[HarfBuzz] inputting text already containing some glpyhs

[Behdad Esfahbod]

On 04/21/11 21:42, Khaled Hosny wrote:
> I was thinking about adding HB support to FontForge's "metrics" view,
> which input can as well be a mix of characters and glyph indices, but I
> did no actual work so far.
> 
> I was wondering if hb_shape() accepts a buffer that contains glyphs
> indices? if it is true, may be one can first convert characters to
> corresponding "base" glyph indices using some FT function (or something
> like Cairo's "toy" text API) and then pass the "pure" glyph indices
> buffer to HB, any ideas?

I thought about this use-case more.  While it's impossible to do shaping on
glyph indices, I think you can get away with what you (and Eric) want by
implementing a custom get_glyph() call back.  Simply encode the glyph indices
in your "character" space by, for example, adding 0x80000000 to the glyph
value.  Then in your get_glyph() function, which would normally map characters
to glyphs, handle the encoded glyphs differently and just return the glyph index.

In the design Ryan and I came up with you can in fact subclass hb_font_funcs_t
objects to do things like this.  I probably get to implement that in a day or two.

behdad


> Regards,
>  Khaled
> 
> On Thu, Apr 21, 2011 at 04:57:35PM -0400, Behdad Esfahbod wrote:
>> Hi Eric,
>>
>> I understand your request, but I don't think we want to handle that.  The way
>> such a case would work, IMO, is that the higher level will break the text into
>> text and glyph-index segments, and only pass the text runs to the shaper.
>> Shaping glyph indices makes little sense.
>>
>> behdad
>>
>> On 04/21/11 16:51, Eric Wasylishen wrote:
>>> Hi,
>>> I have a use case where I want to shape some text which might contain a mix of glyph indices and unicode characters. For example, a string like "hello <glyph 123> world". 
>>>
>>> I get the feeling that just having strings like this is discouraged. There are, however, files in the wild that do this (e.g. Apple RTF files will sometime contain glyph indices).
>>>
>>> What would you think about allowing a Harfbuzz user preparing their buffer to call hb_buffer_add_glyph with a mask to indicate that the glyph being inserted is already a glyph index?  It looks like the first few phases of hb_ot_shape_execute_internal, up to when hb_substitute_default is called, would have to check for this mask and ignore the glyph or treat it as some placeholder unicode character.
>>>
>>> Cheers,
>>> Eric
>>>
>>> P.S. I'm looking at adding Harfbuzz support to GNUstep, which is where this problem comes up.
>> _______________________________________________
>> HarfBuzz mailing list
>> HarfBuzz at lists.freedesktop.org
>> http://lists.freedesktop.org/mailman/listinfo/harfbuzz
>