search

go-text / typesetting

High quality text shaping in pure Go.
Other
88 stars 11 forks source link

Can we use x/text? #11

Closed whereswaldon closed 1 year ago

whereswaldon commented 2 years ago

I just stumbled across golang.org/x/text/unicode/bidi. I don't think I've heard us talk about this library before, so I wanted to ask if that's because we know it isn't adequate to our needs, or because we didn't realize it was available. Does anybody know anything about the x/text implementation?

andydotxyz commented 2 years ago

Fyne uses x/text, and mostly it is just not compatible with the APIs in @benoitkugler repositories so I have to manage the two font definitions side-by-side (we still use x/text for rasterising). If they were able to be compatible that would be a huge relief, but I suspect we can't. Thankfully the bidi package does not seem to depend on their definition of a Font/Face so we may be able to make use of it internally.

benoitkugler commented 2 years ago

If they were able to be compatible that would be a huge relief, but I suspect we can't.

The main reason we can't for now is because x/text Opentype implementation does not support all the layout tables (GPOS/GSUB and AAT, which make the complexity and value of Harfbuzz, by the way). We could maybe try to propose PRs into x/text to fill the gap..

It may be easier to do the other way around and add rasterising into go-text (relying on x/text or implementing from scratch).

The ideal situation, which seems reachable, is that a single font definition (that is a font.Face element) is enough to provide both metrics for layout and sufficient glyphs drawing data/rasterization.

Thankfully the bidi package does not seem to depend on their definition of a Font/Face so we may be able to make use of it internally.

Indeed (and for a good reason which is that, contrary to shaping or font fallback, the bidi algorithm is actually totally font agnostic). x/text/unicode/bidi seems actually in good shape. Would you like me to look closer at how we can use it ? In particular, what I have in mind is starting by checking if the test suite from fribidi is passing with x/text.

whereswaldon commented 2 years ago

@npillmayer Do you know anything about the status of x/text/unicode/bidi?

@benoitkugler If you have time to check it against the fribidi test sutie, I think that would be very interesting data!

whereswaldon commented 1 year ago

I think this isn't necessarily a go-text problem. Gio does use this implementation of bidirectional segmentation and it's doing just fine for us. I'm going to close this for now.