* Pulled in MathJax 1.1 (minus the 112 megabytes of PNG "fonts")
* Pulled in some of <https://en.wikipedia.org/wiki/User:Nageh/mathJax> bits for initialization
This doesn't currently override PNG images, but does replace the text-form if you configure it.
PHP Parse error: syntax error, unexpected T_STRING, expecting ')' in /home/betawiki/projects/mediawiki-ext/Math/Math.i18n.php on line 2235
Full export will readd missing messages.
If enabled, by default loads MathJax from a CDN and forces all equations through MathJax when JavaScript is available -- the image or source form gets used only as <noscript> fallback.
This has a couple of problems for us:
* if scripts are missing -- such as when viewed via MobileFrontend -- you end up with no math at all, as the <script type="math/tex"> bits get ignored but the <noscript> still hides the images.
* while MathJax is loading things, blank spots on the page may appear and stuff jumps around
* if loading new text in via ajax, it doesn't trigger there
Also using some fairly default config, no idea how appropriate it is. Hoping to get some insight from Nageh who's done some of the MathJax work on Wikipedia so far - <https://en.wikipedia.org/wiki/User_talk:Nageh#MathJax_integration_into_stock_MediaWiki>
MathML mode was so incomplete most people thought it simply didn't work (bug 25646).
HTML modes often rendered poorly (eg req bug 24207 to set default to PNG on some sites)
This may cause regressions in that simple "equations" of just variables that came out as HTML
will now render as PNGs that don't get aligned properly with the text baseline. Fixing this is
covered by bug 32694: <https://bugzilla.wikimedia.org/show_bug.cgi?id=32694> to retrieve the
baseline info from dvipng and position images to fit.
Note that because of the way user options are pulled in to the oarser cache key, some folks
may see cached pages with their old settings until they get redone or they save their prefs
again and have it normalized.
The new TEX_FUN1nb introduced by r99741 was not added to the
html_render_deep() pattern matching list. This was spotted by
an ocaml compilation warning:
File "html.ml", line 74, characters 31-2966:
Warning 8: this pattern-matching is not exhaustive.
Here is an example of a value that is not matched:
TEX_FUN1nb (_, _)::_
but \operatorname to fix Bug 31442. More specifically, TEX_FUN1 now
adds braces as it did in previous versions. The result is texvc will
once again correctly sanitize multiple math accents even when no braces are
given (e.g. \dot \vec B) and it will allow math accents to be used with
symbols whose font as been changed (e.g. \tilde \mathcal{M}). Parser
tests are created to ensure these continue to work.
The handling of \operatorname is now handled specially by the
function TEX_FUN1nb (nb stands for no braces and the name is taken to
be consistent with the existing function TEX_FUN2nb). The addition of
braces causes this command to LaTeX incorrectly. The parser test for
\operatorname has been updated to reflect the fact the hash of the
images involved has changed.
* (bug 18912) Add math support for \sen Spanish variant of \sin
* (bug 18912) Fix spacing for \operatorname in math
Reapplies r86962, r87117, r87936, r87941 plus some parser tests.
Note that further batch testing to identify any other potential problems due to the spacing tweaks is a good idea!
Reapplied r87092 (tests added, tweaked for merge, no functional diff) to allow "'" and "`" in \text{...} bits.
I'm not convinced this is a sufficient fix -- there are probably a *lot* more where this came from.
But it doesn't hurt to add them!
May require texlive-lang-greek package to be installed; but it's only loaded when these bits are used.
Reapplied r87298 plus parser test case to confirm the new chars can render and the existing \digamma hasn't changed.
Reapplied r87284 with a tweak to allow \euro through to conservative HTML output.
Parser test case added to confirm that \euro is HTMLable and the other euro variants are renderable to images.
This commit also includes tweaks to HTMLification of \phi, \varphi, and \varepsilon (conservative HTML) and changes the non-conservative HTMLification of number sets from plain letters to the dedicated Unicode chars for their symbols. These changes are not included in tests, but don't alter what input is supported.
Reapplied r86965 (patch by Thenub314), plus parser test case to ensure that \widetilde{x} actually renders. (While we can't compare the contents of the image, the output hash will remain stable in the img url. Forcing $wgMathPath to a fixed fake path for parsertests.
While these mostly look like they seem to do what they claim, *none* of them provide any test cases.
Most don't even include an example in the commit messages.
