth option to dzsave so you can ask for a 1 layer pyramid
2012-11-01 19:11:46 +01:00
* Use @depth to control how low the pyramid goes. This defaults to the
* correct setting for the @layout you select.
*
note dzsave angle restrictions see https://github.com/jcupitt/libvips/issues/968
2018-05-09 17:06:20 +02:00
* You can rotate the image during write with the @angle argument. However,
* this will only work for images which support random access, like openslide,
tiffsave has region_shrink param for new pyr shrink stuff
2018-06-13 19:16:03 +02:00
* and not for things like JPEG. You'll need to rotate those images
note dzsave angle restrictions see https://github.com/jcupitt/libvips/issues/968
2018-05-09 17:06:20 +02:00
* yourself with vips_rot(). Note that the `autorotate` option to the loader
* may do what you need.
*
Add iiif3 (#2483) * add iiif3 support to dzsave some small things have changed from v2
2021-10-19 15:18:40 +02:00
* By default, all tiles are stripped since usually you do not want a copy of
* all metadata in every tile. Set @no_strip if you want to keep metadata.
add no_strip option to dzsave it's sometimes useful to have pyramid tiles with ICC profiles see https://github.com/libvips/libvips/issues/1451
2019-10-21 17:46:20 +02:00
*
Add dzsave zip compression support
2016-06-04 02:14:46 +02:00
* If @container is set to `zip`, you can set a compression level from -1
* (use zlib default), 0 (store, compression disabled) to 9 (max compression).
* If no value is given, the default is to store files without compression.
*
Adds VipsRegionShrink type
2018-03-27 22:35:49 +02:00
* You can use @region_shrink to control the method for shrinking each 2x2
* region. This defaults to using the average of the 4 input pixels but you can
tiffsave has region_shrink param for new pyr shrink stuff
2018-06-13 19:16:03 +02:00
* also use the median in cases where you want to preserve the range of values.
add skip-blanks option to dzsave Skip-blanks was a feature of google layout. This patch makes it into an option you can control which defaults on in gm mode. See https://github.com/libvips/libvips/issues/1185
2018-12-19 15:52:33 +01:00
*
use -1 for do-nothing so that users can have skip-only-exact-blanks see https://github.com/libvips/libvips/issues/1185#issuecomment-449115143
2018-12-21 15:08:54 +01:00
* If you set @skip_blanks to a value greater than or equal to zero, tiles
* which are all within that many pixel values to the background are skipped.
* This can save a lot of space for some image types. This option defaults to
* 5 in Google layout mode, -1 otherwise.
add @id arg to dzsave So you can now set the base of the id property in info.json in IIIF output. Thanks regisrob. See https://github.com/libvips/libvips/issues/1531
2020-01-24 18:12:46 +01:00
*
* In IIIF layout, you can set the base of the `id` property in `info.json`
* with @id. The default is `https://example.com/iiif`.
Add iiif3 (#2483) * add iiif3 support to dzsave some small things have changed from v2
2021-10-19 15:18:40 +02:00
*
* Use @layout #VIPS_FOREIGN_DZ_LAYOUT_IIIF3 for IIIF v3 layout.
Adds VipsRegionShrink type
2018-03-27 22:35:49 +02:00
*
more improvements to dzsave now writes x_files and x.dzi, as per spec (thanks Benjamin) deprecate tile_width and tile_height, just have tile_size now
2012-08-07 14:29:03 +02:00
* See also: vips_tiffsave().
*
* Returns: 0 on success, -1 on error.
*/
int
add @container option to dzsave
2014-04-25 11:26:23 +02:00
vips_dzsave( VipsImage *in, const char *name, ... )
more improvements to dzsave now writes x_files and x.dzi, as per spec (thanks Benjamin) deprecate tile_width and tile_height, just have tile_size now
2012-08-07 14:29:03 +02:00
{
va_list ap;
int result;
add @container option to dzsave
2014-04-25 11:26:23 +02:00
va_start( ap, name );
result = vips_call_split( "dzsave", ap, in, name );
more improvements to dzsave now writes x_files and x.dzi, as per spec (thanks Benjamin) deprecate tile_width and tile_height, just have tile_size now
2012-08-07 14:29:03 +02:00
va_end( ap );
return( result );
}
add dzsave_buffer zip only ... see https://github.com/jcupitt/libvips/issues/415
2016-10-15 20:01:34 +02:00
/**
Update GObject introspection annotations for libvips/foreign directory
2017-09-27 00:39:06 +02:00
* vips_dzsave_buffer: (method)
add dzsave_buffer zip only ... see https://github.com/jcupitt/libvips/issues/415
2016-10-15 20:01:34 +02:00
* @in: image to save
Update GObject introspection annotations for libvips/foreign directory
2017-09-27 00:39:06 +02:00
* @buf: (array length=len) (element-type guint8): return output buffer here
* @len: (type gsize): return output length here
add dzsave_buffer zip only ... see https://github.com/jcupitt/libvips/issues/415
2016-10-15 20:01:34 +02:00
* @...: %NULL-terminated list of optional named arguments
*
* Optional arguments:
*
add @id arg to dzsave So you can now set the base of the id property in info.json in IIIF output. Thanks regisrob. See https://github.com/libvips/libvips/issues/1531
2020-01-24 18:12:46 +01:00
* * @basename: %gchar base part of name
add dzsave_buffer zip only ... see https://github.com/jcupitt/libvips/issues/415
2016-10-15 20:01:34 +02:00
* * @layout: #VipsForeignDzLayout directory layout convention
add @id arg to dzsave So you can now set the base of the id property in info.json in IIIF output. Thanks regisrob. See https://github.com/libvips/libvips/issues/1531
2020-01-24 18:12:46 +01:00
* * @suffix: %gchar suffix for tiles
add dzsave_buffer zip only ... see https://github.com/jcupitt/libvips/issues/415
2016-10-15 20:01:34 +02:00
* * @overlap: %gint set tile overlap
* * @tile_size: %gint set tile size
* * @background: #VipsArrayDouble background colour
* * @depth: #VipsForeignDzDepth how deep to make the pyramid
* * @centre: %gboolean centre the tiles
* * @angle: #VipsAngle rotate the image by this much
* * @container: #VipsForeignDzContainer set container type
* * @compression: %gint zip deflate compression level
use -1 for do-nothing so that users can have skip-only-exact-blanks see https://github.com/libvips/libvips/issues/1185#issuecomment-449115143
2018-12-21 15:08:54 +01:00
* * @region_shrink: #VipsRegionShrink how to shrink each 2x2 region.
* * @skip_blanks: %gint skip tiles which are nearly equal to the background
add no_strip option to dzsave it's sometimes useful to have pyramid tiles with ICC profiles see https://github.com/libvips/libvips/issues/1451
2019-10-21 17:46:20 +02:00
* * @no_strip: %gboolean don't strip tiles
add @id arg to dzsave So you can now set the base of the id property in info.json in IIIF output. Thanks regisrob. See https://github.com/libvips/libvips/issues/1531
2020-01-24 18:12:46 +01:00
* * @id: %gchar id for IIIF properties
add dzsave_buffer zip only ... see https://github.com/jcupitt/libvips/issues/415
2016-10-15 20:01:34 +02:00
*
* As vips_dzsave(), but save to a memory buffer.
*
* Output is always in a zip container. Use @basename to set the name of the
* directory that the zip will create when unzipped.
*
* The address of the buffer is returned in @buf, the length of the buffer in
* @len. You are responsible for freeing the buffer with g_free() when you
* are done with it.
*
* See also: vips_dzsave(), vips_image_write_to_file().
*
* Returns: 0 on success, -1 on error.
*/
int
vips_dzsave_buffer( VipsImage *in, void **buf, size_t *len, ... )
{
va_list ap;
VipsArea *area;
int result;
area = NULL;
va_start( ap, len );
result = vips_call_split( "dzsave_buffer", ap, in, &area );
va_end( ap );
if( !result &&
area ) {
if( buf ) {
*buf = area->data;
area->free_fn = NULL;
}
if( len )
*len = area->length;
vips_area_unref( area );
}
return( result );
}
add dzsave to a target (#2718) * start adding dzsave to a target made a thing for gsf to write to a target * fix stray tabs in dzsave * fix dzsave write to "." Early versions of libgsf did not support writing to ".", so we had an ugly workaround, but this should now be OK. Fixing this ought to make write to target simple * dzsave_target compiles no idea if it works though * seems to work now creates stray 0 length files though, very odd * fix stray files from dzsave next: save to buffer is returning null * fix buffer flush in dzsave all tests pass! * update changelog
2022-05-20 19:38:46 +02:00
/**
* vips_dzsave_target: (method)
* @in: image to save
* @target: save image to this target
* @...: %NULL-terminated list of optional named arguments
*
* Optional arguments:
*
* * @basename: %gchar base part of name
* * @layout: #VipsForeignDzLayout directory layout convention
* * @suffix: %gchar suffix for tiles
* * @overlap: %gint set tile overlap
* * @tile_size: %gint set tile size
* * @background: #VipsArrayDouble background colour
* * @depth: #VipsForeignDzDepth how deep to make the pyramid
* * @centre: %gboolean centre the tiles
* * @angle: #VipsAngle rotate the image by this much
* * @container: #VipsForeignDzContainer set container type
* * @compression: %gint zip deflate compression level
* * @region_shrink: #VipsRegionShrink how to shrink each 2x2 region.
* * @skip_blanks: %gint skip tiles which are nearly equal to the background
* * @no_strip: %gboolean don't strip tiles
* * @id: %gchar id for IIIF properties
*
* As vips_dzsave(), but save to a target.
*
* See also: vips_dzsave(), vips_image_write_to_target().
*
* Returns: 0 on success, -1 on error.
*/
int
vips_dzsave_target( VipsImage *in, VipsTarget *target, ... )
{
va_list ap;
int result;
va_start( ap, target );
result = vips_call_split( "dzsave_target", ap, in, target );
va_end( ap );
return( result );
}