[WIP/RFC] Handle multiple spaces after PUA glyphs by blueyed · Pull Request #1036 · kovidgoyal/kitty

blueyed · 2018-09-28T14:04:56Z

This handles multiple spaces after a PUA glyph (instead of just one).

Before:

After:

It still does not work as in (patched) urxvt, because kitty appears to shrink it to fit its height into the cell. In urxvt it looks like follows:

Where does the shrinking / fitting to cells happen? I would rather have it take the full width of the 3 cells, and be cut at the top/bottom.

blueyed · 2018-09-28T14:07:16Z

kitty/fonts.c

+                render_run(fg, line->cpu_cells + i, line->gpu_cells + i, j + 1, cell_font_idx, true);
+                run_font_idx = NO_FONT;
+                first_cell_in_run = i + 1 + j;
+                prev_width = gpu_cell->attrs & WIDTH_MASK;


Re prev_width: I think this should be 1 always (regardless of this patch) to not continue above in any case.

blueyed · 2018-09-28T14:33:23Z

Where does the shrinking / fitting to cells happen? I would rather have it take the full width of the 3 cells, and be cut at the top/bottom.

This appears to happen in place_bitmap_in_canvas (for fontconfig).

kovidgoyal · 2018-09-28T14:40:53Z

Fitting into the cell is not going to change -- it is needed for proper rendering of glyphs from fallback fonts, which can have different metrics than the main font.

kovidgoyal · 2018-09-28T14:49:01Z

As for handling multiple spaces, IIRC kitty supports a max ligature length of 9 so you should ensure that the fake ligature you create is no longer than that, or bad things will happen down the road (see MAX_NUM_EXTRA_GLYPHS)

blueyed · 2018-09-28T16:00:51Z

MAX_NUM_EXTRA_GLYPHS is check here: https://github.com/kovidgoyal/kitty/pull/1036/files#diff-bf6698a6663f1ce908538b90d1fbe1e7R1014

blueyed · 2018-09-28T16:18:46Z

Fitting into the cell is not going to change -- it is needed for proper rendering of glyphs from fallback fonts, which can have different metrics than the main font.

I've looked into a bit more and the bitmap created from fontconfig is only 25 pixels wide already (render_bitmap: rows=23 width=25 max_width=30, cell_height=20). This is likely due to internally setting the desired font height I assume. (Not sure why rows > cell_height?)
In this case this is already a fallback font btw (via symbol_map).
What I would like to achieve is that the bitmap would take the max_width, and gets cropped at top/bottom - but I can see that this is nothing you would like to have in general.

kovidgoyal · 2018-09-28T16:22:40Z

No, I prefer scaling to cropping. But I am fine with having an option to control that behavior. However, it would only work with freetype since AFAIK there is no way to do that with coretext.

blueyed · 2018-09-28T16:53:08Z

No, I prefer scaling to cropping. But I am fine with having an option to control that behavior.

Great. Do you have some pointers for me where this needs to happen?
Is it related to set_font_size being based on (preferred) height?

kovidgoyal · 2018-09-28T16:56:59Z

Basically you'd need to change the code that fits the glyph(s) into the canvas to crop instead of scaling based on the value of the option

blueyed · 2018-09-28T17:16:37Z

That would be place_bitmap_in_canvas I assume??
render_bitmap before loads the glyph with width=25 already though (for max_width=30 (3 cells)). Therefore I am assuming that font size setting is causing this.
(There appear to be several places where width should be used instead of height then maybe)

kovidgoyal · 2018-09-28T17:29:33Z

sorry dont remember off the top of my head would have to read the code, and I dont have the time for that right now

blueyed · 2018-09-28T17:59:52Z

Ok, will look into this a bit more.

As for this PR: it should be independent from what we discussed above, but I've just noticed that it is broken (added space before):

This appears to be happen because it gets centered in the group of 8 cells (via https://github.com/blueyed/kitty/blob/efc0b830cf5a4430adfa48be9a874086f974a427/kitty/freetype.c#L510-L516).

[ci skip]

kovidgoyal · 2018-09-29T02:40:47Z

yes glyphs are currently centered in the canvas. this works well for most glyphs such as emoji and flags and east asian characters. If you want left alignment for pua glyphs, you will need to change the freetype rendering api to pss in a bool controlling alignment. rememberthat these functions are also used in coretext so make the changes there as well.

Finally, the more cells there are in a ligature, the worse lookup performance is. so I would prefer restricting the max number of spaces to 2 or 3 ince I doubt nerd fonts need more than that.

Not sure why this became necessary though?! Fixes: Linking kitty/fast_data_types ... kitty/freetype.c: In function ‘render_glyphs_in_cells’: kitty/freetype.c:514:82: warning: ‘bm.right_edge’ may be used uninitialized in this function [-Wmaybe-uninitialized] if (num_cells > 1 && right_edge < canvas_width && (delta = (canvas_width - right_edge) / 2) && delta > 1) { ^ kitty/freetype.c:490:21: note: ‘bm.right_edge’ was declared here ProcessedBitmap bm; ^

[ci skip]

blueyed · 2018-09-29T16:42:51Z

I've added some commit to use the actual number of cells (currently only for freetype, based on the bitmap). Does this make sense?

blueyed · 2018-09-29T16:46:40Z

The metrics for U+e0c2 are this btw:

glyph: 1581 bitmap.width: 25 bitmap.rows: 23
horiAdvance: 1600 horiBearingX: 0 horiBearingY: 1152
vertBearingX: -832 vertBearingY: -192 vertAdvance: 1088
width: 1600 height: 1472

(cell height is 20, cell width 10)

blueyed · 2018-09-29T17:08:42Z

If get_glyph_width makes sense like this, it could be used with is_glyph_empty, of course.

kovidgoyal · 2018-09-30T03:37:32Z

I'll review this when I have a little more time.

kovidgoyal · 2018-10-28T06:09:38Z

I'll review this sometime in the next few days. Apologies for the long delay, my personal life has been crazy and I am trying to catch up now.

blueyed · 2018-10-30T11:40:28Z

Wow, thanks for merging - I would have expected some review/comments on it first, and would have squashed at least the fixup then, but ok.. :)
Thanks!

kovidgoyal · 2018-10-30T15:02:22Z

I made some minor changes, but otherwise it looked good, and I am not fussy about commit history :)

I find merging things quickly that require only minor cleanups is better for my long term productivity.

blueyed · 2018-10-30T15:22:52Z

Cool, thanks again!

Handle multiple spaces after PUA glyphs

efc0b83

blueyed commented Sep 28, 2018

View reviewed changes

fixup! Handle multiple spaces after PUA glyphs

d52434c

[ci skip]

blueyed added 5 commits September 29, 2018 16:19

render_glyphs_in_cells: skip centering glyph for PUA

1d50a18

fixup! render_glyphs_in_cells: skip centering glyph for PUA

8e09e80

MAX_NUM_EXTRA_GLYPHS_PUA=4

c64e8fc

render_line: use cells based on bitmap width

12cfe6d

[ci skip]

get_glyph_width for core text?!

3c4cb3c

kovidgoyal merged commit 3c4cb3c into kovidgoyal:master Oct 30, 2018

blueyed deleted the multi-pua branch October 30, 2018 11:39

kovidgoyal mentioned this pull request Dec 7, 2018

Strange behavior with iconic font on 0.13.1 #1210

Closed

Uh oh!

Conversation

blueyed commented Sep 28, 2018 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

blueyed Sep 28, 2018

Choose a reason for hiding this comment

Uh oh!

blueyed commented Sep 28, 2018

Uh oh!

kovidgoyal commented Sep 28, 2018

Uh oh!

kovidgoyal commented Sep 28, 2018

Uh oh!

blueyed commented Sep 28, 2018

Uh oh!

blueyed commented Sep 28, 2018

Uh oh!

kovidgoyal commented Sep 28, 2018

Uh oh!

blueyed commented Sep 28, 2018

Uh oh!

kovidgoyal commented Sep 28, 2018

Uh oh!

blueyed commented Sep 28, 2018

Uh oh!

kovidgoyal commented Sep 28, 2018

Uh oh!

blueyed commented Sep 28, 2018 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

kovidgoyal commented Sep 29, 2018

Uh oh!

blueyed commented Sep 29, 2018

Uh oh!

blueyed commented Sep 29, 2018 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

blueyed commented Sep 29, 2018

Uh oh!

kovidgoyal commented Sep 30, 2018

Uh oh!

kovidgoyal commented Oct 28, 2018

Uh oh!

blueyed commented Oct 30, 2018

Uh oh!

kovidgoyal commented Oct 30, 2018

Uh oh!

blueyed commented Oct 30, 2018

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

blueyed commented Sep 28, 2018 •

edited

Loading

blueyed commented Sep 28, 2018 •

edited

Loading

blueyed commented Sep 29, 2018 •

edited

Loading