CVS log for loncom/xml/lontexconvert.pm

Up to [LON-CAPA] / loncom / xml

Request diff between arbitrary revisions

Keyword substitution: kv
Default branch: MAIN

- For 2.11
  Backport 1.122.

- Default renderer for unconfigured domains set in a single location.

- For 2.11
  Backport 1.120 (part).

- For 2.11
  - Backport 1.121.

- Bug 6881.

- Bug 6867
  - Domain configuration to set default math renderer for <m> tag content.
  - Default renderer for unconfigured domains changed from tth to MathJax.

- For 2.11
  - Backport 1.119

- For 2.11
  - Backport 1.118

- Eliminate warning: "Unescaped left brace in regex is deprecated" when
  starting loncontrol on perl 5.22

- Eliminate jsMath.
  Use MathJax instead in cases where jsMath use is indicated in a resource,
  ina user's preferences, or in course preferences.

removed extra br before table in tth output with display mode (to avoid a large blank space when there is a div or p before)

- For 2.11
  - Backport 1.114, 1.115, 1.116

- Remove trailing white space.

- Remove obsolete comments.

- Bug 5873. Conversion of entities generated by tth to unicode bracket
segment characters.
  - loncommon::decode_user_agent() sets additional item in %env
    (browser.osversion) with information about Windows version.
  - Replacement of entities with unicode equvalents now the default,
    in lontexconvert::unicode_subst()
  - Exceptions: pre-Windows NT 6.0 (i.e., pre-Vista) and mobile devices
    (i.e., ipad, iphone etc.) -- entities replaced with ascii characters
    in those cases.

- For 2.11
  - Backport 1.113

- 'tth' is sixth arg in lontexconvert::algebra().
  Fix change in rev. 1.103.

- For 2.11.
  - Reinstate 1.109 (reverse changes in 1.112.2.1). MathJax now in 2.11.

- For 2.11.
  - Reverse changes in 1.109. MathJax not in 2.11.0.

Remove some debug logging accidently left in.

BZ 5873 - Ugly ascii substitutions for large grouping characters that just
don't work uniformly in moder browsers.  The unicode substitutions are present
but commented out as well in the conversion table.

BZ 5837 -
  1. Add table of conversion string -> unicode equivalents and
     sub to do a conversion.
  2. Use this to turn the entities generated by tth to unicode bracket
segment characters directly.

*** empty log message ***

- Bug 5925.
- Simplify internal HEAD request for /cgi-bin/mimetex.cgi by using loopback.

- Duplicate declaration.

- Bug 6320: Same alignment/new paragraph for eqnarrays as it is used for, e.g. $$ ... $$ (Thanks to Juliane Wenzel for research)
- Replaced deprectated <center> tag with div/CSS

- Backport 1.105.

- Bugs: 5925, 6424.
  - Eliminate separate call for SSL servers to IO::Socket::SSL to retrieve
    vertical alignment for mimetex images. Use IO::Socket::INET for both
    SSL and non-SSL web servers. Rewrite rules may need to be changed on
    SSL web servers to serve HEAD requests for /cgi-bin/mimetex.cgi from port 80.

Removed leftovers of already removed alternative login mode (accessibility).

Making &init_tth() optional when calling lontexconvert::algebra, by adding
an extra argument, and then updating the two places where this function
is called.

I think this change had a purpose at some point. Maybe it still does.
Maybe it doesn't anymore. Maybe I just want to be 3 lines closer to
developer of the month? I have no idea.

Bug 6189
    Too smiley

-- Requires all emoticons to have noses, where applicable.

Bugs 2295 and 6155, related to 6155.  Disallowing smileys that could conflict
with URLs and mathematical expressions.

adding new smiley faces in png format, and replacing
existing ones with new path

Raw mode for formula display

Mimetex alt-argument for screen readers:
* do not remove original math enviroment delimiters
* change double quotes into double primes (so we do not have alt="..."...")
* remove artificial math delimiters

- Omit "User-Agent" part of string so Apache 1.3 (RHEL4 and SLES9) is able to process the HTTP header.

- Add linefeed to HTTP header.

- Use URI::Escape to escape the TeX string passed in the query string to mimetex.cgi.

- Replace use of localhost.localdomain with hostname from &Apache::lonnet::hostname() for lonHostID.

- Replace use of LWP with lower level IO::Socket::SSL and IO::Socket::INET for HEAD request for Vertical-Align from headers. (mimetex.cgi first pass).
   - Reason: mimetex_converted() can be called inside safespace (e.g., by &xmlparse('<m>stuff</m>' in a script block)
     - In this context call to open() - line 144 of HTTP/Request/Common.pm (called by $ua->head()) is trapped in safespace resulting in a run time error.

- &Apache::lonxml::alarm() now used for timeouts.
  - Timeout args for IO::Socket::SSL/INETcan not be used because of dependency on fileno command at line 61 of IO/Select.pm (masked in safespace with run time error).

added call to mimetex_valign which destills the vertical alignment shift needed to vertically align mimetex-formula-images.
Note that two calls are needed to mimetex.cgi.
The first one (within mimetex_align) has the only purpose to strip the needed value from the HTTP header field. The second call gets the actual image.
According to
http://forkosh.com/mimetex.html
this currently seems to be the only way to do it.
Note, however, that the double call is not that detrimental w.r.t. to performance as mimetex on loncapa is configured such that images (and vertical alignment values) are cached.

I have edited the implementation of mimetex to calculate the vertical-align-value for the images, so that the mimetex-gifs are aligned to the surrounding text correctly. I have commented the call, because there is still a hard dependence to a http-connection to the local server. This is a problem (and the timeouts are to high in this solution). We try to get this functionality now working without calling http by doing it local with the mimetex.cgi on the server. I have nevertheless commited this code, so that you can review it. Search for FIXME.

POD comments added/altered

Added ALT-Attributes containing LaTeX-Sourcecode in mimetex-rendering, so that blind people can access this information.

Bug #5770: Added display attribute for algebra tag.

- Problem testing in CSTR - if blank was selected for "math rendering" option, don't return, otherwise tex engine set in user preference will be ignored.

BZ5593: Post process by replacing \leftrightharpoon with the corresponding
UTF-8 character.. works for recent browsers..though known to have
problems with older MAC-OS 10.3 and Firefox 2,9 works for Safari on older
Mac-OS...MAC-OSX 10.4.11 with Firefox2.0.0.11 does work.

As usual forgot to remove the debug logging

BZ5593 - postprocess converted latex so that
       \rightleftharpoons -> ⇔ which is a
       double left/right arrow and is going to have to be
       close enough.

- BUG#5370, need the popup previous tries window to use jamath if the the problem needed jsmath, so created a means to convey that

- add ability for a course to specify a tex engine to be used in the course,
  currently overrides user's setting if they have one.

- was incorreclyt trying to clean out all $ that weren't \$, now correct

- some cleanups to the generated html to make it easier to read

- a <m> placed outside of </problem> could cause an ISE, log the error and go on

- death to $_

- add support for eqnarray* environments to jsMath

- AlgParser.pm 1.15 no longer produces extra () in the bottom of a frac (BUG# 4695)

- BUG#5005, need to shift to maintianing the jsmath processing status to a stack since we can create subpages when creating the main page thus need to track the need for each sub page independently

&Apache::lonnet::unescape -> &unescape
&Apache::lonnet::escape -> &escape

- tth doesn't lie the trailing space on the option list

- calling init_tth from start_page (effectively)
- reducing dependencies from lontextconvert.pm

-making this cmmon usage a simple sub call

- adde type arg

- lontex -> start_page
- eliminate old lontexconvert header/footer routines
- lontex use more standarish form parms to turn things on/off

- Apache 2 compatability

- be case insensitive about the display value

- adding support for resource author to specify a rendering engine
   <m display="tth"> for instance

- grabing update sto AlgParer.pm from WebWorK (revision 1.4)

- ENV -> env

- <algebra> works in FDBK

- stop allowing infinte loops to cause havoc, nesting alram() doesn't work so we need to fake the ability to do it.

- adding tex->image convertor (mimeTeX)

- comment and remove extra logthis

- removing some debugs, and getting it to stop complaining

- adding jsMath as a TeX -> HTML option
- adding the ability for a user to choose between jsMath and tth
- setting tth as the default, but I think the default should be based off browser detection, on anything at all modern jsMath is much sexier

- adding the option to put it into display style

- forgot to update this regexp after the conversion back to <> <= and >=

- lim(a,x=1,3) lim of a as x=1 goes to 4

- form gor int, int(3,a,b) integral from a to b of 3
- simplify int( a dx) -> a

- cos/sin etc (3,4) cos^4(3)

- log(3,4) -> log base 4 of 3

- moodle suggests that extra \left( can end up in the dnominator of a \frac, trusting it

- support sqrt(3,4) -> 4th root of 3

- remove some commented out code
- move the geq leq not= transformations back into AlgParse.pm

- get rid of the zeta -> zita, and do all zeta -> \zeta translations in the AlgParser

- cleanup minus signs

- take care of some nic operators

- add in some greek an sepcial symbol fixups

- meant to use math mode not displaystyle mode

- adding in <algebra>

It does not seem to make any difference in the display of equations, but I
meant to only replace the very first <br clear="all" /> if there's nothing
but whitespace in front of it. Arrgh, missing "^"

Delete only the very first 'clear="all"' - appears to work very well.
1.3.3?

The 'clear="all"' in the leading <br /> of displaystyle equations adds horrible whitespace, so much so that students overlook equations because they think the
page ended.

Displaystyle equations already have

<br /><table>..<table>...</table>..</table><br /> ... that's plenty.

Removing the "clear".

- xhtmlizing cleanup

Reverting to 1.37. regexp on discussion post messages moved back to lonfeedback.pm

Allow posters to edit/delete their own discussion postings.  Controlled by
course parameter set via PARM. Previous versions of postings are preserved.
Deleted postings are hidden. Course Coordinators etc. will be able to see
all versions of postings (interface still needed). Editing currently works for
resource discussions, but not bulletin boards (work in progress).

- if there are any errors at all when doing tth kill the apache child
  - BUG#3030

- fix problem with trying to call &xmlparse('<m>stuff</m>');

- BUG#2809, reports of tth going off into an infinite loop, should catch that and kill off the child cleanly.

&mt cannot translate '['

-localizing the tex conversion

- trying to handle BUG#2522, more of a shot in the dark than anything else.

- retabination

- Fixes BUG#2499, emails with large <m> where getting shoved onto one line because
   - we use <pre>
   - we strip out extra returns before processing

- I tired these changes out with a variety of messages both with and without <m> sections, and post discussion, looks good to me. let me know if you see anything else that looks ugly

- consolidated tth init sections
- course parameter for tth for setting tth parameters

- tth generates
</font



>

this seems to annoy some sub version of IE.

Thus we remove the whitespace in the end font tag
(BUG#2189)

Bug #2295: remove denoser compatibility to make sure smileys mostly only
show when intended to be there.

Internationalization and German localization, spelling mistake.

Bugs #1960 and #2156: <m> in CHAT and FDBK, just like everywhere else.

- better error message
- if the calling function wants an array of results pass the errors back seperately

- how about this instead then.

Bug 1316: tth has problems with linebreaks in formulas. Replaced linebreaks
by space.

Attachments displayed in normal messages.

- backport of 1.18 (HTML entities decode)

Technology breakthrough: fully de-noser compatible smiley generator

- last of BUG#1276, I think this may bomb if you dont have the latest tth.so

- tth doesn't appreciate > and <

- lonhelp now supports proper unicode output
- errors in tex messages don't keep getting repeated

A lot of things happened here:

* everything menu-related was moved out of lonxml, and into lonmenu
* thus, function calls to registerurl, loadevents and unloadevents modified
* fixed lontex.pm, which actually did not call the right routines
* put in calls to lonmenu::menubuttons, which is supposed to insert buttons
  after the body tag in "textual mode" - this function and registerurl will
  be unified (buttons either in Remote or on top of page)

- Fixed the FIXME, -L works properly, and need to do proper unicode for browsers

best wishes to all.

- have the tth init code set the unicode encoding
- new ENV parm needed browser.unicode

Catches frequent tth-segfaults due to bad LaTeX.

What used to be a \n should at least be replaced by a space.

Bug #872

- This will get redone when I get a chance to look at the new tth/m stuff

- Supports both methods of entering edisplya mode

- also returns the error message if there was one

Attempt to get it into LaTeX mode for messages (apparently not working), and
minimal matching in reg exp.

Moved tex convertion of messages into lontexconvert. Having problems with
getting tth into LaTeX mode.

Update to use new &lonxml::registerurl().

GPL

Mac font settings

only need the pathname here

these files are moved from the CVS:modules directory