Merge fea-rs repo #505

cmyr · 2023-10-25T16:52:21Z

This merges the history of the fea-rs repo, which will be subsequently archived.

After this is merged I will:

make sure that fea-rs is setting up its venv during testing correctly
make fea-rs use the workspace deps where appropriate
make crates here depend on the local version of fea-rs
update the metadata & links in the fea-rs cargo.toml and readme
fix anything else that's broken?

This simplifies our imports, and prevents any possible future version conflicts.

This is about updating the API for applying a compilation result to an input font to fit the read/write-fonts division.

This no longer makes sense, as we can just use the write-fonts type and save ourselves some complication.

This moves us off of specifying write-fonts as a local dependency, which means other people can check out this commit and actually build the project, which seems like a reasonable desire. This also cleans up a bunch of dead code etc.

This renames the compile::consts module to compile::common, and moves these things in there. These functions will be shared between validation (which doesn't seem to be happening yet?) and compilation.

This simplifies the logic around generating the name table, where entries come from various sources.

This is a refactor/cleanup, no functional change.

This is a shim that lets us easily build a font with only FEA and a glyph map.

As a consequence we can no longer compile when we are passed a font, instead of a glyph order. Is this better? Is this worse? It is a curious question.

This doesn't currently pass the tests because fontations can't handle mac-japanese encoding, but I'll get around to that later.

- bump version - update deps (and adjust to new API, in the case of xflags) - pin to crates.io version of fontations

- no need to pass glyph map - return builder, so more work can be done afterwards - passing in a font is optional

This further ensures that our output matches the output of fonttools.

This updates our logic to better match how the behaviour of feaLib, which I think is valuable for the time being while we use the feaLib test suite as our benchmark.

e.g. they're all suffixed with 'Builder', and I'm also cleaning up some of the old non-rustic names and removing associated `#[allow]` annotations.

This was blocked on upstream support.

We didn't support named value records for a while, and so this had been omitted, but support was added a few months ago.

(cleaning up some old fixmes)

This will hopefully reduce the chance that we run into differences when reading/writing string values.

This was cluttering up logs when I was debugging pairpos2 splitting.

I'm not sure if this was triggerable during real use, but I was hitting it in some test code where we don't have a real variation model, and so were passing empty deltas.

This matches the behaviour of fontmake, although it deviates from makeotf.

This deletes the var_store module, which has moved to write-fonts.

Not directly needed by fea-rs but fontc (user of fea-rs) needs it and can't update otherwise

bump write-fonts to 0.17.0

bump minor because write-fonts was bumped minor

[clippy] Push the rock back up the hill

This will make it easier for us to debug and profile fontc, by letting us run feature compilation in isolation, on our real inputs. axis info can be specified in a file, where each (non blank, non commented) line contains a tag and three numbers.

rsheeter

LGTM. I think you can delete gitmodules?

Also ignoring _experiments is extremely Colin-specific; I suggest we dont' keep that change.

cmyr added 30 commits November 29, 2022 11:13

Handle contextual ignore statements

cea8e74

Compile reverse chain sub rules

22b573c

Use read/types via write_fonts reexport

997e0b8

This simplifies our imports, and prevents any possible future version conflicts.

Fix for new hhea/vhea API

74d4686

fixup imports

9ff0b85

Improve compiling against existing font

a48e16c

This is about updating the API for applying a compilation result to an input font to fit the read/write-fonts division.

Compile device tables in value records

081c17e

Remove redundant anchor type

ce91561

This no longer makes sense, as we can just use the write-fonts type and save ourselves some complication.

Compile devices in anchor tables

4b2e555

Compile vhea

e4e44d9

Clippy fixups && pin write-fonts to git rev

46a7cf8

This moves us off of specifying write-fonts as a local dependency, which means other people can check out this commit and actually build the project, which seems like a reasonable desire. This also cleans up a bunch of dead code etc.

Shared fns for ss01-ss20 and cv01-cv99 Tags

69f0004

This renames the compile::consts module to compile::common, and moves these things in there. These functions will be shared between validation (which doesn't seem to be happening yet?) and compilation.

Move const tags into common::tags

672778c

Add NameBuilder

f33af88

This simplifies the logic around generating the name table, where entries come from various sources.

Rename STAT -> StatBuilder, add build() fn

07985c8

This is a refactor/cleanup, no functional change.

Add Compilation::build_raw

a712bb1

This is a shim that lets us easily build a font with only FEA and a glyph map.

Update bin/compile to accept glyph order input

be018b9

As a consequence we can no longer compile when we are passed a font, instead of a glyph order. Is this better? Is this worse? It is a curious question.

Test glyph order loading

6790b75

Compile ss01-ss20 feature params

f9e7532

This doesn't currently pass the tests because fontations can't handle mac-japanese encoding, but I'll get around to that later.

Only compile name entries with known encodings

a44137d

Compile cv01-cv99 feature params

86aa9e1

Cleanup for 0.0.1 release

08cbba1

- bump version - update deps (and adjust to new API, in the case of xflags) - pin to crates.io version of fontations

Fix manifest readme entry

738ea57

Update write-fonts, get working with API changes

06ca178

Tweak build API

4a4e92f

- no need to pass glyph map - return builder, so more work can be done afterwards - passing in a font is optional

Write out post table in ttx tests

5372da6

This further ensures that our output matches the output of fonttools.

Fix STAT compilation

75f923d

This updates our logic to better match how the behaviour of feaLib, which I think is valuable for the time being while we use the feaLib test suite as our benchmark.

Compile BASE table

71f74e7

More consistent table builder names

070662f

e.g. they're all suffixed with 'Builder', and I'm also cleaning up some of the old non-rustic names and removing associated `#[allow]` annotations.

Compile OS/2 table

9a53ab1

This was blocked on upstream support.

cmyr and others added 22 commits August 23, 2023 10:37

Remove commented out value record format test case

892dd23

We didn't support named value records for a while, and so this had been omitted, but support was added a few months ago.

Ensure we compile arbitrarily long device records

8c9dbac

(cleaning up some old fixmes)

Store AxisLocation as f32, not Fixed

39f08e3

This will hopefully reduce the chance that we run into differences when reading/writing string values.

Bump version to 0.15.0

853477f

[rustfmt] Run updated formatter

617ec7f

Only print up to some maximum number of errors

5b23c17

Add compiler option to specify number of errors to print

d7bcd76

Change format size log level to 'trace'

98396b3

This was cluttering up logs when I was debugging pairpos2 splitting.

Bump crate version, update write-fonts

9676a6e

Update to new resolver

88c7c0e

Fix crash when trying to write all-zero deltas

497f51c

I'm not sure if this was triggerable during real use, but I was hitting it in some test code where we don't have a real variation model, and so were passing empty deltas.

Accept naked semi ('null statements')

bf56008

This matches the behaviour of fontmake, although it deviates from makeotf.

Bump version to 0.16.1

b42e96d

Update deps, bump version

8609462

This deletes the var_store module, which has moved to write-fonts.

bump write-fonts to 0.17.0

478a183

Not directly needed by fea-rs but fontc (user of fea-rs) needs it and can't update otherwise

Merge pull request #233 from cmyr/bump-write-fonts

6827cf9

bump write-fonts to 0.17.0

bump version

0ff8c5a

bump minor because write-fonts was bumped minor

[clippy] Push the rock back up the hill

19a3667

Merge pull request #237 from cmyr/clippy-time

04fc44f

[clippy] Push the rock back up the hill

Clean up in preparation for move to fontc repo

5f8590e

Merge fea-rs into fontc

739f143

rsheeter approved these changes Oct 25, 2023

View reviewed changes

Fixup test path discovery

e303f72

cmyr force-pushed the merge-fea-rs branch from 086e71e to e303f72 Compare October 25, 2023 17:28

cmyr added this pull request to the merge queue Oct 25, 2023

Merged via the queue into main with commit cb9722c Oct 25, 2023

cmyr deleted the merge-fea-rs branch October 25, 2023 19:34

cmyr mentioned this pull request Oct 25, 2023

[ci] Merge fea-rs actions into main CI workflow #511

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Merge fea-rs repo #505

Merge fea-rs repo #505

cmyr commented Oct 25, 2023

rsheeter left a comment

Merge fea-rs repo #505

Merge fea-rs repo #505

Conversation

cmyr commented Oct 25, 2023

rsheeter left a comment

Choose a reason for hiding this comment