gmane.comp.freedesktop.harfbuzz

Re: Contextual shaping of Malayalam post(pre)/below base forms

Suresh P — 2013-06-19T08:39:17

I have found that the <YA VIRAMA LA> case can be handled with 
decomposition rule using ccmp for the given context. But that looks as a 
work around to me. The <YA VIRAMA RA> is more complex because it 
involves ligation and reordering.

_______________________________________________
HarfBuzz mailing list
HarfBuzz< at >lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/harfbuzz

Re: Contextual shaping of Malayalam post(pre)/below base forms

Richard Wordingham — 2013-06-18T22:46:23

On Tue, 18 Jun 2013 19:33:05 +0530
Suresh P <sureshp< at >gmx.com> wrote:







I think the new logic is missing near, in 0.9.18, line 996 of
hb-ot-shape-complex-index.cc, where the code reads:

  if (indic_plan->mask_array[PREF] && base + 2 < end)
  {
    /* Find a Halant,Ra sequence and mark it for pre-base reordering
  processing. */ for (unsigned int i = base + 1; i + 1 < end; i++) {
      hb_codepoint_t glyphs[2] = {info[i].codepoint, info[i +
  1].codepoint}; if (indic_plan->pref.would_substitute (glyphs,
  ARRAY_LENGTH (glyphs), true, face)) {
info[i++].mask |= indic_plan->mask_array[PREF];
info[i++].mask |= indic_plan->mask_array[PREF];
...

Using the Meera font (Meera_04.ttf, Revision 4.0, date 12 April 2008),
with substitutions reduced to those for pre-base RA, the code enters
this block when processing the string from <YA, VIRAMA, RA>.
Unfortunately, I stopped tracing the logic in sufficient detail after
this point.  I think the reordering is done before the pref lookup is
actually carried out, and

Re: Contextual shaping of Malayalam post(pre)/below base forms

Suresh P — 2013-06-18T14:03:05

It seems neither blwf ligature lookup for <VIRAMA LA> nor pref lookup 
for <VIRAMA RA> honour the contextual lookups. :(  The calt feature I 
used for YA VIRAMA LA with fontforge was:

//
lookup caltContextualAlternatesinMalay {
   lookupflag 0;
     ignore sub [\b_ya \m_U \b_la]'
} caltContextualAlternatesinMalay;

feature calt {
   script mlm2;
      language dflt ;
       lookup caltContextualAlternatesinMalay;

} calt;
//

Also tried [\b_ya] [\m_U]' [\b_la].
yes
The font contains script tags for both versions and I think harfbuzz 
defaults to mlm2.

Re: question about indic matra group rule

datao zhang — 2013-06-18T03:20:23

Re: Contextual shaping of Malayalam post(pre)/below base forms

Richard Wordingham — 2013-06-18T01:38:56

On Mon, 17 Jun 2013 08:47:10 +0530
Suresh P <sureshp-KK0ffGbhmjU< at >public.gmane.org> wrote:






I'll offer my understanding of the situation.  If my explanation
doesn't help you, it may be necessary to examine the actual <GSUB>
table of your font to explain what is happening, and other tables will
provide context for it.

So what you want in exactly one lookup for the <blwf> feature:

In context <ya virama la> do nothing
else in context <virama la> ligate virama la.

(The second, normal condition will invoke a ligature lookup.)

That should do what you want; are you having trouble expressing that
in the lookup definition language you are using?

The same trick should work with the <pref> feature:

In context <ya virama ra> do nothing
else in context <virama ra> ligate virama ra.

Again, the second, normal condition will invoke a ligature lookup.
This ligature lookup should *not* be listed under the <pref> feature.

The OpenType specification at
http://www.microsoft.com/typography/OpenTypeDev/malayalam/intro.

Re: question about indic matra group rule

Richard Wordingham — 2013-06-17T07:52:47

On Mon, 17 Jun 2013 12:06:43 +0800
datao zhang <dataozhang-PkbjNfxxIARBDgjK7y7TUQ< at >public.gmane.org> wrote:
 

I take it that by 'Indic script' you mean 'Indian Indic script'.  There
are many words with 3 matras in Tibetan, and 4 different matras can
occur with simple monosyllables in Tai Tham.  These, however, happen not
to use the generic Indic shaper in HarfBuzz.

Informal Malayalam, however, does double some two-part vowels.


Decomposition is usually easier than composition, and the two-part
vowels need to be decomposed at some stage - consonant vowel ligatures
generally work with parts rather than with whole vowels.

Richard.

question about indic matra group rule

datao zhang — 2013-06-17T04:06:43

Dear behdad:
 
I feel confuse for the syllable rule in  hb-ot-shape-complex-indic-machine.rl:
halant_or_matra_group = (final_halant_group | (h.ZWJ)? matra_group{0,4});
 
I can't image there any case for indic script which 4 matra followed. I can understand you decompose some matra before syllable analysis, ex: Sinhala (0xDDC, after decompose, two matra generated). So you specify more than one matra. But I don't know why it is 4, not 2.
 
Another issue is caused that you decompose matra before syllable analysis and allow more than one matra followed. ex: if we have two matras: (0xDD9,0xDD9). They will be treated as one syllable. I don't know what benefit we can get from such rule. At least it obeys those MS Specific for Indic script.
 
Any comments is appreciated?
 
Thanks!
 
Br,
Dean

Re: Contextual shaping of Malayalam post(pre)/below baseforms

Suresh P — 2013-06-17T03:17:10

Okay, as per the spec.
There is no malfunctioning as such. But is looks weird when one see a 
pre-base RA in front of a base YA ( and a below base LA subjoining YA ) 
because it is never written like this. Instead, in actual use, ligation 
(and reordering) is avoided and written with explicit virama. A test 
case is attached herewith and it applies to any font which uses pre-base 
RA and below base LA. It is better left to the font developer to decide 
the contexts.

Re: lack of documentation drives me crazy

Anthony Carrico — 2013-06-17T00:58:14

Welcome. That was me. If everyone who is concerned about the
documentation contributes, then we'll have it.

As you go up the learning curve make notes here:
  https://github.com/behdad/harfbuzz/wiki

The devs requested gtk-doc for api documentation. I set it up in a git
branch here:
  https://github.com/behdad/harfbuzz/tree/gtk-doc

I'm not sure if that was merged, but of course it will need patches with
actual documentation to be useful.

Welcome.

Re: Skipping Control for Attaching Marks using OpenType

Richard Wordingham — 2013-06-17T00:29:21

On Mon, 17 Jun 2013 07:43:52 +1000
Andrew Cunningham <lang.support-Re5JQEeQqe8AvxtiuMwx3w< at >public.gmane.org> wrote:


My font is currently being set up to work with a Tai Tham shaper.
Other OpenType environments will be added later if I can work out how.
Graphite will be added at some stage.  I am currently using blwf,
pstf, abvs and blws features, which I didn't expect to see working on
Windows 7 for Tai Tham.  I have set up detectors in the font to perform
specific contractions if any of the these or the ccmp, liga, clig, or,
after providing a copy to Behdad, rlig, features is enabled. The blwf
and pstf detectors are broken.

The only application on Windows 7 Service Pack 1 that triggered was
Firefox 19.0.2 (I don't use Windows much), which triggered the
detectors for clig, rlig, abvs and blws.  It gave the same results for
the test sequence I have just provided as HarfBuzz 0.9.18; however, I
believe it is using HarfBuzz despite being on Windows.

In particular, Internet Explorer 9, Notepad, Word 2002, SIL

Re: lack of documentation drives me crazy

Behdad Esfahbod — 2013-06-17T00:20:17

Werner.  You are absolutely right.  I'm sorry!  I'll fix that as soon as I'm
back from Haiti.  Promise!

behdad

On 13-06-15 03:14 PM, Werner LEMBERG wrote:

Re: lack of documentation drives me crazy

Vincent Isambart — 2013-06-16T22:06:30

Someone started some work at https://github.com/behdad/harfbuzz/wiki

Re: Skipping Control for Attaching Marks using OpenType

Andrew Cunningham — 2013-06-16T21:43:52

Hi Richard,

I don't know what OT featurez you are using in your font, but Tai Tham
would be an undefined script in Unis ribe so it would need to limit itself
to features fhat can be used by default shaper.

Essentially, rlig, mark, mkmk, optionally liga and clig.

If your font is using anyother feature fod Tai Tham rendering yoh will not
get what you expect.

Andrew
On 17/06/2013 3:35 AM, "Richard Wordingham" <richard.wordingham-XZoyATsUNX7QFizaE/u3fw< at >public.gmane.orgm>
wrote:

Re: Skipping Control for Attaching Marks using OpenType

Richard Wordingham — 2013-06-16T17:34:54

On Thu, 13 Jun 2013 17:17:49 -0400
Behdad Esfahbod <behdad-muU0i0fqO+Mdnm+yROfE0A< at >public.gmane.org> wrote:



I'm sending an untrimmed test font off-list as I assume most people
don't want it.

I've tested the patch and found that it removes the deviation from the
specification of the lookup.  FWIW, the patch causes no problems known
to me. Unfortunately, my font doesn't seem to trigger any feature use in
Microsoft products, and Firefox 19.0.2 on Windows seems to use
HarfBuzz, so my attempts to compare with Uniscribe have failed.

As a test, I have tried three methods of moving the glyph for TAI THAM
SIGN MAI KANG LAI from immediately following one base consonant to
immediately following the next.  In each method, there are three
contexts to consider - mai kang lai and the next consonant separated by
nothing, by one preposed character (a vowel or MEDIAL RA), or by two
preposed characters (a vowel and MEDIAL RA).  The scheme works by:

(a) Deleting the original mai kang lai by a ligature lookup merging it
wit

Re: lack of documentation drives me crazy

Richard Wordingham — 2013-06-16T15:45:26

On Sun, 16 Jun 2013 15:05:54 +0200 (CEST)
Werner LEMBERG <wl-mXXj517/zsQ< at >public.gmane.org> wrote:

 

As with a lot of things, the quick route to that e-mail is (English)
Wikipedia - that thread is one of the external links in the article on
HarfBuzz.

Somebody did propose starting a HarfBuzz wiki, but I can't remember
where it was supposed to be located.

Richard.

Re: lack of documentation drives me crazy

Werner LEMBERG — 2013-06-16T13:05:54


Thanks!  I think my main problem is that I've never used Cairo.
Otherwise, a lot of concepts would be immediately clear to me, I
believe.


Actually, as the sample code from my previous e-mail shows, I really
need only hb_ft_face_create :-)


Yeah, I've realized this meanwhile.  Fortunately, I only need to work
with FreeType, so I don't need such hassles.


Thanks!  I've tried to find this e-mail, but somehow I failed.


    Werner

Re: lack of documentation drives me crazy

Shriramana Sharma — 2013-06-16T10:08:40

Sorry to not answer your question -- but perhaps they should have a
Google Summer of Documentation for projects like this that require one
desperately but the developer hasn't had the time to do it so far.
Behdad could probably mentor someone else writing the documentation.

(Just a wild thought...)

Re: lack of documentation drives me crazy

Cosmin Apreutesei — 2013-06-16T08:18:30

Hey, I totally share your pain.

Lemme try to answer some of your questions from my very limited
knowledge until Behdad responds.


A hb_ft_face is internal to hb_ft_font and created automatically. It
is needed by the font object to query font metrics and read sfnt
tables. I don't know why you would create one yourself.


hb_font_create and hb_face_create create "abstract" objects. If you
create them like this, you need to implement and assign a bunch of
callbacks to them to make them work. This way harfbuzz is extendable
so you can use your OS font subsystem instead of freetype for reading
fonts, or hook in to your svg parser so you can use svg fonts, etc.
This is similar to prototype-based inheritance of dynamic laguages.


Check this out:
http://lists.freedesktop.org/archives/harfbuzz/2009-August/000359.html

first tests of hb_ot_layout_lookup_collect_glyphs

Werner LEMBERG — 2013-06-16T08:03:11

[harfbuzz 79d1007a501fd63c0ba4d51038c513e6b8b94740]


Behdad,


the attached program gives the also attached output if applied to
`pala.ttf', version 1.00 (which is ooold).  Interestingly,
`hb_ot_layout_collect_lookups' returns incorrect values: According to
ttx (output also attached), the lookups affected by `smcp' are only 8
and 9, while my demo program reports 7 also...  And indeed, the first
four glyphs in the list are incorrect.  Note that the function seems
to work fine with other fonts like Libertine, or even `pala.ttf'
version 5.00.

I'm sending you `pala.ttf' v1 privately.


    Werner
#include <stdio.h>

#include <ft2build.h>
#include FT_FREETYPE_H

#include <hb.h>
#include <hb-ot.h>
#include <hb-ft.h>


int
main(int argc,
     char** argv)
{
  int ret = 0;

  FT_Error ft_error;
  FT_Library ft_library;
  FT_Face ft_face;

  hb_face_t* face;
  hb_set_t* lookups;
  hb_codepoint_t lookup_index;

  hb_set_t* glyphs;
  hb_codepoint_t glyph_index;

  hb_tag_t features[] =
  {
    HB_TAG ('s', 'm', 'c',

lack of documentation drives me crazy

Werner LEMBERG — 2013-06-15T19:14:34

Behdad,


please, PLEASE add some documentation to harfbuzz!  It really drives
me crazy that I have to read the source code of harfbuzz to get some
clues.  This is an endless time sink :-(

While the source code is written very cleanly, the purpose of it is
not as obvious as you might think.  For example, what's the difference
between

  hb_ft_font_create
  hb_ft_face_create

and what's the difference of those two to the

  hb_font_create
  hb_face_create

functions?  When should I use the former, when the latter?  Or both?
In which order?

Or: What is a blob?  Why should I use it?  Etc., etc.

IMHO, providing at least a skeleton documentation (i.e., short
sentences which describe the various functions and structures with a
few words) is *essential*, and I call it a really critical bug that
absolutely nothing is available.  And no, the stuff in the `util'
directory is not a valid substitute for lack of documentation, since
those examples aren't documented either (and written in high-level,
non-trivial C++ c

Fwd: Javanese script in Harfbuzz

R.S. Wihananto — 2013-06-14T21:42:22

Here is the font. It contains simple OpenType substitution using
'blwf' and 'pstf' features. virama + consonant -> conjunct. Consonant
cluster in Javanese script formed like other Brahmic script:
C1 + virama + C2 -> C1 + conjunctC2.
The virama is ꧀ U+A9C0. For example: ꦩ꧀ꦧ mba, ꦩ꧀ꦥ mpa.

wihananto

On Fri, Jun 14, 2013 at 3:19 AM, Behdad Esfahbod <behdad-muU0i0fqO+Mdnm+yROfE0A< at >public.gmane.org> wrote: