Add hb_buffer_create_similar()

[behdad]

Right now one can create new buffer and append existing buffer to it, but in utils I also have this piece:

  static void copy_buffer_properties (hb_buffer_t *dst, hb_buffer_t *src) 
  { 
    hb_segment_properties_t props; 
    hb_buffer_get_segment_properties (src, &props); 
    hb_buffer_set_segment_properties (dst, &props); 
    hb_buffer_set_flags (dst, hb_buffer_get_flags (src)); 
    hb_buffer_set_cluster_level (dst, hb_buffer_get_cluster_level (src)); 
  } 

[ebraminio]

I needed this (and #1480) for some Kashida related testing thing, so.

[behdad]

Yes. Let's sort out the API first though. Should it create a new buffer? Or do we want a, essentially, assign version? I don't have a strong opinion.

[behdad]

The usecase in utils needs to copy just a range of the buffer. That's done using hb_buffer_append which takes a start/end. So, maybe we just need a function that copies the non-content properties like above. Just need to pick a name for it.

[ebraminio]

I like something like Java's .clone() wording, hb_buffer_clone to return a new and modifiable buffer. My usecase was something like to copy a non-shaped buffer, shape it, maybe copy again and add Kashida glyph to places I like, and render it using hb-render, which I know, is superfluous usecase anyway and can be done completely without these and may is wrong layer to apply it, I however had no rush experimenting with the thing so.

[behdad]

Clone wouldn't address the need in util.

[behdad]

The usecase in utils needs to copy just a range of the buffer. That's done using hb_buffer_append which takes a start/end. So, maybe we just need a function that copies the non-content properties like above. Just need to pick a name for it.

What is also not address by hb_buffer_append is the effect on pre/post-context.

[behdad]

What is also not address by hb_buffer_append is the effect on pre/post-context.

And that's #1843

[behdad]

Getting ideas from cairo: hb_buffer_create_similar(). I like that.

[behdad]

hb_buffer_create_similar()

One problem with it: since, unfortunately, hb_buffer_clear_contents() also clears segment-properties, there's a need for two different two different functions. Maybe one doesn't need to be a function:

• hb_buffer_create_similar() should copy everything other than contents, content-type, and segment-properties.

• contents and content-type are handled by hb_buffer_append. Maybe segment-properties, if not set (individually, script, language, direction), should also be set by hb_buffer_append? Then the code in util will become automatic.

[behdad]

@khaledhosny wdyt? We already changed hb_buffer_append to take care of pre/post-context. I think taking care of segment-properties makes sense, since we consider segment-properties as part of content.

[khaledhosny]

@khaledhosny wdyt? We already changed hb_buffer_append to take care of pre/post-context. I think taking care of segment-properties makes sense, since we consider segment-properties as part of content.

Makes sense.