This is a mess of a function, and the tests show it. There are lots
of side-effects and context-sensitivity, which can be addressed in
later work. The interface with ReferenceStack is too wide.
Change-Id: I00cab2a555b2a9efd32d937979cd722d43ac1005
I was able to track this code down to I093d85d from 2012, which was done
right after the ParserAfterParse hook was introduced. I believe the
redundant code path was left to keep the Cite extension compatible with
old MediaWiki versions that did not had this hook yet.
I also noticed this code path is most probably entirely redundant with
the current version of MediaWiki. The *only* thing this code does is
blocking the ParserBeforeTidy hook from doing the same thing a second
time if the ParserAfterParse hook was called before. But it does *not*
block any other compination, e.g. if the two hooks are called the other
way around, or the same hook twice.
In core, it looks like it is impossible for the ParserBeforeTidy hook
being fired without the ParserAfterParse hook being fired before. If this
is true, this is in fact dead code.
Change-Id: Iacf8b600c7abdeaf89c22c2fc31e646f57245e47
Encapsulate the language interfaces, this will be used to replace
global wfMessage calls in future patches.
Change-Id: I7857f3e5154626e0b29977610b81103d91615f65
The new extends="…" feature is using numbers like "1.2". These should be
localized in languages like Hebrew that uses other symbols for the digits.
But the "." should not change.
The existing feature when a <ref> is reused multiple times does have the
same "issue". But it seems this is intentional, because it is covered by
a test. Note this is not visible in German, because German uses custom
labels "a", "b", and so on.
This patch also improves the so called "smoke" tests and makes one cover
numbers up to "1,10" for a <ref> that is reused that often.
Bug: T239725
Change-Id: Iffcb56e1c7be09cefed9dabb1d6391eb6ad995ce
If `extends` is encountered before the parent ref, we reserve the
sequence number and leave a placeholder to record the link between
ref name and number. This is necessary to render a list like,
"[1] [2.1] [2]", or to use subreferencing when the parent ref is
declared in the references tag.
When a placeholder is encountered during references section rendering,
it means that the parent was never declared.
Change-Id: I611cd1d73f775908926a803fae90d039ce122ab6
Pass the full ref structure from ReferenceStack to FootnoteMarkFormatter,
to give it control over the final rendering. This is aligned with how
the FootnoteBodyFormatter directly scans over groupRefs.
Change-Id: I3294fd9366f01daa4250a5d481f4adbae84c72b1
This was carrying the entire footnote marker, but subreferences need
to extract just the first (group ref sequence) part. Storing number
and extendsIndex in two separate fields gives us more flexibility
during rendering, for example these might use two different symbol sets.
Change-Id: I75bd6644c336036f9e84ba91e1c35e05bc1ca7f3
This was a bug which would affect book references, if the same group
and parent ref name combination occur twice in an article.
Change-Id: I608f58aac0cec31c8650835fc80195a87bc851d3
Validation blocks (name==null && text==null), so it should not be a
test case. Give the text a non-null value.
Also adds a check for missing test data.
Change-Id: I0f02206e2221805f5a2f8eaa163ed237cfb8d777
This patch does two things:
* Add strict PHP 7 type hints to most code.
* Narrow the interface of the checkRefsNoReferences() method to not
require a ParserOptions object any more.
Change-Id: I91c6a2d9b76915d7677a3f735ee8e054c898fcc5
There was a call in the API that was *not* using normalizeKey(). Now
that the API is gone, we can inline this.
This patch also contains a bunch of cleanups that might already been
resolved in the previous patches.
Change-Id: Id3767b5830268c8cfe9c10efabfa4a31e9dafeb8
Forked from Icd933fc983.
Bugs and unimplemented features are documented as TODOs in the parser test
fixtures.
Bug: T237241
Change-Id: I9427e025ea0bcf2fa24fd539a775429cc64767cc
This API was never used in Wikimedia production, and would have caused
performance problems. Removing the dead code will simplify our refactoring.
Bug: T238195
Change-Id: I7088f257ec034c0d089e0abdaa5a739910598300
I noticed a possible issue related to the $this->refSequence counter
in the patch Ida9612d. Some of these counters might get messes up, but
there was never a test that checked what will happen to the *next*
reference then.
I checked the test cases in this patch with a very old version of the
codebase.
Change-Id: If6e56f727dce5d0e5e38e048e602437597248a42
We realized the trim() are not needed. This does not leave much behind
in the existing refArg() method, except that it checks for unknown keys.
I tried a few strategies and ended using the pretty new possibility to
have keys in list(), as well as use [] instead of list(). Both is
supported since PHP 7.1.
Change-Id: I569bfa14e68b64402519bd39022c197553881dde
We noticed the group="…" attribute was the only one that was not
trimmed. Does this mean it was possible to have two groups "a" and
" a"? It turns out: no. This was never possible because the parser
already trims all attributes before calling this code.
I tried to come up with the worst possible test case, but it succeeds,
even with very old versions of this codebase.
I suggest to remove the extra trimming from this codebase and rely on
what the parser provides.
Note the content is special and *not* trimmed by default.
Change-Id: Idff015447d7156ba7b5c03a5c423f199a71349f2
These exist two times, one time in the unit/ folder as a unit test, and
another time in the parent folder as an integration test. This confused
me already several times.
Change-Id: I147b8af8a7edba2582496468b4878faecc6d8110
Functional changes:
* hasGroup() will return false when a group exists, but is empty. This
is in line with what other methods like getGroups() already do.
Shouldn't have any effect on the existing code, but feels more clean
and consistent.
* getGroupRefs() won't fail any more when asked for an unknown group.
Tests:
* Add missing @covers for the constructor.
* Simplify test setup by always returning a spy. All tests need it
anyway.
* Cover 3 more methods.
Change-Id: Ie93e9af6258b757d842b30b0b059344733aad434
That was annoying me. Since we're passing a bare list, alphabetical
order helps make the code and tests readable.
Change-Id: I6384094e429e0e2a6fa810fdc28ae0643a0ccf7c
Most of this state is used to manage interactions with other state,
and encapsulation allows us to hide data structures and access behind
self-explanatory function names.
The interface is still much wider than I'd like, but it can be improved in
future work.
There is one small behavior change in here: in the `follows` edge case
demonstrated by I3bdf26fd14, we prepend if the splice point cannot be
used because it has a non-numeric key. I believe this was the original
intention of the logic, and is how the numeric case behaves. I've verified
that when array_splice throws a warning about non-numeric key, it fails to
add anything to the original array, so the broken follows ref disappeared.
Bug: T237241
Change-Id: I091a0b71ee9aa78e841c2e328018e886a7217715
I realized especially the method name html() was wrong. It does not
return HTML. What it returns is still wikitext and must still be parsed.
It only applies some early steps of the parsing process, e.g. expanding
extension <tags>.
Change-Id: I2c403a77eef843940f34f0933e4bfe58e6200ce5
* This fixes the refArg() function. If there is nothing wrong with the
follow="…" attribute, it should not return null.
* However, *everything* is false if an unknown error (e.g. an unknown
attribute) occurs.
* A trivial check for `if ( $follow )` is fine because all keys are
guaranteed to not be the string "0".
Change-Id: Ia4e37781e01db1ee6615ffc30bb68e47023c6634
One of the test cases was duplicated, but a lot of the possible code
paths never had tests, including the happy code path!
I found this issue while trying to rework some of the more confusing
loops in this codebase. These changes are still part of this patch. All
loops still do the same as before, but are (I hope) more readable now.
Bug: T238187
Change-Id: I85baeadd9b149025a14c7522bcc4182339c66972
… and make the error message for bad dir="…" shorter and more to the
point.
Now I understand why the error reporting was not done when $text was
empty: the error was actually appended to $text, which messes with
everything else that also works with the $text variable! This even
includes the API. This error message was exposed via the API. That was
certainly a bug.
With this patch, all error checking for the dir="…" attribute is now
done way down, when rendering the <references> section.
Note this also fixes a bug where the dir="…" was *not* rendered when
previewing a section.
Change-Id: I4ab0cb510973ed879c606bfaa394aacc91129854
This fixes a whole bunch of inconsistencies:
* The dir attribute is now trimmed, as most others already are. This is
an actual user-facing change.
* The internal representation is now false in case the value was invalid,
not an empty string any more.
* Null means the attribute was not present. This is now always used,
even in the return values that are meant to represent an error state. No
existing behavior changes.
* The internal representation does not contain an HTML snippet any more,
but the raw value "ltr" or "rtl", or null. Note this might influence the
API, because the API actually exposes the internal representation.
However, we are pretty sure the API is not used anywhere. Even if,
exposing HTML code was most certainly an unwanted and unexpected effect
of the patch that introduced the dir attribute. This does make this a
bugfix, I would argue.
Change-Id: Ic385d9ab36fa0545c374d3d63063028ae4e449d4
This patch does intentionally not touch any file name. Some of the
file names are a little weird now, e.g. \Cite\Cite. These can more
easily be renamed in later patches.
I used https://codesearch.wmflabs.org/search/?q=new%20Cite%5C( and it
looks like this code is not used anywhere else.
Change-Id: I5f93a224e9cacf45b7a0d68c216a78723364dd96
The use case we care about is this:
<ref extends="some_book"> </ref>
It doesn't make sense that works, but the following doesn't:
<ref extends="some_book"></ref>
We decided that both need to behave the same.
For consistency this patch is applying the same change to all references,
no matter if they use the extends attribute or not. This is an actual
change and might make existing wikitext render differently. However, I
would like to argue that all wikitext that was using this was broken. The
effect of a <ref> </ref> with some whitespace is that the <references>
section at the end of the article will contain – well – an empty footnote.
Bug: T237241
Change-Id: Iaee35583eabcb416b0a06849b89ebbfb0fb7fef9
Note this codebase appears to be dual-licensed. Some files mention MIT,
but extension.json and some other files mention GPL.
Since WMDE typically uses GPL, I will continue to mark the files we
created as such.
Change-Id: I126da10f7fb13a6d4c99e96e72d024b2e5ecee06
The main motivation here is to cover the fallback code that was moved
in I20c814d. At some point we might touch this code again.
Bug: T238194
Change-Id: I0ab8a34b09790f42b10376eb3730c3b3c4ef53d2
1. Most existing CiteTests can be unit tests. They run so much faster
this way.
2. I modified some test cases to cover all trim() in the code.
3. The strict type hint in CiteHooks is removed because the parameter
is not used. Having a hard type hint for what is effectively dead code
makes the code more brittle for changes done outside of this codebase.
Change-Id: I1bff1d6e02d9ef17d5e6b66aeec3ee42bba99cf4
This fixes a series of issues:
* There is nothing about a "frame" in the Cite class any more.
* There is no addModules() call in the Cite class any more.
Change-Id: I20c814d46c26825c5c07eab0a5586de3a531eee7
To be honest I don't get why this lazy registration was done in the
first place. None of the 4 other hooks should ever be called before
the ParserFirstCallInit hook got called.
Also, under which circumstances can the ParserFirstCallInit hook be
called more than once?
Both scenarios would be wrong, as far as I'm concerned. Either I'm
missing something, or this code can indeed be simplified. Maybe it was
something to make it more compatible with older MediaWiki versions?
The only reason I can think of is: in all situations that do not
involve a parser, having the 4 extra hooks registered is pointless.
Does this waste space and/or runtime in the $wgHooks registry?
Change-Id: I5ef1495f4ce7bce940fa5f8e700af3d2c4851a01
Can use a shortcut where we pass the expected value directly. Verified
that we're still asserting equality.
Change-Id: I63512488c50e599df23d5dae2a5064218e311e90
Note it doesn't make a difference if this is behind the feature flag or
not. It should always be forbidden, and in fact is: Either the follows
attribute is unknown, or the combination is forbidden.
Bug: T236256
Change-Id: Iebbb2d1d5bab183ab0590b8a7a7f6e79d319b72c
What we find critical is:
* That all tests relevant for book referencing are in a separate file.
* That unimplemented stuff is marked with TODOs.
Not having to move tests to another file allows for nice diffs.
I tried to order the tests as good as I could. E.g. have all tests with
a group="…" next to each other, followed by all with a follow="…".
Change-Id: Idc1d9e7843b341235ab3d8ebe398e01946eb1845
Any time the book referencing attribute is used in a page,
permanently tag that page with the `ref-extends` property, so
that it can be watched and cleaned up if necessary.
Bug: T237531
Change-Id: Ice5d9d8f7a305702cdc7c2a55d4147c4f79b5881
Incremental patch which extracts the refines attribute from the tag.
Doing this now to allow the calling function to have responsibility
for doing something with the attribute value.
Bug: T237531
Change-Id: I59bb409bedd8e6ed06268e705e02e8ffb45b1f0e
Encapsulate the feature tests in dedicated files. These are picked
up by the test runner for matching glob `tests/parser/*.txt`, as can
be shown by,
phpunit.php --testsuite parsertests --filter=bookRef
Also adds TODO comments to some tests, documenting how the current output
will not match the fully implemented code's results.
Bug: T236256
Change-Id: Ie3e769c84856256180754aeff417da893a84b479
These tests document the current status quo, and are meant to change
with every patch that makes the code for refined references more
feature complete.
Bug: T236256
Change-Id: I8c11b1decc36b86e7f7d1919cc39d0c16a200055
Note this is intentionally testing a private method. As of now, the
code is so heavily entangled, it's not yet possible to test individual
aspects without calling private methods. The plan is to slowly increase
the overall test coverage, and the start restructuring the code as
necessary.
Change-Id: Ib3b01bddaffd0469fb66979c67c8114a5807df6d
Allows the "refines" attribute when the feature flag is set, but doesn't
render. This is part of our rollback strategy, so that we aren't left
with invalid wikitext in case of undeployment.
Bug: T236257
Change-Id: I936be0e62dccb46caeb84162d2c5166956fd9916
This change allows to change the editsection HTML by
I305e3313ca2f931a2ea9cee34194b8cb93b90b0e without failing in Jenkins.
The parser test gets restored in the new format by the follow-up change
Ibb4341b405f0d6fa6883c992c5dd3a9e594c9efc.
Change-Id: I337d7f7c0cd134a3766343565e2c30edf1d70f7e
Adding a step to wait for loaded modules. This was manily taken from
the Popups-extension browser test, but with a fixed chain for the
results.
Let's hope we do not need to pause with this.
See as well I274bdee0b3c39c418a2b61881d56f89889c53485
Bug: T220318
Change-Id: I5bdc1951a23a51e9a4deea69af2bfc96843f5a1d
We can resolve this bug by either replacing the bogus "return false"
with the intended "return [ false, … ]". Or rely on the code a few
lines below that also bails out with a "return [ false, … ]" when to
many parameters ($cnt is not 0 then) are present. The tests prove both
solutions are equally valid.
Bug: T211576
Change-Id: Iadd55c134dede7042cfd152c69bc8f27b59d8912
This is split from I642d38e and does nothing but adding test cases
that document the current (broken) behavior.
Bug: T211576
Change-Id: Iee313d26e7bed6deb34101e37736a1c697947905
This change does have two consequences:
1. A few more whitespace characters act as separators. This should not
have any consequence in real life situations, and is mainly done to
make the code easier to read and less surprising.
2. Sequences of two or more whitespace characters previously resulted
in partly *empty* results. This was a potential source of errors. The
additional + fixes this.
Change-Id: Ib58326109c740dd0cbd05d8fddb4af2145f232fe
This also includes putting the backling tests in their
own file due to the common setup needed.
Bug: T205271
Change-Id: Ica98c31bf0fddde9eb80f9c53b3dc089cb60e30c
Core sanitizes link targets and removes double spaces and underscores.
But the corresponding id="…" attributes are not sanitized the same
way. This results in broken links. This patch is not perfect (two
references with name="a_b" and name="a__b" will conflict), but the
best solution I can think of at the moment.
Bug: T184912
Change-Id: I9dbc916ad99269517d84c8ffb8581628d44a9f4e
This adds a test which checks if the extension appears on the
Special:Version page.
This is an initial step for adding more browser tests to check
changes done in I26fe41c328157233cc5b06d38d2ba0f7b036a853
Change-Id: I9a9d1cd2a25277f2c430f4e80b51b72c1621f91b
Adding option for dir attribute in ref tags. The value must be a valid
direction ('ltr' or 'rtl', case insensitive) or the direction will be
stripped out.
The directionality of the li element is set using a css class accordingly.
Bug: T15673
Change-Id: Iff480bc8cc4f81403b310e8efecd43e29d1d4449
Don't use the \Database alias, use the namespaced version when calling
Database::getCacheSetOptions.
And document why the remaining issue is suppressed.
Change-Id: I80a102f2e82efedcfa999d8e714bfe049263ffeb
When using HTML5 ids, we need to take greater care to properly escape the
id (or derived strings) before passing them back through
Parser::recursiveTagParse().
Bug: T176170
Change-Id: I89a4f8ba24b867f2d5ccdc2bf9a4312ab9b385a9
This is based on the popular 'count' parameter from Template:Reflist on
English Wikipedia, which has also been adopted by many other wikis.
That template's 'count' parameter allows maximum flexibility on a per-
page basis. This was important because the template can't know how many
references the list will contain. Users typically manually add (and
later, increment) the 'count' parameter when the list exceeds a certain
threshold.
The template currently sets an exact column count (via the CSS3
property `column-count`).
This patch improves on that by instead using the closely related CSS3
`column-width` property. This automatically derives the column count
based on the available space in the browser window. It will thus create
two or three columns on a typical desktop screen, and two or no columns
on a mobile device.
The specified width is the minimum width of a column. This ensures that
the list is not split when rendered on a narrow screen or mobile device.
It also hooks into the raw list before parsing and adds the class only
when the list will contain more than a certain number of items. This
prevents very short lists from being split into multiple columns.
Templates like Template:Reflist on English Wikipedia currently are not
able to set inline styles on the list element directly, which is why
they set it on a `<div>` wrapping the `<references />` output. Because
of this, the feature of the Cite extension must not be enabled at the
same time, as that would result in both the template's wrapper and the
references list being split. The end result would involve sitations with
three columns split in four sub-columns, creating a complicated mess of
nine intermixed columns.
To provide a smooth migration for wikis, this feature can be disabled by
default using `$wgCiteResponsiveReferences = false`. Each individual
template createing reference list can then be migrated, by removing the
wrapper column styles and instead settting the new "responsive"
attribute, like so: `<references responsive />`.
Once any conflicting templates have been migrated, the default for the
wiki can be swapped by setting `$wgCiteResponsiveReferences = true`.
If wikis wish for some templates to keep their custom column splitting
behaviour, templates can also opt-out by setting `responsive="0"`, which
will make sure that it will keep behaving the current way even after the
feature becomes enabled by default for the wiki.
In summary, when disabled by default, pages can opt into this system
with `<references responsive />`. When enabled by default, pages can opt
out of the system with `<references responsive=0 />`.
* Deprecate cite_references_prefix/cite_references_suffix.
This message is rarely used and opens up compatibility hazards.
It was already removed by Parsoid, but the PHP implementation
still had it. It's typically used to add inline styles to the
wrapper which is more appropiately done in Common.css (or
obsoleted as part of the skin or Cite extenion itself nowadays
depending on what style in question).
It was also a HTML-style message with separated open and close
segments, which is an anti-pattern in itself.
* Declare module target explicitly and include mobile. The absence of
this stylesheet caused subtle BiDi/RTL bugs on mobile.
Bug: T33597
Change-Id: Ia535f9b722e825e71e792b36356febc3bd444387
