[WIP] BiDi and shaping support for Font::draw/Font::get_string_size #21791

bruvzg · 2018-09-06T07:25:50Z

Changes Font::draw, Font::draw_halign, Font::get_string_size functions and Label, LineEdit controls to support bi-directional text and complex scripts using ICU and HarfBuzz libraries.
Adds Font::draw_paragraph function, and ShapedString/ShapedAttributedString classes for text display and input handling.

🔹 Third-party libraries

HarfBuzz (full library except non-relevant font backends (CoreText, Uniscribe) and built-in UCDN), 1.8.8, Old-MIT
ICU4C ("common" library part and built-in unicode base data), 62.1, Unicode License

🔸 Updates

7 Sep 2018 - Fixed build for ARMv7 Linux, added glyph offsets to get_string_size calculation.
27 Sep 2018 - Added multiline and rich text shaping APIs.
12 Oct 2018 - Added Label, and LineEdit, text shaping moved from font to separate ShapedString (for plain text), ShapedAttributedString (for rich text)

🔹 TODO list

Related:
#10546, #3081, #9961, #982

bruvzg · 2018-09-06T09:01:39Z

Some visual comparison of 3.1 and this PR:

🔹 3.1 alpha 1

🔹 93a888e + this PR

🔸 References to compare rendering

Noto Nastaliq Urdu test page:
http://behdad.org/urdu/

Random Wikipedia articles:
https://he.wikipedia.org/wiki/פאדי_קונסידיין
https://km.wikipedia.org/wiki/ខេត្ដព្រះសីហនុ
https://th.wikipedia.org/wiki/อาณาจักรโชซ็อนโบราณ

Update: Some additional examples of opentype features (Fonts: Kleymissky, Noto Sans, Fira Code)

🔹 3.1 alpha 1

🔹 this PR

akien-mga · 2018-09-13T09:47:08Z

COPYRIGHT.txt

@@ -173,6 +173,26 @@ Comment: The FreeType Project
 Copyright: 1996-2017, David Turner, Robert Wilhelm, and Werner Lemberg.
 License: FTL

+Files: ./thirdparty/harfbuzz/


Sections are ordered alphabetically based on the "Files" path, so this should go after ./thirdparty/glad/.

akien-mga · 2018-09-13T09:49:41Z

COPYRIGHT.txt

+ 2005, David Turner
+ 2004, 2007, 2008, 2009, 2010, Red Hat, Inc.
+ 1998-2004, David Turner and Werner Lemberg
+License: MIT


The text of Harfbuzz's MIT variant should be included, as it's not the same as the "classical" MIT license Godot uses, or as Debian names it the Expat license.

Since there are dozens of MIT licenses with different wordings, I'd suggest to name this one MIT-HarfBuzz for clarity.

For the reference, I checked https://www.debian.org/doc/packaging-manuals/copyright-format/1.0/#license-specification and https://spdx.org/licenses/ but it doesn't include any identifier for HarfBuzz's MIT variant.
Debian itself documents it as "MIT" in https://metadata.ftp-master.debian.org/changelogs/main/h/harfbuzz/harfbuzz_1.8.8-2_copyright, but since we might end up adding more thirdparty code under other MIT variants, I prefer to be more explicit.

akien-mga · 2018-09-13T09:56:22Z

COPYRIGHT.txt

@@ -388,6 +408,41 @@ License: BSD-3-clause



+License: Unicode


Licenses are listed alphabetically too, so this should go just above License: Zlib.

akien-mga · 2018-09-13T09:56:45Z

SConstruct

@@ -190,6 +190,9 @@ opts.Add(BoolVariable('builtin_squish', "Use the built-in squish library", True)
 opts.Add(BoolVariable('builtin_thekla_atlas', "Use the built-in thekla_altas library", True))
 opts.Add(BoolVariable('builtin_zlib', "Use the built-in zlib library", True))
 opts.Add(BoolVariable('builtin_zstd', "Use the built-in Zstd library", True))
+opts.Add(BoolVariable('builtin_hb', "Use the built-in HarfBuzz library", True))


I'd prefer builtin_harfbuzz.

Should also be listed alphabetically.

akien-mga · 2018-09-13T09:59:00Z

SConstruct

@@ -190,6 +190,9 @@ opts.Add(BoolVariable('builtin_squish', "Use the built-in squish library", True)
 opts.Add(BoolVariable('builtin_thekla_atlas', "Use the built-in thekla_altas library", True))
 opts.Add(BoolVariable('builtin_zlib', "Use the built-in zlib library", True))
 opts.Add(BoolVariable('builtin_zstd', "Use the built-in Zstd library", True))
+opts.Add(BoolVariable('builtin_hb', "Use the built-in HarfBuzz library", True))
+opts.Add(BoolVariable('builtin_icu', "Use the built-in ICU library", True))
+opts.Add(BoolVariable('use_sil_graphite2', "Use the external SIL Graphite library (LGPL licensed)", False))


I don't see any LGPL text in src/hb-graphite2.{cc,h}, it has the same copyright header as HarfBuzz.

akien-mga · 2018-09-13T10:01:49Z

platform/javascript/detect.py

-    # Don't use dynamic_cast, necessary with no-rtti.
-    env.Append(CPPDEFINES=['NO_SAFE_CAST'])
+    # These flags help keep the file size down
+    env.Append(CPPFLAGS=["-fno-exceptions"]) #, '-fno-rtti' - dynamic_cast required by ICU


I don't think we can/should do that. CC @eska014

Yeah, turning off runtime type information saves about 200KiB last I tested, so we should avoid re-enabling it. Does ICU have some build option that avoids usage of RTTI?

akien-mga · 2018-09-13T10:03:41Z

thirdparty/README.md

@@ -124,6 +124,27 @@ Files extracted from upstream source:
 - the include/ folder
 - `docs/{FTL.TXT,LICENSE.TXT}`

+## ICU4C


Should be listed alphabetically too (and with two lines between sections).

reduz · 2018-09-16T23:01:13Z

Hi Bruvzg, very glad that you are trying to get this to work, I think this is a lot of progress. How do you want to continue from here? I can imagine Label and RichTextlabel will need more complex code.

Also, one thing I was never too clear about is how big is ICU. I suppose in the range of <1mb it could be compiled in with the option to compile out. If larger, it could be built-in for the editor, but loaded via gdnative on exported projects (or optionally compiled in on an export template).

bruvzg · 2018-09-27T10:28:40Z

How do you want to continue from here? I can imagine Label and RichTextlabel will need more complex code.

I have added separate functions to break lines and justify for Label, and separate shape_spans for RichText. But didn't do anything to controls itself yet.

Also, one thing I was never too clear about is how big is ICU. I suppose in the range of <1mb it could be compiled in with the option to compile out. If larger, it could be built-in for the editor, but loaded via gdnative on exported projects (or optionally compiled in on an export template).

Currently total size impact of HarfBuzz and ICU on macOS release export template without LTO is 1.42 MB, some parts of ICU probably could be removed.

ICU break iterator data is about 6 MB for all languages, full data is 26MB, but it is optional (only used for breaking lines and justification of some languages) and can be loaded as resource when needed (Can be set in application/config/icudata project setting). Also full ICU distribution includes tool to customize data file.

With ICU data:

Without ICU data:

I don't see any LGPL text in src/hb-graphite2.{cc,h}, it has the same copyright header as HarfBuzz.

I didn't included actual library (libgraphite2) code itself, since it's LGPL in most cases lib should be linked dynamically if someone need it.

Not sure how to handle this best, graphite won't be useful for most users, it's only used for rare languages and require special smart fonts.

bruvzg · 2018-12-29T22:54:52Z

FYI: I have uploaded standalone version of the same thing at - https://github.com/bruvzg/godot_tl

Works both as builtin module (clone into Godot's modules and rebuild engine) and gdnative module (dynamic library).
Same backend with some fixes + paragraph + rich text edit control prototype.
Includes optional graphite shaper support (LGPL/MPL).
In addition to String works with raw UTF-8/UTF-16/UTF-32 PoolVector (Godot String on Windows uses 16-bit wchar_t but does not support surrogate pairs on any level and is effectively broken for everything outside BMP).
Uses slightly modified font classes (based on Godot font classes, stripped of unneeded stuff), to provide smarter script based font fallback/substitution.

…(LineEdit, Label), line breaking and line justification using ICU and HarfBuzz libraries.

slevy85 · 2019-04-11T17:23:25Z

Hi,
I am not sure it is the right place for this, but I have used this PR and I got an issue when using Stretch Mode in 2D :
Project Settings/Display/Window/Stretch

The letters get mixed together :

Edit : It doesn't happened with the default Godot font, so maybe it is a font issue, I used Amiri fonts to test

realkotob · 2019-09-02T10:45:43Z

@bruvzg If I need to add a Label with arabic text, should I use this PR or use the standalone version?

MohammadKhashashneh · 2019-10-01T15:38:38Z

Hi, anyone can update us with the current state of this pull request? it's been a while since the last exchange. @bruvzg Are you still on it? how can I help?

MohammadKhashashneh · 2019-10-02T12:01:30Z

Ok, I did a little research and to answer my own question here is what I found:

The branch builds and it looks really promising. glad to see actual Arabic support happening. great work @bruvzg
Aside from the above mentioned todo list, the following issues are currently present:
- Could not build with the use_staitc_icu_data=true option
- No default fallback/last resort when not specifying a font (had to specify a font to show Arabic script)
- No fallback also when using mixed text (2 scripts)
- crashes here and there (when running the scene, when modifying font properties...)
There exist a separate module https://github.com/bruvzg/godot_tl (didn't try yet)
that is actually a bit more active and fixes some issues, more controls supported, fallback...

Now my new questions @reduz @akien-mga @bruvzg

How to proceed from here? Is fixing and continuing on this branch and ultimately merging still an option?
should we just move on and use the separate module? can't it be officially included in godot ?

I'm willing to help either way so I appreciate it if someone could answer the above.

akien-mga · 2019-10-04T11:02:15Z

How to proceed from here? Is fixing and continuing on this branch and ultimately merging still an option?

should we just move on and use the separate module? can't it be officially included in godot ?

The plan is to have all these features built-in in the engine, I think it fulfills an important enough use case to be made available without hassle in the base engine.

But to do so, the implementation and API need to be designed to fit closely to Godot's design principles and architecture for rendering text - or Godot's architecture for text needs to evolve based on a consensus. At the time of this PR the consensus could not be reached, but later on @reduz and @bruvzg more or less reached a consensus on how to do it. Now what's needed is contributor time, but also the ability to refactor text APIs and probably break compatibility, which means waiting for Godot 4.0's development cycle.

MohammadKhashashneh · 2019-10-05T06:09:19Z

Great! Thank you @akien-mga for the clarification.
If that's the case then 4.0 would be Ideal to properly shake things up and do it right.
Ok , I believe in the meantime is to get aquanted more with this PR tiny details and get ready.

Will someone post the requirements/vision somewhere?

OmarAglan · 2019-10-30T10:37:38Z

@bruvzg Any Update On This?
I worked on it a bit and it's working with some addons from my side (i will not pull request it).

bruvzg · 2019-10-30T11:55:26Z

Any Update On This?

Nothing is being done here. If you need a bit more maintained version - https://github.com/bruvzg/godot_tl

aaronfranke · 2020-05-13T06:08:24Z

@bruvzg What is the status of this PR? Is there any desire to continue this work and add it to Godot, or should users who want this feature simply be directed to your repository?

If not, abandoned pull requests will be closed in the future as announced here. Feel free to close this PR if you don't want to rebase and continue this work.

jitspoe · 2020-10-14T02:57:34Z

Is this the closest thing to getting Arabic support or is there something else being worked on? I saw this as a gdscript solution, but it seems like it would make more sense to support it natively in-engine: https://github.com/3akev/godot-arabic-text

Also, the gdscript version uses regular expressions, which I believe cause problems with Switch ports. (Is that still a thing?)

bruvzg · 2020-10-14T04:57:24Z

@jitspoe current version is here #41100 (for 4.0 / master), there's also https://github.com/bruvzg/godot_tl (for 3.2).

nightblade9 · 2020-10-14T13:42:07Z

I saw this as a gdscript solution, but it seems like it would make more sense to support it natively in-engine: https://github.com/3akev/godot-arabic-text

AFAIK that project was intended as a band-aid solution (it's mostly a port of some Python bidi libraries/code) until we can get something better. I believe there is an issue in the godot-feature-requests repo for proper bidi support.

bruvzg requested review from akien-mga and leonkrause as code owners September 6, 2018 07:25

bruvzg mentioned this pull request Sep 6, 2018

[WIP] Add support for bidi text output #10546

Closed

bruvzg force-pushed the min-shaping branch from 47498ee to 4cf4f2c Compare September 6, 2018 08:22

akien-mga added feature proposal topic:core labels Sep 6, 2018

akien-mga added this to the 3.1 milestone Sep 6, 2018

akien-mga requested a review from reduz September 6, 2018 09:34

bruvzg force-pushed the min-shaping branch 3 times, most recently from 2349a44 to 7c79abe Compare September 7, 2018 11:42

akien-mga reviewed Sep 13, 2018

View reviewed changes

bruvzg force-pushed the min-shaping branch 2 times, most recently from 7b573fc to 20f16f1 Compare September 27, 2018 10:28

bruvzg force-pushed the min-shaping branch from 20f16f1 to c4150e4 Compare September 27, 2018 16:17

bruvzg changed the title ~~BiDi and shaping support for Font::draw/Font::get_string_size~~ [WIP] BiDi and shaping support for Font::draw/Font::get_string_size Sep 27, 2018

bruvzg force-pushed the min-shaping branch from c4150e4 to c3c125c Compare October 12, 2018 10:07

bruvzg mentioned this pull request Oct 17, 2018

Most emojis don't work on Windows #23093

Closed

bruvzg force-pushed the min-shaping branch 3 times, most recently from a68b98c to 57d9eda Compare October 29, 2018 08:57

kalmufti mentioned this pull request Jan 25, 2019

Arabic Translation GDQuest/godot-power-pitch#86

Open

bruvzg force-pushed the min-shaping branch from 10c55c2 to a0d53f2 Compare February 5, 2019 06:54

bruvzg requested a review from a team as a code owner February 5, 2019 06:54

bruvzg force-pushed the min-shaping branch 2 times, most recently from 819be30 to efed218 Compare March 12, 2019 21:12

BiDi and shaping for DynamicFont and BitmapFont and complex controls …

bf338b2

…(LineEdit, Label), line breaking and line justification using ICU and HarfBuzz libraries.

bruvzg force-pushed the min-shaping branch from efed218 to bf338b2 Compare March 13, 2019 07:30

bruvzg mentioned this pull request Apr 6, 2019

How to use this module in a Label control ? bruvzg/godot_tl#3

Open

starry-abyss mentioned this pull request Apr 24, 2019

Overlapping display of the upper vowel and the tone character in Thai Language #28371

Closed

akien-mga modified the milestones: 3.2, 4.0 Oct 4, 2019

aaronfranke marked this pull request as draft April 8, 2020 23:46

bruvzg closed this May 13, 2020

aaronfranke added the archived label May 13, 2020

akien-mga added the salvageable label May 13, 2020

bruvzg mentioned this pull request Nov 10, 2020

[Complex Text Layouts] Implement TextServer interface. #41100

Merged

bruvzg deleted the min-shaping branch November 5, 2021 12:41

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[WIP] BiDi and shaping support for Font::draw/Font::get_string_size #21791

[WIP] BiDi and shaping support for Font::draw/Font::get_string_size #21791

bruvzg commented Sep 6, 2018 •

edited

Loading

bruvzg commented Sep 6, 2018 •

edited

Loading

akien-mga Sep 13, 2018

akien-mga Sep 13, 2018

akien-mga Sep 13, 2018

akien-mga Sep 13, 2018

akien-mga Sep 13, 2018

akien-mga Sep 13, 2018

akien-mga Sep 13, 2018

leonkrause Sep 16, 2018

akien-mga Sep 13, 2018

reduz commented Sep 16, 2018

bruvzg commented Sep 27, 2018

bruvzg commented Dec 29, 2018

slevy85 commented Apr 11, 2019 •

edited

Loading

realkotob commented Sep 2, 2019 •

edited

Loading

MohammadKhashashneh commented Oct 1, 2019 •

edited

Loading

MohammadKhashashneh commented Oct 2, 2019 •

edited

Loading

akien-mga commented Oct 4, 2019

MohammadKhashashneh commented Oct 5, 2019

OmarAglan commented Oct 30, 2019

bruvzg commented Oct 30, 2019

aaronfranke commented May 13, 2020 •

edited

Loading

jitspoe commented Oct 14, 2020

bruvzg commented Oct 14, 2020

nightblade9 commented Oct 14, 2020 •

edited

Loading

[WIP] BiDi and shaping support for Font::draw/Font::get_string_size #21791

[WIP] BiDi and shaping support for Font::draw/Font::get_string_size #21791

Conversation

bruvzg commented Sep 6, 2018 • edited Loading

bruvzg commented Sep 6, 2018 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

reduz commented Sep 16, 2018

bruvzg commented Sep 27, 2018

bruvzg commented Dec 29, 2018

slevy85 commented Apr 11, 2019 • edited Loading

realkotob commented Sep 2, 2019 • edited Loading

MohammadKhashashneh commented Oct 1, 2019 • edited Loading

MohammadKhashashneh commented Oct 2, 2019 • edited Loading

akien-mga commented Oct 4, 2019

MohammadKhashashneh commented Oct 5, 2019

OmarAglan commented Oct 30, 2019

bruvzg commented Oct 30, 2019

aaronfranke commented May 13, 2020 • edited Loading

jitspoe commented Oct 14, 2020

bruvzg commented Oct 14, 2020

nightblade9 commented Oct 14, 2020 • edited Loading

bruvzg commented Sep 6, 2018 •

edited

Loading

bruvzg commented Sep 6, 2018 •

edited

Loading

slevy85 commented Apr 11, 2019 •

edited

Loading

realkotob commented Sep 2, 2019 •

edited

Loading

MohammadKhashashneh commented Oct 1, 2019 •

edited

Loading

MohammadKhashashneh commented Oct 2, 2019 •

edited

Loading

aaronfranke commented May 13, 2020 •

edited

Loading

nightblade9 commented Oct 14, 2020 •

edited

Loading