This piece of code is only relevant in case when:
- the intro section is requested (either in plaintext or html);
- the parse result for the full page is available in the parser cache;
- the full extract is not available in the TextExtracts WAN cache;
- the intro is also not available in the TextExtracts WAN cache.
In this case getFirstSection() is called with the parser output,
which is different from the the convertText() output it is called
with in other code paths, and still contains <h*> tags. A quick
regex is used to extract the first section. This stops at any <h2>.
A TOC also contains a <h2> (which will be removed later via
$wgExtractsRemoveClasses). This one needs to be ignored in case
the TOC is placed before the first section using e.g. the __TOC__
keyword.
The patch changes the regex so it ignores a h2 with
id="mw-toc-heading", but keeps working in plaintext mode when <h*>
tags are not present (the code path when the intro section is
requested, and the full extract is available in the TextExtracts
WAN cache but the intro extract isn't).
Bug: T269967
Change-Id: I0a495d06cf1725744e556e81f17047fb53f53521
Fixes:
1) TextExtracts\Test\ApiQueryExtractsTest::testMemCacheHelpers
strtr(): Passing null to parameter #1 ($string) of type string is deprecated
Change-Id: Id39cb89bb5d49fbea543181da678b191f70311ac
All services can be injected by bumping minimum
version of mediawiki (1.34 would be enough, but
since that is no longer supported require 1.35)
Change-Id: I8bb1573a02932ef5f2871606e94a41afe073fd00
When the text is short enough to be returned as it is, it's very
confusing to see it with an ellipsis added at the end. There is
no more text. It should not look like there is more text.
Change-Id: I7ef205fde6c358a1cbcbb41346a1c9e2a856d8fd
Remove use of deprecated MWTidy::isEnabled() and internal
MWTidy::singleton() methods. See I3584181070da7ed4888beaaf04e083114aca1eab
for context.
Bug: T198214
Change-Id: I511068cc7b2398773a837f66e08def206cbb5626
composer:
* mediawiki/mediawiki-codesniffer: 28.0.0 → 29.0.0
The following sniffs are failing and were disabled:
* MediaWiki.Commenting.FunctionComment.MissingDocumentationPrivate
npm:
* eslint-config-wikimedia: 0.12.0 → 0.15.0
* grunt-eslint: 21.0.0 → 22.0.0
Additional changes:
* Also sorted "composer fix" command to run phpcbf last.
* Removing manual reportUnusedDisableDirectives for eslint.
Change-Id: I351f0a333fd5f06e47f0748aa25cb3fff63cc67f
By turning the (?:…) into (?=…) they become lookaheads and are not
part of the returned string in $tail any more. This is exactly what we
want here. All we want is to *know* if the dot, question or exclamation
mark is followed by a space. But we don't need the space captured.
Change-Id: I4be715c4c084165e5ab25da77609f12ffce4d385
I argue that the code fixing unclosed HTML tags is – even if optional –
an integral part of the code that potentially breaks these HTML tags in
the first place. Notice how much code disappears in the ApiQueryExtracts
class.
Additionally, the new approach uses an interface instead of a static
function call that is impossible to mock and hard to test.
Change-Id: Ic1a65995f4dba11d060a8738d642905cbfc79271
Note how only two files mentioned the license before. For consistency
it should be either all or none. Both solutions would be possible. Even
*not* mentioning the license anywhere in these files would be fine from
a legal perspective, as long as the relevant file COPYING is still
there in the root folder of this extension.
The overly long "deed" text does not serve much of a purpose. It's not a
complete, legally relevant license text. It's hard to read as the fact
this is "GPL2+" is surprisingly hard to find. The @license tag solves
these problems, and is recognized by documentation generators.
Change-Id: I7844be0c5f4f3d7562156cd9f34fe466552a9c9d
This is a straightforward baseline patch that does nothing but moving
existing code around, without touching it. I'm not even trying to
remove the "static" keyword. The actual refactoring will be done in
the next patch. I hope with this the changes I do in the refactoring
become more visible and much easier to review.
Change-Id: Idba859ec0c24f3622ea8fb8d7a9b11843d1e3827
This is especially covering:
* The behavior on different kinds of HTML encoded characters.
* Trimming, or more precisely the fact that the formatter currently does
not do all the trimming. There is a trim() call in the test that should
not be there. This will be reworked in a following patch.
Additionally:
* Prefer strict assertSame() where possible.
* Remove two assertions that don't help making the test more safe.
* Utilize the new `// phpcs:…` syntax.
Change-Id: Ic361f2644300dcef3fc972a2dc2ae19ad34fa513
This patch does two things:
* Add a testMemCacheHelpers() test for the setCache(), getFromCache(),
as well as cacheKey() methods.
* Simplify the existing test setup utilizing the PHPUnit4And6Compat
layer. I'm also replacing the Config mock with an actual HashConfig,
because that's a really trivial object that does not need to be mocked.
Change-Id: I7e8a62651a53919d338a7d921f548e941b671670
Instead of trying to come up with a (probably hypercomplex) integration
test for the entire API module, I start to test a few more critical
details. Yes, this is now testing private implementation details. This
is meant as a migration path. See, it's hard to test code like this
without rewriting it, and hard to rewrite it without tests. This patch
aims to give us a little more confidence for future rewrites.
Change-Id: I0405e13eba31124fcb5af83159d7438fe0ad9879
For a new ParserOptions object, ->setEditSection( true ) is already the
default. Since it's being deprecated in MediaWiki core change
Ied5fe1a61, just remove the call.
Change-Id: Id517a8746d4cc4b0b26a98a74244e4d8c711358c
