Note: I'm not sure if this code is copied from somewhere else. If it
is this change is better done in the source.
Change-Id: Ic84be9b8b52a190eb5a41f2c7ff8fd6c04724318
Personally I find them confusing because there is apparently no way
to "stop" an array_walk(), and that's why it constantly needs to
check if it even needs to run. Why not use a syntax we can actually
stop? I find this more readable and it makes the code actually run
faster. On average 50% of the iterations get skipped now.
Change-Id: I47b6d5b8f4ad5c9d0f47b389c638196e734809bb
We can now serve the JSON data almost directly (although, curiously,
we need a main file in the module that is JavaScript, not JSON),
and without using global state in the JS code.
Change-Id: I203b3e396b28f4b4ef49c3c0918a4025082308c4
Removes functionality to
* select PNG rendering mode
* automated fallback to PNG images for SVG rendering
* PNG mode related class and tests
However, PNG images received from mathoid / restbase are still stored.
Bug: T311620
Change-Id: I595926027433182cc0396570bc3f1ce0cd2cafb3
Needed for I869af06896b9757af18488b916211c5a41a8c563, where I am
trying to change LanguageFactory in MediaWiki core not to use
MWException.
I truly feel mocked after working with this code. See if you can find
the two lines in this diff where the meaningful changes are.
Change-Id: Ifcb31dbb7113ce57526f06558cde2abedee317d7
TexUtil gets instanciated and loads a json for each node in Literal and Fun1. This is not necessary since it takes much RAM and time to load the corresponding JSON from files.
Bug: T321599
Change-Id: I91855ebc4437f075302ae13d367ca4902cc0f3ba
For development history of this changeset see:
Id96a4b1b55e3959aab81f4ba436c5ac125f2a1bb
Bug: T312528
Change-Id: I61cfdbd63f8d50b072ada05927a134686fdd53d3
For development history of this changeset see:
Id96a4b1b55e3959aab81f4ba436c5ac125f2a1bb
Bug: T312528
Change-Id: I0e64a3fc4b4556bc4ea6271bd6b778ed2930ab1a
3 backslashes work as well according to
https://stackoverflow.com/a/28063081
However, it might be confusing as the behaviour would change in double qouted strings. Also mw-core seems to mostly use 4 backslashes.
https: //codesearch.wmcloud.org/core/?q=%5B%5E%5C%5C%5D%5C%5C%5C%5C%5C%5C%5B%5E%5C%5C%5D&i=nope&files=.php%24&excludeFiles=&repos=
Change-Id: I9be37c386f8ca0f9d0a7be641484654cf9a2f7fa
Texutil related functionalities and tests will come in other changeset.
Related code:
fb56991251/lib/nodes/
Bug: T312528
Change-Id: Iead338a31403348603442b43ae802270f6b1d675
Implement a rest API endpoint that displays the popup that is currently
only shown on a special page.
Code was revived from Idd22057a88312bf1a1cb5546d0a6edca5678d80d
Bug: T288076
Bug: T233099
Change-Id: I65fcbf25ac5818f6c649daf494c719921247e8f5
Add support for the new properties 'P9758' (symbol represents) and
'P2534' (in defining formula) for the special page that shows
in depth information about tagged formulae.
Bug: T312893
Change-Id: I39551e04487f77cca60c89e8c95293032fd5bd6a
The MathEntitySelector gets its config from wbRepo which caused a hard
dependency on the Wikibase client.
If the Wikibase client config is not available a default value (wikidata)
is used to determine the URL of the foreign repo.
As the dropdown will never do write operations, we restrict the repo
access to anonymous read only.
Bug: T313143
Change-Id: Iba33dfd32a78f4ad7c2e99a1f56218458ab884b4
MathWikibaseConfig introduced additional complexity by passing through
some services to MathWikibaseConnector.
The remaining options were just a list of properties that could also be defined as ServiceOptions.
Bug: T313646
Change-Id: Ib9083c0cb62f6d972befc7f5dc3ed47f55cf92a5
* add some basic tests
* migrate MathWikibaseConnectorTest to a unit test
Bug: T313331
Depends-On: Ic3b8d4f685d5cf648a02696284b6ee499502a12a
Change-Id: I38d6425eb5e2c52ae4362c4b8656223a8d9d90a5
* add some basic tests
* MathWikibaseConfig should also be transformed into a service in a follow
up to allow for better testing
Bug: T313331
Change-Id: Iec7446a06d9a117cdeba9d26616f4fdf8e8540ca
Lookups returned by this factory will resolve redirects; the old factory
is deprecated and will soon be removed.
Bug: T312223
Change-Id: I3d03e2e7ad646c1048f914915b84836f958b7f6e
It is good practice to have a folder in the root for all sql files.
Also use the engine folder to store the patches
Change-Id: Ic26c042f0a5a8ca6e0568b38c4b4b2b4f0686efb
Originally this was added in T103269, but according
to T103269#6769990 this could not be needed anymore.
I've tested this with the original test case, and it
all works correctly without running the regex.
Change-Id: I5b3b11e71f0d6e55f158fa0d941977c823d12d51
Set up both database tables (mathoid and mathlatexm) regardless of the enabled modes. This will allow to change configuration without having to rerun db update scripts for the price of having potentially empty tables in the db.
Bug: T307896
Change-Id: I5ed50e53835c378f9f497a6ff3a4195299259d5b
Extensions using Phan need to be updated simultaneously with core due
to T308443.
Bug: T308718
Depends-On: Id08a220e1d6085e2b33f3f6c9d0e3935a4204659
Change-Id: I6d355e7037186e840a76f319886ed870419d9fc8
Adds a new configuration variable that can be used to run RESTBase
independently from Math.
Bug: T306572
Change-Id: Ie40bd36a53fa5018364cf9268fee29533f4aa23b
The global function wfReadOnly() has been deprecated in favor of the
new ReadOnlyMode service. Its usages should be replaced.
Bug: T283978
Change-Id: I8f7c026961a6b65e2e45a675c54f0481ccdf09e0
Old versions of Math returned
<dl>
<dd>iso-8859-1</dd>
</dl>
for empty math tags. This is no longer the case today.
Moreover, a problem in the png2svg conversion was fixed
in mathoid recently Iaf41f61335b4af51f758b64427c6e661ee36fb1b
This version of mathoid is now used in production.
Therefore, we can remove the handling of the special case.
Bug: T10372
Change-Id: Id835f7041a19f9519e7d5640b69a02b4b8a3397e
Instead of modifying a page rendering hash,
use the proper mechanism of parser option
registration. The cache is still split on parser
option value. The 'math' option is lazy, since
not all pages contain 'math' tag and we want to
avoid an user preference lookup for every parse.
WARNING: this will invalidate ParserCache for
all pages that contain <math> tags, but according
to Moritz, this is about 1% of all pages, so we
should be fine.
Setting up the test is still rather hard and weird
due to very hacky rendering mode to user option
converter, and math renderer interaction with the DB.
This will be cleaned up later.
Depends-On: I92b3e18fabef4eecac2ec2a4844f1be2716e5d89
Change-Id: I3bce04684070ad306685dabbc51267def25773cd
This is doing a bulk of the job of creating a service
for working with math modes configuration. There's
still things to do, like moving math mode names,
and perhaps depending on default user options to
provide a bit more convenience, but that will be the
next step.
Change-Id: I1259a93651920f44104f2f5135e3e620c858be8d
Additionally - remove RestbaseInterface::checkBackend.
When HTTP requests are banned from unit tests this method
makes no sence.
MathoidCli tests were not reenabled cause it's a bit hard
to mock shell acceess. Hopefully as part of Shellbox
transition we can have MockShell trait in core and reenable
those tests
Bug: T265628
Depends-On: I350ed59af603a0504704af3265787a4be8f2dc2a
Change-Id: I294d0ed905193850dc38cb3071eb9b31c8863121
The overall aim is to make renderer and checker
proper services and stop storing each operation
state in them. This patch reduces the number of
places where we store the lastError. I've decided
to stop here in the first pass so that the patch
doesn't grow out of proportion.
Change-Id: Ice7e1f9f2f074d62ef1819355a510ce0b0335d88
Instead of having a separate method for getting the post data
in MathLateXML, we can just override getPostData method from
parent and make the makeRequest method signature much simpler.
Change-Id: I07b683856885396ff7b6b5a40df79399828c8214
This is not the job of an extension to be a load balancer.
This feature was already deprecated and is not used in WMF prod.
Change-Id: I950dbaf45bc40d9d9ed61b1c98b339d3d38326fc
This hook is no longer needed by core so it's not
called anymore.
See I5124789fac333a664b73b4b4a1e801ecc0a618ca
Change-Id: I9a554b4e70a2564834e1df4058b1ad5cb98a56e3
Note that this changes the URL, which is directly shown to the user
(SpecialMathWikibase::buildPageRepresentation() uses it as the link
text). I think that’s okay (though it would probably be nicer to use
only the item ID as the text).
Bug: T286539
Change-Id: I50c49040846a5e08e5a354641fff35598034ceb9
User::setOption() is deprecated and should be replaced with UserOptionsManager::setOption()
Bug: T277818
Change-Id: I8bfcf9e732b6067d5461792d78c8b41ac31de3a9
* Reenable math validator tests
* Introduce $wgMathUseRestBase which defaults to true
and determines if mathoid is used via RESTbase or directly
* The effect of this change can be tested by setting $wgMathUseRestBase
to false, and then creating a new statement of a property with type math.
One should see a connection to mathoid for every new string and one should
also see the WAN cache for the check endpoint to fill up. Note that
setting WANObject cache to the database simplifies the investigation
of the WAN cache content.
Bug: T274436
Change-Id: I40109e9e09464da78bf855646fb800ee87c1ea67
This additional hook does not seem to be needed.
- wgMathValidModes: All access to this information already uses
getValidModes() and thus there is no need to write this back
to the input variable (use outside the extension is not supported
and it is already not used as cache either).
- wgMathDisableTexFilter: All access already uses getDisableTexFilter()
Move the last bit of logic there instead.
- wgDefaultUserOptions: Duplicate of this already exists in the
onGetPreferences hook higher up.
This seems almost too easy, so I suspect someone tried this before
and intentionally didn't finish it because of some non-obvious
issue.
See T189966 for related perf anaysis.
Change-Id: I4d7fdf6fcae3d44a37bf80d9c72fff2b4361f0c9
* Keep status code and response in memory
* Allow external cache preloading.
This will be required for the one-stop mathoid batch rendering
* Duplicate functionality from the restBASE interface
** Error rendering
** Validity checks
** getter for validated tex
Depends-On: Ic779e671ada0d0306e0935e578b47c431abd520e
Change-Id: I066975b9ba03f5861f6c92687cba4743044ea82a
This patch is a result of an observation that when <math> tags being
placed before {{#tree:..}} call they are not rendering correctly and
displaying -postMath strip markers instead, so never being replaced by
the actual math markup.
The reason is that the TreeAndMenu do parse operations within its parser
function callback and considering the MathHooks::onParserAfterTidy is
programmed to clear all the MathHooks::$tags data after first call of
the hook it looses all the data stored.
This issue also should appear of <math> is used with any other extension
function that calls parser `parse`, `recursiveTagParse`, etc
The patch modified MathHooks::onParserAfterTidy callback making it
remove tag data only after successful replacement instead of clearing
all the tags data on the first call
Change-Id: Icd115c61492d2ee9aa2004ad1940ff837bd1059a
The current wikidata entity selector requires WikibaseRepo to be enabled on the client wiki, which is only the case for very small installations.
We implement our custom entity selector that
* displays the item label and
* uses the description as tool-tip.
The implementation was inspired by
mw.widgets.TitleInputWidget.
Bug: T229939
Change-Id: I35d6cc317cf2984d9de6e062f79030f5c31c0811
For now, the hook handler isn't actually used (T248551) but
when it is at some point it shouldn't use $wgUser
Bug: T248550
Bug: T248551
Change-Id: I826e91411deb91bbd8c58147b32fda3f3e7ed6a8
Expose the math checking service via the REST API.
Port the behaviour of the RESTbase check endpoint to a new endpoint
that uses the MediaWiki API.
* this endpoint does not (yet) provide actual rendering (only checking)
* consequently it does not supply the x-resource-location header
Bug: T252390
Change-Id: I28bcb9a925953062b43cd3481536784431edcd8a
Use WANObjectCache rather than genuine DB tables for greater
flexibility.
Note, this commit is the second step of Ifd89e2d9cb30fa15cc318f077b434207ed6b7214
Bug: T252390
Change-Id: Ifa7d943875f05088ef697a5a18ad292342a9b4ad
The current default for Message::__toString() is to parse, so this
keeps that behaviour unchanged.
Change-Id: Icf83608b031390126b5e8d29a5600d37589cf411
This allows it to respect the proposed limit on HTTP request timeout.
Bug: T245170
Depends-On: I8252f6c854b98059f4916d5460ea71cf4b580149
Change-Id: I1c3d96720709253ad15bb8528cdd132571de2e4e
Introduce new namespace \MediaWiki\Extension\Math\ and move checking
code to this new namespace.
Note, this commit is the first step of Ifd89e2d9cb30fa15cc318f077b434207ed6b7214
Bug: T252390, T187154
Change-Id: I8c93f4b21d4a72ef2d52d4774d69e4ef55d037ec
* Currently PHP issues a unspecific warning if mathoid-cli does not
return a PNG image.
* This change suppresses this warning.
Bug: T200493
Change-Id: Ice7d12ca9131d76fe233fd332e37484feaf79883
Discussions in the Wikimedia Math Community show that there is consensus
to deprecate the use of MediaWiki specific LaTeX macros that conflict
with LaTeX macros from commonly used packages.
In this change, we show the deprecation warnings generated by mathoid.
Bug: T197842
Change-Id: I24dbb446665fdc227d0e7342fdbf8829b4c1bda4
Actual changes:
* Let the special page accept the subpage parameter, e.g.
[[Special:MathWikibase/Q6203]].
* Fix some of the message escaping issues in SpecialMathWikibase.
Messages that are just plain text and not supposed to contain any
parameters, wikitext, HTML, or something like this should be
properly escaped, either via Html::rawElement( $msg->escaped() ) or
better Html::element( $msg->plain ). The later approach is cheaper.
Other code cleanups that don't affect production:
* Reduce the amount of duplicate code in SpecialMathWikibase.
* Make use of the preg_match() return value.
* Remove default parameters from parent::__construct() call.
* Remove some auto-generated comments.
* Remove some comments that don't add new information, but just repeat
what the code already says.
* Make use of PHPUnit's assertCount().
We can split this patch into smaller patches, if you prefer. This
should especially be done if one of the changes is controversial, to
not block the other changes. Please tell me.
Change-Id: I8c20014435af3d2ac9c7381fe5fd00b9fe49b2c2
$wgContLanguageCode has been an unconfigurable copy of $wgLanguageCode
since 2006 (MW 1.8).
Bug: T247674
Change-Id: I1d2ec1e63320086fdf9ab7c6321f949c99c1ab2a
The following sniffs are failing and were disabled:
* MediaWiki.Commenting.FunctionComment.MissingDocumentationPrivate
Change-Id: I57bb0f5bc4167ea4b4a50d666ad95c02fcc7d8c9
With this commit, the Math extension only depends on WikibaseClient
and no longer on WikibaseRepo.
Bug: T208758
Change-Id: Iac8f2568e0da1a4aff4f4b22d124626c8ecca8eb
Add a special page and an API endpoint to fetch data from Wikibase
items with a given qID. The special page summarizes information from
Wikibase. The API endpoint allows to request the information
directly. Both, the API endpoint and the special page, fetch the data
from a new helper class for consistency.
Bug: T208758
Bug: T229939
Change-Id: Idd22057a88312bf1a1cb5546d0a6edca5678d80d
When $wgMathoidCli is misconfigured the error message was misleading.
It was reported that the specified file was not executable, even though
other reasons are possible (e.g., nodejs misconfigured).
Therefore, we capture STDERR and add it to the error log.
Bug: T198564
Change-Id: I7d0ae0f3ab4789124f9ff1533e7712ca31233f9e
Gets warning information from restbase and adds tracking categories
for deprecated chemical syntax.
Refactor mechanism to add tracking categories, and avoiding rechecking
already checked formulae.
Depends on: Ieca66f45ae7685d61eece1624bd7ff65ccad2eaa
Change-Id: I10e78ab79015dc1331f645c60b25bbbd237e23fe
The replacement OutputPage::addWikiTextAsInterface() method was added in
MW 1.32; the minimum required MW version has been bumped accordingly.
Bug: T198214
Change-Id: Ie9932b7398d0a9184094a99aec060d78f62dd6f0
Create MathDataUpdater class that checks if statements
are using math and adds the stlyes required for math in that case.
This patch has no effect until
I5bc0622ee7338f3215d14e15331a0a1931ca1ae0
is merged.
After both changes are in effect
I0abd6acbfa12fd503d11476f0c8c8c8a8a851cdc
can be merged.
Bug: T173949
Change-Id: I0e24bbb53e6e01d549f534744780ca1afc49fdd7
Note that a class is not resolved this was. …::class is not a function
call. It's more like a named string constant. Technically still a
string. The advantage is that IDEs and tools like linters can much easier
understand that these strings refer to a class, and list them in usage
reports, renames, and such.
Change-Id: I5225543dbb837685a1840837cb2772dd576cca38
* deprecate and fix pickHost. It was broken and never used.
* simplify inputTypeSelection
* remove superfluous is_array check for the result of explode
Change-Id: I392f22f074facfe30b97d53a3002f464a471b67e
In T74240 support for LaTeX based rendering of the <math/>-tags was
removed. This change eliminates the unused code from the repository.
Bug: T195871
Change-Id: Ic57d83fa49f090b574ce8b82fd2ebc84a5350318
* Use the exactly same routines to deliver png images that are used in
mathml mode.
* Change the output to use mathoids png image rather than the mathml
and svg output.
* Locally tested on Firefox and Chrome: Depending on the mode either
the SVG or the PNG path is used.
Bug: T74240
Change-Id: I4b1cac92eb9a02190f316faab6621940951603d5
The motivation for this patch is to make the code less complex, better
readable, and less brittle.
Example:
public function onExampleHook( Parser &$parser, array &$result ) {
/* This is the hook handler */
}
In this example the $result array is meant to be manipulated by the
hook handler. Changes should become visible to the caller. Since PHP
passes arrays by value, the & is needed to make this possible.
But the & is misplaced in pretty much all cases where the parameter is
an object. The only reason we still see these & in many hook handlers
is historical: PHP 4 passed objects by value, which potentially caused
expensive cloning. This was prevented with the &.
Since PHP 5 objects are passed by reference. However, this did not
made the & entirely meaningless. Keeping the & means callees are
allowed to replace passed objects with new ones. The & makes it look
like a function might intentionally replace a passed object, which is
unintended and actually scary in cases like the Parser. Luckily all
Hooks::run I have seen so far ignore unintended out-values. So even if
a hook handler tries to do something bad like replacing the Parser
with a different one, this would not have an effect.
Removing the & does not remove the possibility to manipulate the
object. Changes done to public properties are still visible to the
caller.
Unfortunately these & cannot be removed from the callers as long as
there is a single callee expecting a reference. This patch reduces the
number of such problematic callees.
Change-Id: I21d53c989ea487607dc69e6b3365c023ef6729f5
Without this change it would no longer load in safe mode (T185303)
because the module would be missing.
I think this has been copy-pasted from VisualEditor, see commit
I6d097ccbf1dc2462843219adcf96bf8313e30289 there for explanation.
Bug: T185303
Change-Id: I6644a361d1450c394a7f82dc3dad97515982846d
Without this change it would no longer load in safe mode (T185303)
because the module would be missing.
I think this has been copy-pasted from VisualEditor, see commit
I6d097ccbf1dc2462843219adcf96bf8313e30289 there for explanation.
Bug: T185303
Change-Id: Ieed46488c94c5809e9f17f667c6926a21ffe838b
