Unicode math properties #2425
Unicode math properties #2425
Conversation
…ning, ...) associated with a unicode char
…ning, ...) associated with a unicode char
…d with the glyph, after decoding and font adjustment
…il digestsion; use new decodeMathChar API
…lass with any already assigned grammatical role
There was a problem hiding this comment.
I actually quite like the direction of the PR. Maybe I don't yet understand enough to register the controversial piece?
I left a few minor comments, and I can't say I understand the full details of the TeX interplay yet, but the grammar-near pieces seem encouraging.
I especially like the changes in the tests, where the math XML markup has become significantly more enhanced.
| ($local ? Tokens(T_CS('\char'), $value->revert, T_CS('\relax')) : $$self{cs})); } | ||
| else { # Else math mode, mathDecode! | ||
| my ($glyph, $f, $rev, %props) = LaTeXML::Package::decodeMathChar($nvalue); | ||
| if (!defined $props{name}) { # TEMPORARY HACK ????? |
There was a problem hiding this comment.
Is there anything to add to that comment? E.g. what is missing for the hack to go away?
| $whatsit->setProperty(font => LookupValue('font')->specialize($glyph)) if $glyph; | ||
| my ($glyph, $f, $rev, %props) = decodeMathChar($n); | ||
| $whatsit->setProperty(glyph => $glyph) if $glyph; | ||
| $whatsit->setProperties(%props); |
There was a problem hiding this comment.
Very minor, but you could also guard that setter call:
$whatsit->setProperties(%props) if %props;
| DefConstructor('\mathclose Digested', "?#1(<ltx:XMWrap role='CLOSE'>#1</ltx:XMWrap>)()", bounded => 1); | ||
| DefConstructor('\mathpunct Digested', "?#1(<ltx:XMWrap role='PUNCT'>#1</ltx:XMWrap>)()", bounded => 1); | ||
| DefConstructor('\mathinner Digested', "?#1(<ltx:XMWrap role='ATOM'>#1</ltx:XMWrap>)()", bounded => 1); | ||
| our %mathclass_subclass = ( |
There was a problem hiding this comment.
This nested hash has the feeling of a mini-grammar... Interesting.
How about three other ideas:
PUNCT => { PERIOD => 1 }ID => { NUMBER => 1 }BIGOP => { SUMOP => 1, INTOP => 1, LIMITOP => 1, DIFFOP => 1 }
There was a problem hiding this comment.
Cool!, yep, that's the idea!
| my $font = $curfont->merge(%$fontinfo); | ||
| $fontinfo = $defn && $defn->isFontDef; | ||
| if ($$basefontinfo{size} != $curfont->getSize) { # If we've gotten an explicit font SIZE change; Adjust! | ||
| $fontinfo = {%$fontinfo}; $$fontinfo{size} = $curfont->getSize; } } |
There was a problem hiding this comment.
There is a small theoretical chance $fontinfo is undef here, so maybe default it to {} ?
$fontinfo = ($defn && $defn->isFontDef) || {};| $fontinfo = {%$fontinfo}; $$fontinfo{size} = $curfont->getSize; } } | ||
| my $font = $curfont->merge(%$fontinfo); | ||
| if ($downsize > 0) { $font = $curfont->merge(scripted => 1); } | ||
| if ($downsize > 1) { $font = $curfont->merge(scripted => 1); } |
There was a problem hiding this comment.
These merge calls have the same value? Do you intend it to do the same merge twice when >1 ?
There was a problem hiding this comment.
yep, twice for scriptscript
| #====================================================================== | ||
| UTF(0x5C) => { role => 'ADDOP', meaning => 'set-minus' }, # \backslash | ||
| UTF(0xAC) => { role => 'BIGOP', meaning => 'not' }, # \lnot | ||
| UTF(0xAC) => { role => 'BIGOP', meaning => 'not' }, # \neg |
There was a problem hiding this comment.
0xAC is given 3 separate times here? Unsure how that happened. There may be other cases, I see 0x5C is also there twice.
There was a problem hiding this comment.
editing mistake. fixed.
| "ker" => { role => 'OPFUNCTION', meaning => 'kernel' }, # \ker # | ||
| "lg" => { role => 'OPFUNCTION' }, # \lg # | ||
| "lim" => { role => 'LIMITOP', meaning => 'limit', need_scriptpos => 1 }, # \lim # | ||
| "lim inf" => { role => 'LIMITOP', meaning => 'limit-infimum', need_scriptpos => 1 }, # \liminf # |
There was a problem hiding this comment.
Interesting, first I assume single characters, then single words (no spaces), but I see some entries even have a space.
There was a problem hiding this comment.
yeah, these don't actually get fired, yet; it would need to be hooked into the math ligature that combines "s,i,n" into "sin"; and then looked up! I suspect "lim inf" will likely fail that test. I kinda left them all in as a reminder...
* New unicode_math_properties(char) gets the math properties (role, meaning, ...) associated with a unicode char * New unicode_math_properties(char) gets the math properties (role, meaning, ...) associated with a unicode char * Revise API for decodeMathChar to return the math properties associated with the glyph, after decoding and font adjustment * Revise CharDef->new API to take only the code, deferring decoding until digestsion; use new decodeMathChar API * Updates for improved decodeMathChar and CharDef->new API * Update fontmap for some bigops * Have \mathop and friends intelligently merge the requested TeX math class with any already assigned grammatical role * Update tests for more 'semantic' when low-level glyphs are accessed * Refinements & clarifications suggested by D.Ginev * Add guard
Here's an interesting, perhaps controversial, PR: It enhances
decodeMathCharto also lookup math properties (role, meaning,...) for the decoded glyph from the Unicode module. This provides some parity between symbols defined "semantically" by bindings and low-level\mathchar, in the simple cases. This covers most of what TeX's plain defines, as it defines them!