CHANGELOG.md

Change Log

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog and this project adheres to Semantic Versioning.

Unreleased

Added

• alloc build feature, for no_std builds with an allocator. Name::to_string and the gvar-alloc heap spill no longer require std. Thanks to iriswebb.
• (cmap) Subtable4::codepoints_iter and Subtable12::codepoints_iter. Unlike codepoints, they allow early exit. Both take self by value. Thanks to MoSal.
• CFFError::SubroutineCallLimitReached. Thanks to scadastrangelove.
• core::error::Error is implemented for FaceParsingError in no_std builds now. Thanks to iriswebb.
• (CFF) cff::Table::parse_with_upem, which applies the font matrix using the face's units per em. cff::Table::parse keeps its original one-argument signature. Thanks to LaurenzV.
• (OS/2) os2::Table::panose, returning the ten raw PANOSE digits plus FamilyType and is_bold/is_italic/is_monospaced. The digits are public because their meaning depends on the family type and PANOSE defines more values than this crate models. Face::is_bold, Face::is_italic and Face::is_monospaced deliberately do not consult it: PANOSE is a design classification, not a style flag. Thanks to inferiorhumanorgans.

Changed

• (loca) loca::Table::len returns u32 instead of u16, and the Table::Short/Table::Long payloads are LazyArray32 instead of LazyArray16. Thanks to asibahi.
• The gvar-alloc build feature implies alloc instead of std now. A default-features = false, features = ["gvar-alloc"] build no longer satisfies the std/no-std-float guard and must select one of them explicitly. Thanks to iriswebb.
• Face::set_variation returns None for an axis tag the face does not expose. It previously returned Some(()) unconditionally, contradicting its own documentation. This is a behavioural change and will not produce a compile error. Thanks to asibahi.
• CFFError is #[non_exhaustive] now, so that future hardening limits can add variants without breaking downstream code. Together with the new SubroutineCallLimitReached variant, this does break existing exhaustive matches on it.
• Face::is_italic no longer treats a nonzero post.italicAngle as italic. It now consults the two flags the spec treats as authoritative: OS/2.fsSelection and head.macStyle bit 1, the latter of which was not parsed at all before. The angle is a slant value that the spec only says "should" be 0 for upright faces, and real fonts violate that while rendering perfectly upright, so it produced false positives. This is a behavioural change and will not produce a compile error.
• (head) head::Table has a new public is_italic field, which breaks exhaustive struct destructuring of it.
• (COLR) push_clip and pop_clip are called when painting v0 glyphs, matching v1. A Painter that relied on v0 never clipping will render differently. Thanks to valadaptive.
• Bump MSRV to 1.88 and move to edition 2024.
• Test fonts and development tools are excluded from the published package, which is substantially smaller as a result. Thanks to weiznich.
• Dependencies are pinned to a minor version rather than an exact one. Thanks to alerque.
• Documented the Face::style fallback to Style::Normal when OS/2 is absent. Thanks to dereified.
• Documented every field of RawFaceTables.

Fixed

• (CFF) Bound the total number of subroutine invocations per glyph. STACK_LIMIT bounds nesting depth only, so work grew as fanout^depth: a 255-byte CFF table hung for over 30 seconds. Thanks to scadastrangelove.
• (GSUB/GPOS) Reject self-referential extension lookups. An extension lookup pointing at the extension type recursed unboundedly: a stack-overflow abort in debug and an infinite loop in release, from 8 bytes of input. Thanks to scadastrangelove.
• (loca) Support tables with 65536 offsets. A well-formed 65535-glyph font was rejected, glyf was dropped entirely and every outline returned None. Thanks to asibahi.
• (glyf/gvar) Cap the total number of component visits when outlining composite glyphs. MAX_COMPONENTS bounds chain depth only, so shared child glyphs let sibling components force branching^depth re-walks under the cap. Thanks to scadastrangelove.
• (COLR) Cap the total number of paint-graph node visits in v1 painting. The recursion stack bounds path depth but not fan-out, and layers_count is a u8, so a small font could force an enormous traversal. Thanks to scadastrangelove.
• (CFF2) Guard the BLEND operator against an empty argument stack. Thanks to scadastrangelove.
• (CFF) Glyphs containing the deprecated dotsection operator are no longer rejected. Previously the whole charstring failed with UnsupportedOperator, so glyphs like i, j, ! and . in fonts converted from Type 1 produced no outline at all. Thanks to tobocop2.
• (GSUB/GPOS) LazyOffsetArrayIter16 skips NULL offsets instead of ending iteration. A NULL offset hid every later entry, so valid fonts silently lost sequence rule sets. Thanks to asibahi.
• (GSUB/GPOS) LazyOffsetArrayIter16::count no longer over-reports by the number of skipped entries and agrees with collect().len() now.
• (avar) map_value arithmetic no longer overflows. Unvalidated i16 coordinates could panic in debug or silently invert the sign in release. Thanks to scadastrangelove.
• (fvar) Read the HIDDEN_AXIS flag from bit 0, not bit 3. Hidden axes were never reported for conforming fonts. Thanks to SamusAranX.
• (C API) ttfp_get_glyph_name rejects glyph names longer than the documented 256-byte buffer instead of panicking across the extern "C" boundary, which is a non-unwinding abort. Thanks to scadastrangelove.
• (glyf) Composite components using point matching no longer desync the parser. The two component arguments were consumed only on the ARGS_ARE_XY_VALUES path, so an unsupported point-matching component left the stream positioned mid-arguments and every subsequent read in that glyph decoded from the wrong offset. Point matching itself remains unimplemented, but it no longer corrupts the components that follow it.
• (fvar) Face::set_variation works on a face with exactly 64 axes. Its guard rejected an axis count equal to the coordinate-storage limit, while every other site clamps to that limit inclusively, so such a face could not be varied on any axis at all.
• Stream::read_bytes bounds the requested length against the buffer instead of asserting on offset + len. The assert was meant to catch caller bugs, but malformed fonts reach it just as easily through a CFF INDEX count, a cbdt PNG length, a GPOS class-pair product and others, so debug builds aborted on input that release builds rejected cleanly — noise for anyone fuzzing a debug build. Checking first also makes the addition sound on 32-bit targets, which the assert never did.
• (HVAR/VVAR/COLR) A delta-set index map no longer advances past its buffer before reading. mapCount is a raw u32 and the index is scaled by the entry size.
• (COLR) Variation delta indices are added with overflow checking. A varIndexBase near u32::MAX aborted in debug and wrapped to an unrelated delta index in release.
• (CFF) Glyph outlines are transformed by the font matrix. Fonts with a non-default FontMatrix were outlined at the wrong scale. Thanks to tbodt.
• (CFF2) Fonts without a vstore can outline glyphs. The Top DICT entry is optional per spec, but variation scalars were resolved at index 0 before the first operator ran, which fails on an absent store — so every glyph of a static CFF2 font returned None. Scalars are now resolved on first blend instead, which is the only operator that needs them.
• (gvar) Outlining a variable glyph no longer aborts under -Zrandomize-layout. Two debug_assert!s bounded the size of internal iterator structs, but Rust does not guarantee repr(Rust) layout and randomization pushed them past the bound, so every debug build of every application using the flag panicked on a budget it could not influence. The bounds are now checked in this crate's own test suite instead.

0.25.1 - 2024-11-29

Changed

• Fix some typos and documentation related lints
• Set MSRV in manifest to actual current value (1.63.0)
• Update forge URL for ttf-parser and rustybuzz to harfbuzz org

0.25.0 - 2024-10-04

Added

• STAT table parsing. Thanks to inferiorhumanorgans.
• UnicodeRanges internal field is public now.
• cargo-c metadata to C API. Thanks to lu-zero.

Changed

• Face::is_italic checks for italic angle as well.
• Face::italic_angle returns just a f32 and not Option<f32> now.
• Bump MSRV to 1.59

Fixed

• Only apply avar table to the variation axis being set. Thanks to maxmelander.

0.24.1 - 2024-08-05

Added

• (glyf) glyf::Table::bbox. Thanks to LaurenzV.

Fixed

• (kerx) kerx::SubtablesIter wasn't updating the current subtable index.
• (GPOS) gpos::AnchorMatrix allows nullable/optional matrices now. Thanks to LaurenzV.

0.24.0 - 2024-07-02

Changed

• Make core_maths dependency optional. When building for no_std one must enable no-std-float build feature now.

0.23.0 - 2024-07-02

Changed

• Use core_maths instead of libm. Should simplify the build process. Thanks to LaurenzV.

Removed

• no-std-float build flag. Should be handled automatically now.

0.22.0 - 2024-06-29

Added

• Face::glyph_phantom_points
• hvar::Table::right_side_bearing_offset. Thanks to LaurenzV.
• vvar::Table::advance_offset. Thanks to LaurenzV.
• vvar::Table::top_side_bearing_offset. Thanks to LaurenzV.
• vvar::Table::bottom_side_bearing_offset. Thanks to LaurenzV.
• vvar::Table::vertical_origin_offset. Thanks to LaurenzV.
• colr::Table::clip_box. Thanks to LaurenzV.

Changed

• no_std build of ttf-parser requires --features=no-std-float now. This is because we need trigonometry functions to flatten transforms in COLR. Thanks to LaurenzV.
• colr::Painter no longer has push_translate, push_scale, push_rotate and push_skew. Only push_transform left. Thanks to LaurenzV.
• Split hvar::Table into hvar::Table and vvar::Table. Previously, we treated both HVAR and VVAR tables as identical, but VVAR actually has additional fields. Thanks to LaurenzV.
• Rename hvar::Table::side_bearing_offset into hvar::Table::left_side_bearing_offset. Thanks to LaurenzV.

Fixed

• Face::glyph_hor_advance and Face::glyph_ver_advance include gvar's phantom points when HVAR/VVAR tables are missing. Affects only variable fonts.
• (CFF) Allow MoveTo with width commands in nested subroutines.
• opentype_layout::LookupFlags::mark_attachment_type parsing.
• (CFF) Allow empty charsets in cff::parse_charset. Thanks to LaurenzV.
• (gvar) Empty sub-glyphs/components is no longer an error. Thanks to LaurenzV.
• (GSUB/GPOS) Allow NULL offsets in ChainedContextLookup Format2 subtables. Thanks to LaurenzV.
• Face::glyph_y_origin properly handles variable fonts now. Thanks to LaurenzV.
• (kerx) Fix AnchorPoints parsing. Thanks to LaurenzV.

Removed

• push_translate, push_scale, push_rotate and push_skew from colr::Painter. Use colr::Painter::push_transform instead. Thanks to LaurenzV.

0.21.1 - 2024-05-11

Fixed

• Delta set length calculation in variable fonts. Thanks to LaurenzV.
  Got broken in the previous version.

0.21.0 - 2024-05-10

Added

• COLR / CPAL v1 support. Thanks to LaurenzV.

Changed

• Replace Face::is_bitmap_embedding_allowed with Face::is_outline_embedding_allowed. The old one had a bool flag flipped. Thanks to Fuzzyzilla.
• Increase lenience of embed permissions for older OS/2 versions. Thanks to Fuzzyzilla.
• Bump MSRV to 1.51

0.20.0 - 2023-10-15

Added

• COLR / CPAL v0 support. Thanks to laurmaedje.
• Face::is_color_glyph
• Face::color_palettes
• cpal::Table::palettes

Changed

• svg::SvgDocumentsList returns svg::SvgDocument and not just &[u8] now. Thanks to wjian23.
• Face::set_variation allows duplicated axes now.
• cpal::Color is called cpal::BgraColor now, and cpal::Table::get returns it.

0.19.2 - 2023-09-13

Added

• cff::Table::glyph_cid

0.19.1 - 2023-06-20

Fixed

• cff::Table::glyph_width returns a correct width when subroutines are present.

0.19.0 - 2023-04-17

Added

• bdat, bloc, EBDT and EBLC tables support. Thanks to dzamkov.
• BitmapMono, BitmapMonoPacked, BitmapGray2, BitmapGray2Packed, BitmapGray4, BitmapGray4Packed, BitmapGray8 and BitmapPremulBgra32 variants to RasterImageFormat.

Fixed

• CBLC table parsing. Thanks to dzamkov.

0.18.1 - 2023-01-10

Fixed

• (MATH) Handle NULL offsets. Thanks to laurmaedje.

0.18.0 - 2022-12-25

Added

• name::Name::language
• Language enum with all Windows languages.

0.17.1 - 2022-10-15

Added

• Face::permissions
• Face::is_subsetting_allowed
• Face::is_bitmap_embedding_allowed
• Face::unicode_ranges
• os2::Table::permissions
• os2::Table::is_subsetting_allowed
• os2::Table::is_bitmap_embedding_allowed
• os2::Table::unicode_ranges

Changed

• Using a non-zero index in Face::parse for a regular font will return FaceParsingError::FaceIndexOutOfBounds now. Thanks to Pietrek14.

0.17.0 - 2022-09-28

Added

• MATH table support. Thanks to ruifengx and laurmaedje.

Fixed

• (CFF) Fix large tables parsing.

0.16.0 - 2022-09-28

Added

• CFF Encoding support.
• cff::Table::glyph_index
• cff::Table::glyph_index_by_name
• cff::Table::glyph_width
• cff::Table::number_of_glyphs
• cff::Table::matrix
• post::Table::glyph_name
• post::Table::glyph_index_by_name
• post::Table::names
• Face::glyph_index_by_name
• RawFace fields and TableRecord struct are public now.

Changed

• Face::from_slice was replaced by Face::parse.
• RawFace::from_slice was replaced by RawFace::parse.
• post::Table::names is a method and not a field now.
• Use post::Table::glyph_name instead of post::Table::names.get().

Fixed

• (hmtx/vmtx) Allow missing additional side bearings.
• (loca) Allow incomplete table.
• Reduce strictness of some table length checks.
• (post) post::Names::len was returning a wrong value. Now this method is gone completely. You can use post::Table::names().count() instead.

0.15.2 - 2022-06-17

Fixed

• Missing advance and side bearing offsets in HVAR/VVAR is not an error. Simply ignore them.

0.15.1 - 2022-06-04

Fixed

• (cmap) cmap::Subtable4::glyph_index correctly handles malformed glyph offsets now.
• (cmap) cmap::Subtable4::codepoints no longer includes 0xFFFF codepoint.
• (SVG) Fixed table parsing. Thanks to Shubhamj280

0.15.0 - 2022-02-20

Added

• apple-layout build feature.
• ankr, feat, kerx, morx and trak tables.
• kern AAT subtable format 1.
• RawFace

Changed

• The parser module is private now again.

0.14.0 - 2021-12-28

Changed

• (cmap) cmap::Subtable::glyph_index and cmap::Subtable::glyph_variation_index accept u32 instead of char now.
• (glyf) ~7% faster outline parsing.

0.13.4 - 2021-11-23

Fixed

• (CFF) Panic during seac resolving.
• (CFF) Stack overflow during seac resolving.

0.13.3 - 2021-11-19

Fixed

• (glyf) Endless loop during malformed file parsing.

0.13.2 - 2021-10-28

Added

• gvar-alloc build feature that unlocks gvar table limits by using heap. Thanks to OrionNebula

0.13.1 - 2021-10-27

Fixed

• Face::line_gap logic.

0.13.0 - 2021-10-24

Added

• Complete GSUB and GPOS tables support. Available under the opentype-layout feature.
• Public access to all supported TrueType tables. This allows a low-level, but still safe, access to internal data layout, which can be used for performance optimization, like caching.
• Style enum and Face::style method.
• Face::glyph_name can be disabled via the glyph-names feature to reduce binary size.

Changed

• Improved ascender/descender/line_gap resolving logic.
• Face methods: has_glyph_classes, glyph_class, glyph_mark_attachment_class, is_mark_glyph and glyph_variation_delta are moved to gdef::Table.
• The Names struct is no longer an iterator, but a container. You have to call into_iter() manually.
• The VariationAxes struct is no longer an iterator, but a container. You have to call into_iter() manually.
• Most of the Name struct methods become public fields.
• Face::units_per_em no longer returns Option.
• (cmap) Improved subtable 12 performance. Thanks to xnuk

Removed

• (c-api) ttfp_glyph_class, ttfp_get_glyph_class, ttfp_get_glyph_mark_attachment_class, ttfp_is_mark_glyph, ttfp_glyph_variation_delta and ttfp_has_table.
• TableName enum and Face::has_table. Tables can be access directly now.
• Face::character_mapping_subtables. Use Face::tables().cmap instead.
• Face::kerning_subtables. Use Face::tables().kern instead.

Fixed

• Iterator::count implementation for cmap::Subtables, name::Names and LazyArrayIter32.

0.12.3 - 2021-06-27

Changed

• (glyf) Always use a calculated bbox.

0.12.2 - 2021-06-11

Fixed

• Face::glyph_bounding_box for variable glyf.
• (glyf) Do not skip glyphs with zero-sized bbox.

0.12.1 - 2021-05-24

Added

• Support Format 13 subtables in cmap::Subtable::is_unicode. Thanks to csmulhern
• Derive more traits by default. Thanks to dhardy

0.12.0 - 2021-02-14

Changed

• Face::ascender and Face::descender will use usWinAscent and usWinDescent when USE_TYPO_METRICS flag is not set in OS/2 table. Previously, those values were ignored and hhea::ascender and hhea::descender were used. Now hhea table values will be used only when OS/2 table is not present.
• Face::outline_glyph and Face::glyph_bounding_box in case of a glyf table can fallback to a calculated bbox when the embedded bbox is malformed now.

0.11.0 - 2021-02-04

Added

• FaceTables, which allowed to load Face not only from a single chunk of data, but also in a per-table way. Which is useful for WOFF parsing. No changes to the API. Thanks to fschutt

0.10.1 - 2021-01-21

Changed

• Update a font used for tests.

0.10.0 - 2021-01-16

Added

• variable-fonts build feature. Enabled by default. By disabling it you can reduce ttf-parser binary size overhead almost twice.

Changed

• (gvar) Increase the maximum number of variation tuples from 16 to 32. Increases stack usage and makes gvar parsing 10% slower now.

Fixed

• (CFF) Fix seac processing. Thanks to wezm

0.9.0 - 2020-12-05

Removed

• kern AAT subtable 1 aka kern::state_machine. Mainly because it's useless without a proper shaping.

0.8.3 - 2020-11-15

Added

• Face::glyph_variation_delta
• The parser module is public now. Made private again in 0.15.0.
• (C API) ttfp_glyph_variation_delta

Fixed

• Iterator::nth implementation for cmap::Subtables and Names.

0.8.2 - 2020-07-31

Added

• cmap::Subtable::codepoints

Fixed

• (cmap) Incorrectly returning glyph ID 0 instead of None for format 0
• (cmap) Possible invalid glyph mapping for format 2

0.8.1 - 2020-07-29

Added

• Face::is_monospaced
• Face::italic_angle
• Face::typographic_ascender
• Face::typographic_descender
• Face::typographic_line_gap
• Face::capital_height

0.8.0 - 2020-07-21

Added

• Allow true magic.
• FaceParsingError
• NormalizedCoordinate
• Face::variation_coordinates
• Face::has_non_default_variation_coordinates
• Face::glyph_name can lookup CFF names too.
• Face::table_data
• Face::character_mapping_subtables

Changed

• (CFF,CFF2) 10% faster parsing.
• Face::from_slice returns Result now.
• Name::platform_id returns PlatformId instead of Option<PlatformId> now.
• The cmap module became public.

Fixed

• Face::width parsing.
• Possible u32 overflow on 32-bit platforms during Face::from_slice.
• (cmap) Face::glyph_variation_index processing when the encoding table has only one glyph.

0.7.0 - 2020-07-16

Added

• (CFF) CID fonts support.
• (CFF) seac support.
• Font::global_bounding_box

Changed

• Rename Font to Face, because this is what it actually is.
• Rename Font::from_data to Font::from_slice to match serde and other libraries.
• Rename Name::name_utf8 to Name::to_string.

Removed

• Font::family_name and Font::post_script_name. They were a bit confusing. Prefer:

  face.names().find(|name| name.name_id() == name_id::FULL_NAME).and_then(|name| name.to_string())
  

0.6.2 - 2020-07-02

Added

• Name::is_unicode
• Font::family_name will load names with Windows Symbol encoding now.

Fixed

• Font::glyph_bounding_box will apply variation in case of gvar fonts.

0.6.1 - 2020-05-19

Fixed

• (kern) Support fonts that ignore the subtable size limit.

0.6.0 - 2020-05-18

Added

• sbix, CBLC, CBDT and SVG tables support.
• Font::glyph_raster_image and Font::glyph_svg_image.
• Font::kerning_subtables with subtable formats 0..3 support.

Changed

• (c-api) The library doesn't allocate ttfp_font anymore. All allocations should be handled by the caller from now.

Removed

• Font::glyphs_kerning. Use Font::kerning_subtables instead.
• (c-api) ttfp_create_font and ttfp_destroy_font. Use ttfp_font_size_of + ttfp_font_init instead.

  ttfp_font *font = (ttfp_font*)alloca(ttfp_font_size_of());
  ttfp_font_init(font_data, font_data_size, 0, font);

• Logging support. We haven't used it anyway.

Fixed

• (gvar) Integer overflow.
• (cmap) Integer overflow during subtable format 2 parsing.
• (CFF, CFF2) DICT number parsing.
• Font::glyph_*_advance will return None when glyph ID is larger than the number of metrics in the table.
• Ignore variation offset in Font::glyph_*_advance and Font::glyph_*_side_bearing when HVAR/VVAR tables are missing. Previously returned None which is incorrect.

0.5.0 - 2020-03-19

Added

• Variable fonts support.
• C API.
• gvar, CFF2, avar, fvar, HVAR, VVAR and MVAR tables support.
• Font::variation_axes
• Font::set_variation
• Font::is_variable
• Tag type.

Fixed

• Multiple issues due to arithmetic overflow.

0.4.0 - 2020-02-24

A major rewrite.

Added

• Font::glyph_bounding_box
• Font::glyph_name
• Font::has_glyph_classes
• Font::glyph_class
• Font::glyph_mark_attachment_class
• Font::is_mark_glyph
• Font::glyph_y_origin
• Font::vertical_ascender
• Font::vertical_descender
• Font::vertical_height
• Font::vertical_line_gap
• Optional log dependency.

Changed

• Font::outline_glyph now accepts &mut dyn OutlineBuilder and not &mut impl OutlineBuilder.
• Font::ascender, Font::descender and Font::line_gap will check USE_TYPO_METRICS flag in OS/2 table now.
• glyph_hor_metrics was split into glyph_hor_advance and glyph_hor_side_bearing.
• glyph_ver_metrics was split into glyph_ver_advance and glyph_ver_side_bearing.
• CFFError is no longer public.

Removed

• Error enum. All methods will return Option<T> now.
• All unsafe.

Fixed

• glyph_hor_side_bearing parsing when the number of metrics is less than the total number of glyphs.
• Multiple CFF parsing fixes. The parser is more strict now.

0.3.0 - 2019-09-26

Added

• no_std compatibility.

Changed

• The library has one unsafe block now.
• 35% faster family_name() method.
• 25% faster from_data() method for TrueType fonts.
• The Name struct has a new API. Public fields became public functions and data is parsed on demand and not beforehand.

0.2.2 - 2019-08-12

Fixed

• Allow format 12 subtables with Unicode full repertoire in cmap.

0.2.1 - 2019-08-12

Fixed

• Check that cmap subtable encoding is Unicode.

0.2.0 - 2019-07-10

Added

• CFF support.
• Basic kerning support.
• All cmap subtable formats except Mixed Coverage (8) are supported.
• Vertical metrics querying from the vmtx table.
• OpenType fonts are allowed now.

Changed

• A major rewrite. TrueType tables are no longer public.
• Use GlyphId instead of u16.

Removed

• GDEF table parsing.

0.1.0 - 2019-06-18

Added

• Initial release. Parsing of cmap, glyf, head, hhea, hmtx, loca, maxp, name, OS/2, post, vhea and vmtx, with GDEF and font-collection support.