Returning a non-null value from the new `posts_pre_query` filter will cause
`WP_Query` to skip its database query, so that posts data can be provided from
elsewhere. This is useful in cases where post data may be mirrored in a
separate location, such as an external search application.
Developers should note that the `WP_Query` properties generally used to
calculate pagination - specifically, `found_posts` and `max_num_pages`, which
are determined by default in `set_found_posts()` - must be provided explicitly
when using the `posts_pre_query` filter; since `WP_Query` will not be
contacting the database, it will have no access to `SELECT FOUND_ROWS()`.
The `WP_Query` instance is passed to `posts_pre_query` by reference, so that
these properties can be set manually if needed.
Props jpdavoutian, tlovett1.
Fixes#36687.
git-svn-id: https://develop.svn.wordpress.org/trunk@37692 602fd350-edb4-49c9-b593-d223f7449a82
`wp_get_canonical_url()` encapsulates the URL logic of `rel_canonical()` to provide a consistent way to retrieve the canonical URL for a post.
The new filter `get_canonical_url` allows to customize the canonical URL.
Props joostdevalk, jipmoors, DrewAPicture, ocean90.
Fixes#36168.
git-svn-id: https://develop.svn.wordpress.org/trunk@37685 602fd350-edb4-49c9-b593-d223f7449a82
Allows the fetching of terms based on `term_taxonomy_id`, or an array of
`term_taxonomy_ids`.
Props spacedmonkey.
Fixes#37074.
git-svn-id: https://develop.svn.wordpress.org/trunk@37683 602fd350-edb4-49c9-b593-d223f7449a82
When an object with private properties is cast directly to an array, those properties are no longer available with their original keys.
Props @flixos90.
See #36717.
git-svn-id: https://develop.svn.wordpress.org/trunk@37667 602fd350-edb4-49c9-b593-d223f7449a82
* One test per method
* Clarify existing tests.
* Add test for passing a "blog slug" string to `get_blog_details()`.
* Shared fixture of sites.
* Reduce number of sites created to only those necessary.
* Remove unnecessary networks creation.
See #36566.
git-svn-id: https://develop.svn.wordpress.org/trunk@37666 602fd350-edb4-49c9-b593-d223f7449a82
* Convert existing tests into a data provider and clarify expectations.
* Add shared test fixtures in preparation for future tests.
This passes with the `wp_get_sites()` from 4.5 and the deprecated version in trunk.
See #36566.
git-svn-id: https://develop.svn.wordpress.org/trunk@37662 602fd350-edb4-49c9-b593-d223f7449a82
When custom pagination parameters are passed to `wp_list_comments()`, a
secondary query must be performed to fetch the proper comments. See [36157].
This query should show comments of the same `comment_status` as the default
query initialized in `comments_template()`: show only comments that are
approved, or those that are unapproved but belong to the current user.
Props smerriman.
Fixes#37048.
git-svn-id: https://develop.svn.wordpress.org/trunk@37655 602fd350-edb4-49c9-b593-d223f7449a82
To ensure a clean run, the test suite drops all tables before installing, by simply looping over the table list and dropping them if they exist. This works well for Core, but may fail when a plugin has created a table with foreign key constraints in a previous test run.
Many plugins choose to base their test suite on the Core setup, so making life easier for them is a plus, even if Core doesn't directly need this change.
Props javorszky.
Fixes#37046.
git-svn-id: https://develop.svn.wordpress.org/trunk@37654 602fd350-edb4-49c9-b593-d223f7449a82
Defer to the new `get_sites()` replacement, offering fresh (...or cached) `WP_Site` objects via the new `WP_Site_Query`.
Props flixos90.
Fixes#36994.
git-svn-id: https://develop.svn.wordpress.org/trunk@37653 602fd350-edb4-49c9-b593-d223f7449a82
`term_id` and `term_taxonomy_id` become offset when running the entire test
suite.
Introduced in [37647].
See #37009.
git-svn-id: https://develop.svn.wordpress.org/trunk@37649 602fd350-edb4-49c9-b593-d223f7449a82
The change can cause fatal errors under certain conditions, like when the subclass has a different function signature for `widget()` or doesn't even implement the method.
See #35981.
git-svn-id: https://develop.svn.wordpress.org/trunk@37648 602fd350-edb4-49c9-b593-d223f7449a82
Move the REST API JSONP callback validation check into a separate function named `wp_check_jsonp_callback()`. This allows plugins to use the built-in validation when handling JSONP callbacks.
Extremely Important Note: If you send JSONP in your custom response, make sure you prefix the response with `/**/`. This will mitigate the Rosetta Flash exploit. You should also send the `X-Content-Type-Options:nosniff` header, or even better, use the REST API infrastructure.
Props rmccue.
Fixes#28523.
git-svn-id: https://develop.svn.wordpress.org/trunk@37646 602fd350-edb4-49c9-b593-d223f7449a82
[37623] used the wrong parameter name (count=true instead of fields=count).
For greater flexibility and forward compatibility with other potential changes
to the return value of `get_terms()`, we now do a looser check: any non-array
value is excluded from the filter.
Fixes#36992.
git-svn-id: https://develop.svn.wordpress.org/trunk@37634 602fd350-edb4-49c9-b593-d223f7449a82
Different tests are used for subdomain and subdirectory installs as domain and path are searched differently for each. Only trailing wildcard searches are tested because leading wildcards are not yet supported.
See #36675.
git-svn-id: https://develop.svn.wordpress.org/trunk@37633 602fd350-edb4-49c9-b593-d223f7449a82
- We already match URLs on their own line, add another regex to match URLs in their own paragraphs.
- Always exclude the `\s<>"` characters when matching.
- Add more unit tests.
Props iseulde, azaozz.
Fixes#25387.
git-svn-id: https://develop.svn.wordpress.org/trunk@37627 602fd350-edb4-49c9-b593-d223f7449a82
Hierarchical comment queries work by first fetching the IDs of top-level
comments, and then filling the descendant tree one level at a time based on the
top-level results. When top-level comment IDs are found in the cache,
`WP_Comment_Query` does not generate the SQL used to fetch these comments. In
this case, the `fill_descendants()` query does not have enough information
to fill children. As a result, descendant comments were failing to be filled
in cases where the top-level comments were found in the cache.
This was a minor bug previously, because comment caches were not maintained
between pageloads. Since comment caches are now persistent [37613], the problem
becomes evident anywhere that a persistent object cache is in use.
The solution is to cache parent-child relationships, so that when top-level
comments are found in the cache, descendant comments should be found there as
well.
Fixes#36487.
git-svn-id: https://develop.svn.wordpress.org/trunk@37625 602fd350-edb4-49c9-b593-d223f7449a82
Use of the 'get_terms' filter was consolidated in [37572], with the
introduction of `WP_Term_Query`. At that time, the result of 'count=true'
queries began being filtered by 'get_terms'. This breaks existing 'get_terms'
callbacks, which often assume that the returned value will be an array or a
`WP_Error` object.
Props JustinSainton.
Fixes#36992.
git-svn-id: https://develop.svn.wordpress.org/trunk@37623 602fd350-edb4-49c9-b593-d223f7449a82
The logic for determining the appropriate character set and collation to use is becoming more complex, particularly with the recent additions of [37522] and [37523]. As `init_charset()` has side effects, and makes use of constants instead of parameters, it's not possible to unit test this logic.
This commit splits the logic part of `init_charset()` out into a new method, `wpdb::determine_charset()`, along with appropriate unit tests.
See #32105, #37522.
Fixes#36917.
git-svn-id: https://develop.svn.wordpress.org/trunk@37601 602fd350-edb4-49c9-b593-d223f7449a82
Prior to [37572], arguments passed to `get_terms()` were passed immediately
through `wp_parse_args()`, which made it possible to pass arguments as a
querystring (`hide_empty=0`) rather than an array
(`array( 'hide_empty' => false )`). [37572] moved default argument parsing
into `WP_Term_Query`, while assuming that arguments passed to `get_terms()`
would be formatted as an array.
To provide compatibility, we now parse all args passed to `get_terms()` into
an array before processing.
See #35381.
git-svn-id: https://develop.svn.wordpress.org/trunk@37599 602fd350-edb4-49c9-b593-d223f7449a82
`CHAR` is redundant, since the `meta_value` column is `LONGTEXT`. Meanwhile,
use of `CAST()` causes MySQL to ignore any index that the administrator may
have added to the column.
A number of automated tests were doing searches for `CAST` in the SQL strings
generated by `WP_Meta_Query` (for reasons unrelated to the `CAST()` behavior).
These tests have been updated to expect the new query format.
Props ericlewis.
Fixes#36625.
git-svn-id: https://develop.svn.wordpress.org/trunk@37594 602fd350-edb4-49c9-b593-d223f7449a82
Since [37573], object relationship caches (`{$taxonomy}_relationships`)
contain term IDs rather than term objects. See #36814. As such, it's no longer
necessary to clear these caches when a term is updated; none of the data that's
changed on update (name, description, count, etc) is stored in the relationship
cache.
Fixes#36251.
git-svn-id: https://develop.svn.wordpress.org/trunk@37593 602fd350-edb4-49c9-b593-d223f7449a82
The process of lazy-loading can be resource intensive for object that have
terms in large numbers of taxonomies and are running a persistent object cache.
This new parameter allows the feature to be disabled in these cases.
Props DBrumbaugh10Up.
See #36953.
git-svn-id: https://develop.svn.wordpress.org/trunk@37589 602fd350-edb4-49c9-b593-d223f7449a82
In order to allow non-web initializations of WordPress (such as through wp-cli) to modify things like the check for maintenance mode, plugins.php and the associated functions must be available much earlier. The use of these functions earlier than the loading of plugins is not recommended in most use cases.
Fixes#36819. See #34936.
Props jorbin, danielbachhuber for documentation.
git-svn-id: https://develop.svn.wordpress.org/trunk@37588 602fd350-edb4-49c9-b593-d223f7449a82
`dbDelta()` compares the index definitions against the result of `SHOW INDEX FROM $table_name`. This requires a specific format so indices are not unnecessarily re-created. This format wasn't ensured, until now.
* Parse the raw index definition to extract the type, name and columns so a normalized definition can be built (#20263, #34873).
* Standardize on uppercase types (#34871) and on 'KEY'. 'INDEX' is only a synonym for 'KEY'.
* Escape index names with backticks (#20263).
* Normalize columns: Ignore ASC and DESC definitions (#34959), remove whitespaces (#34869) and escape column names with backticks (#20263).
* Add backticks to all index change queries (#20263).
Props ocean90, pento, kurtpayne.
Fixes#20263, #34869, #34871, #34873, #34959.
git-svn-id: https://develop.svn.wordpress.org/trunk@37583 602fd350-edb4-49c9-b593-d223f7449a82
`dbDelta()` already supported spatial fields (by virtue of not checking field types), so it's nice to round that out with spatial key support, too.
Fixes#36948.
git-svn-id: https://develop.svn.wordpress.org/trunk@37574 602fd350-edb4-49c9-b593-d223f7449a82
Previously, objects containing all data about a term were stored in each
object's term cache. Besides being wasteful, this approach caused invalidation
issues, as when a modified term count required a flush for all objects
belonging to the term.
Backward compatibility is maintained for plugins that continue to put object
data directly into the `{$taxonomy}_relationships` cache bucket.
Fixes#36814.
git-svn-id: https://develop.svn.wordpress.org/trunk@37573 602fd350-edb4-49c9-b593-d223f7449a82
[34529] introduced logic intended to prime the termmeta cache for certain
values of the `fields` parameter. There were a few bugs:
* The `all_with_object_id` param was misspelled.
* `term_id` was used instead of `ids`.
* The values being passed to `update_termmeta_cache()` in the case where `fields=ids` was not correct.
All of these would result in a failure to pre-fetch termmeta in some cases.
Props dlh.
Fixes#36932.
git-svn-id: https://develop.svn.wordpress.org/trunk@37567 602fd350-edb4-49c9-b593-d223f7449a82
This changeset adds a unit test factory so that bookmark/link fixtures can be
created during tests.
Why are we writing tests for functionality that has been deprecated for years?
Because it's the Right Thing to Do.
See #18356.
git-svn-id: https://develop.svn.wordpress.org/trunk@37563 602fd350-edb4-49c9-b593-d223f7449a82
When upgrading to `utf8mb4`, `TEXT` fields will be upgraded to `MEDIUMTEXT` (and likewise for all other `*TEXT` and `*BLOB` fields). This is to allow for the additional space requirements of `utf8mb4`.
On the subsequent upgrade, after the `utf8mb4` upgrade, `dbDelta()` would try and downgrade the fields to their original size again. At best, this it a waste of time, at worst, this could truncate any data larger than the original size. There's no harm in leaving them at their original size, so let's do that.
Fixes#36748.
git-svn-id: https://develop.svn.wordpress.org/trunk@37525 602fd350-edb4-49c9-b593-d223f7449a82
`wpdb::init_charset()` doesn't lend itself to being tested, so the unit test added in [37521] won't work under most circumstances.
See #32405.
git-svn-id: https://develop.svn.wordpress.org/trunk@37522 602fd350-edb4-49c9-b593-d223f7449a82
Some sites prefer to use locale-specific location settings. For example, the Swedish WordPress package use `utf8_swedish_ci`, instead of `utf8_unicode_ci`. When upgrading the connection to `utf8mb4`, we were overriding this to be `utf8mb4_unicode_ci`, instead of maintaining the use of the `_swedish_ci` variant.
The locale-specific collations do have extra collation rules just for that language, so it's useful to maintain compatibility.
Fixes#32405.
git-svn-id: https://develop.svn.wordpress.org/trunk@37521 602fd350-edb4-49c9-b593-d223f7449a82
Without a fixture that does not match the search term, it's possible for the
test to pass even if the search clause isn't built properly.
See #13992, #35381.
git-svn-id: https://develop.svn.wordpress.org/trunk@37519 602fd350-edb4-49c9-b593-d223f7449a82
