Use any installed Noto font when named families lack glyphs (closes #1722)

SystemFontProvider.NOTO_FONT_PATTERNS enumerates about two dozen Noto
families by name, and MultiFontManager only ever asked for those. Any
script outside that list fell through to the glyphless Occulta fallback
even when the correct font was installed, which is the common case on
macOS: it ships around a hundred script-specific Noto faces in
/System/Library/Fonts/Supplemental, almost none of which we knew how to
ask for. The reporter had the fonts and still got told to install them.

Add an optional GlyphSearchingFontProvider protocol (find_font_with_glyphs)
implemented by SystemFontProvider, BuiltinFontProvider and
ChainedFontProvider, and a new selection phase that uses it once the named
families fail. The system scan enumerates every Noto face, reusing the
existing variable-font/-Regular/-VF filename classification, and keeps
only the font it selects so a full scan does not retain every font file on
the system. Fonts found this way are remembered and retried by name for
later words. Capability detection is isinstance-based so third-party
providers keep working unchanged.

The warning itself was also unactionable: it named neither the characters
nor the script, which is why the reporter had to ask which package to
install. It now identifies what it could not render, e.g.
'Ꮳ' U+13E3 CHEROKEE LETTER TSA. A word mixing scripts that no single font
covers now gets a distinct message saying that installing fonts will not
help, rather than sending the user after fonts they already have.

Finally, the documented macOS install command was wrong: `brew install
font-noto` does not exist, since Homebrew has no single Noto package, only
a cask per family. The Fedora package name was also corrected, as
google-noto-fonts-common ships no actual fonts.
This commit is contained in:
James R. Barlow
2026-07-26 23:25:51 -07:00
parent f0a3a74374
commit 5d49f75c56
9 changed files with 672 additions and 16 deletions
+19 -3
View File
@@ -663,9 +663,25 @@ provides text shaping for proper multilingual support. These replace the
legacy hOCR-based renderer. Install with: `pip install fpdf2 uharfbuzz`
**fonts-noto** (or an equivalent comprehensive font package) is recommended
for proper text rendering, especially for non-Latin scripts. On Debian/Ubuntu:
`apt install fonts-noto`. On Fedora: `dnf install google-noto-fonts-common`.
On macOS with Homebrew: `brew install font-noto`.
for proper text rendering, especially for non-Latin scripts. OCRmyPDF bundles
a Latin font only, and discovers the rest from the fonts installed on your
system.
- Debian/Ubuntu: `apt install fonts-noto`
- Fedora: `dnf install google-noto-fonts-all`
- macOS with Homebrew: Homebrew has no single Noto package; each family is a
separate cask. Install at least
`brew install --cask font-noto-sans font-noto-serif`, plus a cask per
additional script you OCR, for example
`brew install --cask font-noto-sans-arabic font-noto-sans-cjk`. Run
`brew search font-noto` to list them all.
If OCRmyPDF warns that no installed font has glyphs for some of the text, the
message names the characters it could not render, for example
`'Ꮳ' U+13E3 CHEROKEE LETTER TSA`. Install the Noto font for that script — here,
`fonts-noto-core` on Debian or `font-noto-sans-cherokee` on Homebrew. The text
layer remains searchable and copyable either way; only its appearance when
highlighted in a PDF viewer is affected.
**pypdfium2**, if present, provides fast PDF page rasterization using
the pdfium library (the same library used by Google Chrome). It is
+3 -1
View File
@@ -47,7 +47,9 @@ OCRmyPDF has the following runtime dependencies:
**For text rendering** (expressing OCR results in PDF):
- `fpdf2` (Python package) - Required for text layer rendering
- `uharfbuzz` (Python package) - Required for text layer rendering
- `font-noto` (system package) - Recommended for text layer rendering
- Noto fonts (system package) - Recommended for text layer rendering.
`fonts-noto` on Debian/Ubuntu, `google-noto-fonts-all` on Fedora; Homebrew
has no single Noto package, only per-family casks such as `font-noto-sans`.
**Other dependencies**:
- `unpaper` (system binary) - Optional, enables `--clean` and `--clean-final`
+21
View File
@@ -3,6 +3,27 @@
# v17
## v17.9.0
- OCRmyPDF now uses any Noto font installed on the system, not just the two
dozen script families it knows by name ({issue}`1722`). Previously a document
in, say, Cherokee or Vai was rendered with the glyphless fallback font even
though the matching font was installed — a common situation on macOS, which
ships around a hundred script-specific Noto faces. When the named fonts
cannot cover a word, OCRmyPDF now searches the installed fonts for one that
can.
- The "no installed font has glyphs" warning now names the characters it could
not render, with their codepoints and Unicode names, so it is clear which
font to install. Text that mixes scripts no single font covers is now
reported as such, instead of advising the user to install fonts they may
already have.
- Fixed the macOS font installation instructions, which recommended a Homebrew
package (`font-noto`) that does not exist ({issue}`1722`). Homebrew has no
single Noto package; each family is a separate cask. The Fedora package name
was also corrected to `google-noto-fonts-all`.
- Font providers may now implement the optional `GlyphSearchingFontProvider`
protocol to participate in coverage-based font search.
## v17.8.1
- Improved the `--tesseract-pagesegmode` help text to point to
+2
View File
@@ -18,6 +18,7 @@ from ocrmypdf.font.font_provider import (
BuiltinFontProvider,
ChainedFontProvider,
FontProvider,
GlyphSearchingFontProvider,
)
from ocrmypdf.font.multi_font_manager import MultiFontManager
from ocrmypdf.font.system_font_provider import SystemFontProvider
@@ -25,6 +26,7 @@ from ocrmypdf.font.system_font_provider import SystemFontProvider
__all__ = [
"FontManager",
"FontProvider",
"GlyphSearchingFontProvider",
"BuiltinFontProvider",
"ChainedFontProvider",
"MultiFontManager",
+60 -1
View File
@@ -7,7 +7,7 @@ from __future__ import annotations
import logging
from pathlib import Path
from typing import Protocol
from typing import Protocol, runtime_checkable
from ocrmypdf.font.font_manager import FontManager
@@ -52,6 +52,34 @@ class FontProvider(Protocol):
...
@runtime_checkable
class GlyphSearchingFontProvider(Protocol):
"""Optional capability: find a font by glyph coverage rather than by name.
A provider only knows a limited set of logical font names, but it may have
access to many more fonts than it can name (e.g. the ~100 script-specific
Noto faces macOS installs). Implementing this lets MultiFontManager use
them as a last resort instead of falling back to glyphless rendering.
Providers that do not implement this are used as-is; the capability is
detected at runtime with ``isinstance``.
"""
def find_font_with_glyphs(self, text: str) -> tuple[str, FontManager] | None:
"""Find a font that has glyphs for every character in text.
The returned name must subsequently resolve through ``get_font()``, so
that callers can cache the selection by name.
Args:
text: Text the font must fully cover
Returns:
(logical font name, FontManager), or None if no font covers text
"""
...
class BuiltinFontProvider:
"""Font provider using builtin fonts from ocrmypdf/data directory."""
@@ -119,6 +147,18 @@ class BuiltinFontProvider:
"""Get the glyphless fallback font."""
return self._fonts['Occulta']
def find_font_with_glyphs(self, text: str) -> tuple[str, FontManager] | None:
"""Find a bundled font that covers text, ignoring glyphless Occulta."""
if not text:
return None
codepoints = {ord(c) for c in text}
for name, font in self._fonts.items():
if name == 'Occulta':
continue
if all(font.has_glyph(cp) for cp in codepoints):
return name, font
return None
class ChainedFontProvider:
"""Font provider that tries multiple providers in order.
@@ -170,6 +210,25 @@ class ChainedFontProvider:
result.append(name)
return result
def find_font_with_glyphs(self, text: str) -> tuple[str, FontManager] | None:
"""Ask each capable provider in turn for a font that covers text.
Providers that don't implement the search are skipped.
Args:
text: Text the font must fully cover
Returns:
(logical font name, FontManager) from the first provider with a
match, or None if no provider found one
"""
for provider in self.providers:
if not isinstance(provider, GlyphSearchingFontProvider):
continue
if found := provider.find_font_with_glyphs(text):
return found
return None
def get_fallback_font(self) -> FontManager:
"""Get the glyphless fallback font.
+127 -11
View File
@@ -10,6 +10,7 @@ language hints and glyph coverage analysis.
from __future__ import annotations
import logging
import unicodedata
from pathlib import Path
from ocrmypdf.font.font_manager import FontManager
@@ -17,6 +18,7 @@ from ocrmypdf.font.font_provider import (
BuiltinFontProvider,
ChainedFontProvider,
FontProvider,
GlyphSearchingFontProvider,
)
from ocrmypdf.font.system_font_provider import SystemFontProvider
@@ -33,9 +35,17 @@ class MultiFontManager:
Font selection strategy:
1. Try language-preferred font (if language hint available)
2. Try fallback fonts in order by glyph coverage
3. Fall back to Occulta.ttf (glyphless fallback)
3. Ask the provider for any installed font that covers the text
4. Fall back to Occulta.ttf (glyphless fallback)
"""
# How many uncoverable characters to name in the missing-font warning
MAX_REPORTED_CHARS = 3
# How many characters of a word to look up individually when composing that
# warning; each lookup may scan every font installed on the system
MAX_EXAMINED_CHARS = 8
# Language to font mapping
# Keys are ISO 639-2/3 codes or Tesseract language codes
LANGUAGE_FONT_MAP = {
@@ -173,6 +183,9 @@ class MultiFontManager:
self._selection_cache: dict[tuple[str, str | None], str] = {}
# Track whether we've warned about missing fonts (warn once per script)
self._warned_scripts: set[str] = set()
# Fonts found by glyph coverage rather than by name, tried before
# repeating the (expensive) provider search
self._discovered_fonts: list[str] = []
@property
def fonts(self) -> dict[str, FontManager]:
@@ -208,7 +221,8 @@ class MultiFontManager:
Uses a hybrid approach:
1. Language-based selection (if language hint available)
2. Ordered fallback through available fonts by glyph coverage
3. Final fallback to Occulta.ttf (glyphless)
3. Provider search over every installed font, by glyph coverage
4. Final fallback to Occulta.ttf (glyphless)
Args:
word_text: The text content of the word
@@ -233,19 +247,50 @@ class MultiFontManager:
if result := self._try_font(preferred, word_text, cache_key):
return result
# Phase 2: Try fallback fonts in order
for font_name in self.FALLBACK_FONTS:
# Phase 2: Try fallback fonts in order, then anything a previous
# coverage search turned up
for font_name in [*self.FALLBACK_FONTS, *self._discovered_fonts]:
if font_name in tried_fonts:
continue
tried_fonts.add(font_name)
if result := self._try_font(font_name, word_text, cache_key):
return result
# Phase 3: Glyphless fallback (always succeeds)
# Phase 3: Ask the provider to search every installed font. The named
# families cover common scripts only, but systems ship many more (macOS
# installs ~100 Noto faces), and those should be used before giving up
# on rendering the text at all. See issue #1722.
if found := self._search_font_by_coverage(word_text):
font_name, font = found
self._selection_cache[cache_key] = font_name
return font
# Phase 4: Glyphless fallback (always succeeds)
# Warn if we're falling back for non-ASCII text (likely missing font)
self._warn_missing_font(word_text, line_language)
self._selection_cache[cache_key] = 'Occulta'
return self.font_provider.get_fallback_font()
def _search_font_by_coverage(self, text: str) -> tuple[str, FontManager] | None:
"""Search the provider for any font covering text, if it supports it.
Args:
text: Text the font must fully cover
Returns:
(font name, FontManager), or None if unsupported or nothing matched
"""
provider = self.font_provider
if not isinstance(provider, GlyphSearchingFontProvider):
return None
found = provider.find_font_with_glyphs(text)
if found is None:
return None
font_name, _font = found
if font_name not in self._discovered_fonts:
self._discovered_fonts.append(font_name)
return found
def _warn_missing_font(self, word_text: str, line_language: str | None) -> None:
"""Warn user about missing font for non-Latin text.
@@ -264,26 +309,97 @@ class MultiFontManager:
self._warned_scripts.add(warn_key)
uncoverable = self._uncoverable_characters(word_text)
if not uncoverable:
# Every character has a font, but no single font has them all.
# Telling the user to install fonts would be wrong advice here.
log.warning(
"Text mixing scripts that no single installed font covers (%r) "
"was added as an invisible text layer: it stays searchable and "
"copyable, but appears blank when highlighted in a PDF viewer. "
"Installing more fonts will not help; OCRmyPDF uses one font "
"per word.",
word_text,
)
return
missing = self._describe_characters(uncoverable)
if line_language and line_language in self.LANGUAGE_FONT_MAP:
font_family = self.LANGUAGE_FONT_MAP[line_language].removesuffix('-Regular')
log.warning(
"No installed font has glyphs for the detected '%s' text, so "
"it was added as an invisible text layer: it stays searchable "
"No installed font has glyphs for the detected '%s' text (%s), "
"so it was added as an invisible text layer: it stays searchable "
"and copyable, but appears blank when highlighted in a PDF "
"viewer. Install the %s font family (via your OS package "
"manager or https://fonts.google.com/noto) for full rendering.",
line_language,
missing,
font_family,
)
else:
log.warning(
"No installed font has glyphs for some of the detected text, "
"so it was added as an invisible text layer: it stays "
"No installed font has glyphs for some of the detected text "
"(%s), so it was added as an invisible text layer: it stays "
"searchable and copyable, but appears blank when highlighted "
"in a PDF viewer. Install the matching Noto fonts "
"(https://fonts.google.com/noto) for full rendering."
"in a PDF viewer. Install a Noto font covering that script "
"(https://fonts.google.com/noto) for full rendering.",
missing,
)
def _uncoverable_characters(self, word_text: str) -> list[str]:
"""Find the characters of word_text that no installed font can render.
Args:
word_text: The word that fell back to glyphless rendering
Returns:
The distinct uncoverable characters, in order of first appearance,
considering at most MAX_EXAMINED_CHARS of them
"""
candidates = [
char
for char in dict.fromkeys(word_text) # de-duplicate, keep order
if not char.isspace() and not self._is_char_renderable(char)
]
# The named fonts missed these, but the provider may still have a font
# for them, so confirm before telling the user to install anything. The
# search walks every installed font, hence the cap on how many
# characters we are willing to look up for one warning.
return [
char
for char in candidates[: self.MAX_EXAMINED_CHARS]
if self._search_font_by_coverage(char) is None
]
def _describe_characters(self, chars: list[str]) -> str:
"""Describe characters by codepoint and Unicode name.
Naming the codepoints tells the user which font to install even for
scripts OCRmyPDF has no language mapping for, which the generic
"install the matching Noto fonts" advice did not. See issue #1722.
Args:
chars: Characters to describe
Returns:
Human-readable description, truncated to MAX_REPORTED_CHARS
"""
described = ", ".join(
f"{char!r} U+{ord(char):04X} {unicodedata.name(char, 'unnamed character')}"
for char in chars[: self.MAX_REPORTED_CHARS]
)
if len(chars) > self.MAX_REPORTED_CHARS:
described += f", and {len(chars) - self.MAX_REPORTED_CHARS} more"
return described
def _is_char_renderable(self, char: str) -> bool:
"""Check whether any font already known to us has a glyph for char."""
for font_name in [*self.FALLBACK_FONTS, *self._discovered_fonts]:
font = self.font_provider.get_font(font_name)
if font is not None and self._has_all_glyphs(font, char):
return True
return False
def _has_all_glyphs(self, font: FontManager, text: str) -> bool:
"""Check if a font has glyphs for all characters in text.
+146
View File
@@ -209,6 +209,13 @@ class SystemFontProvider:
self._not_found: set[str] = set()
# Cached font directories (computed lazily)
self._font_dirs: list[Path] | None = None
# Cached (logical name, path) of every Noto face on the system, in the
# order the coverage search should try them (computed lazily)
self._noto_candidates: list[tuple[str, Path]] | None = None
# Memoized results of find_font_with_glyphs(), keyed by codepoint set
self._coverage_cache: dict[frozenset[int], str | None] = {}
# Font files that failed to load, so we only complain about them once
self._unloadable: set[Path] = set()
def _get_platform(self) -> str:
"""Get the current platform identifier.
@@ -352,6 +359,145 @@ class SystemFontProvider:
return best[1]
return None
@staticmethod
def _family_base(stem: str) -> str | None:
"""Get the Noto family base a filename stem is the Regular face of.
Args:
stem: Filename without extension, e.g. 'NotoSansCherokee-Regular'
Returns:
The family base ('NotoSansCherokee') or None if the stem is not a
Noto font, or is a weight/slope variant such as '-Bold' or
'-Italic' that should not stand in for the family.
"""
head = stem.split('[', 1)[0] # drop variable-font axes, e.g. '[wght]'
if head.endswith('-Regular'):
head = head[: -len('-Regular')]
elif head.endswith('-VF'):
head = head[: -len('-VF')]
elif '-' in head:
return None
return head if head.startswith('Noto') else None
@classmethod
def _candidate_sort_key(cls, base: str) -> tuple[int, int, str]:
"""Rank a family base for the coverage search.
Sans comes before serif before everything else, and plain families come
ahead of their narrower UI and Mono cousins.
"""
if base.startswith('NotoSans'):
family_rank = 0
elif base.startswith('NotoSerif'):
family_rank = 1
else:
family_rank = 2
narrow_use = base.endswith('UI') or base.startswith('NotoSansMono')
return (family_rank, int(narrow_use), base)
def _get_noto_candidates(self) -> list[tuple[str, Path]]:
"""Enumerate every Noto family installed on the system.
Scans each font directory once and keeps the best-ranked file per
family, so a family present in several directories or in several
variants contributes a single candidate.
Returns:
List of (logical font name, path) in the order to try them.
"""
if self._noto_candidates is not None:
return self._noto_candidates
best: dict[str, tuple[int, Path]] = {}
for font_dir in self._get_font_dirs():
if not font_dir.exists():
continue
try:
paths = sorted(font_dir.rglob('Noto*'))
except OSError:
# Skip directories we can't read
continue
for path in paths:
if path.suffix.lower() not in self._FONT_EXTENSIONS:
continue
base = self._family_base(path.stem)
if base is None:
continue
kind = self._classify_variant(path.stem, base)
if kind is None:
continue
rank = self._VARIANT_RANK[kind]
if base not in best or rank < best[base][0]:
best[base] = (rank, path)
self._noto_candidates = [
(f'{base}-Regular', path)
for base, (_rank, path) in sorted(
best.items(), key=lambda item: self._candidate_sort_key(item[0])
)
]
return self._noto_candidates
def find_font_with_glyphs(self, text: str) -> tuple[str, FontManager] | None:
"""Find any installed Noto font that covers every character in text.
``NOTO_FONT_PATTERNS`` enumerates the couple dozen scripts OCRmyPDF
knows by name, but systems ship far more: macOS alone installs around a
hundred script-specific Noto faces in
``/System/Library/Fonts/Supplemental``. This is the last resort that
makes those usable, so a document is only rendered glyphless when no
installed font can actually cover it. See issue #1722.
This walks every Noto face on the system and is therefore expensive;
results are memoized, and callers should only reach it after the named
fonts have failed.
Args:
text: Text that the returned font must fully cover
Returns:
(logical font name, FontManager) of the first covering font, or
None if nothing installed covers the text.
"""
if not text:
return None
needed = frozenset(ord(c) for c in text)
if needed in self._coverage_cache:
cached_name = self._coverage_cache[needed]
if cached_name is None:
return None
if cached := self._font_cache.get(cached_name):
return cached_name, cached
for font_name, path in self._get_noto_candidates():
font = self._font_cache.get(font_name)
if font is None:
if path in self._unloadable:
continue
try:
font = FontManager(path)
except Exception as e:
log.debug("Skipping unreadable font %s: %s", path, e)
self._unloadable.add(path)
continue
if all(font.has_glyph(cp) for cp in needed):
# Keep only fonts we actually use; the rest are released so a
# full scan doesn't retain every font file on the system.
self._font_cache[font_name] = font
self._not_found.discard(font_name)
self._coverage_cache[needed] = font_name
log.debug(
"Found system font %s at %s (glyph coverage match)",
font_name,
path,
)
return font_name, font
self._coverage_cache[needed] = None
return None
def get_font(self, font_name: str) -> FontManager | None:
"""Get a FontManager for the named font (lazy loading).
+114
View File
@@ -569,3 +569,117 @@ def test_missing_font_warning_explains_consequences(font_dir, caplog):
# the text stays searchable but renders blank when highlighted.
assert 'searchable' in msg.lower()
assert 'highlight' in msg.lower() or 'select' in msg.lower()
# --- Coverage-driven last-resort font search (#1722) ---
class _FakeProviderWithSearch(_FakeFontProvider):
"""FontProvider that can also search unlisted fonts by glyph coverage."""
def __init__(self, fonts, searchable):
super().__init__(fonts)
self._searchable = searchable
self.search_calls: list[str] = []
def find_font_with_glyphs(self, text):
self.search_calls.append(text)
for name, font in self._searchable.items():
hb = font.get_hb_font()
if all(hb.get_nominal_glyph(ord(c)) for c in text):
self._fonts[name] = font # discovered fonts become resolvable
return name, font
return None
def test_unlisted_font_found_by_coverage_search():
"""A script outside FALLBACK_FONTS is rendered if the font is installed."""
searchable = {'NotoSansCherokee-Regular': _FakeFontManager('Cherokee.ttf', 'ᏣᎳᎩ')}
provider = _FakeProviderWithSearch({}, searchable)
manager = MultiFontManager(font_provider=provider)
font = manager.select_font_for_word('ᏣᎳᎩ', None)
assert font.font_path.name == 'Cherokee.ttf'
assert provider.search_calls == ['ᏣᎳᎩ']
def test_coverage_search_result_is_cached():
"""The expensive coverage search runs once per distinct word."""
searchable = {'NotoSansCherokee-Regular': _FakeFontManager('Cherokee.ttf', 'ᏣᎳᎩ')}
provider = _FakeProviderWithSearch({}, searchable)
manager = MultiFontManager(font_provider=provider)
manager.select_font_for_word('ᏣᎳᎩ', None)
manager.select_font_for_word('ᏣᎳᎩ', None)
assert len(provider.search_calls) == 1
def test_named_fonts_take_precedence_over_coverage_search():
"""The coverage search is a last resort, not a substitute for named fonts."""
fonts = {'NotoSans-Regular': _FakeFontManager('NotoSans.ttf', 'abc')}
searchable = {'NotoSansMono-Regular': _FakeFontManager('NotoSansMono.ttf', 'abc')}
provider = _FakeProviderWithSearch(fonts, searchable)
manager = MultiFontManager(font_provider=provider)
font = manager.select_font_for_word('abc', None)
assert font.font_path.name == 'NotoSans.ttf'
assert provider.search_calls == []
def test_provider_without_coverage_search_still_falls_back():
"""Providers predating find_font_with_glyphs() keep working (duck-typed)."""
manager = MultiFontManager(font_provider=_FakeFontProvider({}))
font = manager.select_font_for_word('ᏣᎳᎩ', None)
assert font.font_path.name == 'Occulta.ttf'
def test_missing_font_warning_names_the_missing_characters(font_dir, caplog):
"""The warning must identify what could not be rendered (#1722).
The user's real question is "which font package do I install?" — naming the
offending codepoints and their Unicode names answers it even for scripts
OCRmyPDF has no language mapping for.
"""
manager = MultiFontManager(font_provider=BuiltinFontProvider(font_dir))
with caplog.at_level(logging.WARNING):
manager.select_font_for_word('ᏣᎳᎩ', None)
msg = caplog.text
assert 'U+13E3' in msg # CHEROKEE LETTER TSA
assert 'CHEROKEE' in msg.upper()
def test_missing_character_warning_ignores_covered_characters(font_dir, caplog):
"""Only the uncoverable characters are reported, not the whole word."""
manager = MultiFontManager(font_provider=BuiltinFontProvider(font_dir))
with caplog.at_level(logging.WARNING):
manager.select_font_for_word('aᏣb', None)
msg = caplog.text
assert 'U+13E3' in msg
assert 'U+0061' not in msg # 'a' is covered by the builtin Latin font
def test_mixed_script_word_warning_does_not_advise_installing_fonts(caplog):
"""A word no single font covers is reported as such, not as a missing font.
Every character here has an installed font; telling the user to install
more would be wrong advice and is exactly the confusion behind #1722.
"""
fonts = {'NotoSans-Regular': _FakeFontManager('NotoSans.ttf', 'ab')}
searchable = {'NotoSansCherokee-Regular': _FakeFontManager('Cherokee.ttf', '')}
manager = MultiFontManager(font_provider=_FakeProviderWithSearch(fonts, searchable))
with caplog.at_level(logging.WARNING):
font = manager.select_font_for_word('aᏣb', None)
assert font.font_path.name == 'Occulta.ttf'
msg = caplog.text
assert 'mixing scripts' in msg
assert 'will not help' in msg
assert 'U+' not in msg # no codepoints to blame; nothing to install
+180
View File
@@ -363,6 +363,140 @@ class TestSystemFontProviderVariableFonts:
assert provider.get_font('NotoSansSC-Regular') is None
class TestSystemFontProviderUnlistedFamilies:
"""Test the coverage-driven search over Noto families we don't enumerate.
``NOTO_FONT_PATTERNS`` names only the couple dozen most common scripts, but
macOS ships ~100 script-specific Noto fonts and Homebrew/Linux distros offer
even more. Those fonts must still be usable when the enumerated families
cannot cover the text. See issue #1722.
"""
@pytest.fixture
def real_font_bytes(self):
"""Bytes of a real, loadable font covering ASCII."""
font_path = (
Path(__file__).parent.parent
/ "src"
/ "ocrmypdf"
/ "data"
/ "NotoSans-Regular.ttf"
)
if not font_path.exists():
pytest.skip("Builtin font not available")
return font_path.read_bytes()
def _provider_for(self, tmp_path, filenames, real_font_bytes):
"""Build a provider whose only font dir is tmp_path with given files."""
for name in filenames:
(tmp_path / name).write_bytes(real_font_bytes)
provider = SystemFontProvider()
provider._font_dirs = [tmp_path]
return provider
def test_finds_unlisted_family_by_coverage(self, tmp_path, real_font_bytes):
"""A Noto family absent from NOTO_FONT_PATTERNS is still usable."""
provider = self._provider_for(
tmp_path, ['NotoSansCherokee-Regular.ttf'], real_font_bytes
)
assert 'NotoSansCherokee-Regular' not in provider.NOTO_FONT_PATTERNS
found = provider.find_font_with_glyphs('A')
assert found is not None
name, font = found
assert name == 'NotoSansCherokee-Regular'
assert font.font_path.name == 'NotoSansCherokee-Regular.ttf'
def test_finds_unlisted_variable_family(self, tmp_path, real_font_bytes):
"""Bracketed variable filenames are eligible for the coverage search."""
provider = self._provider_for(
tmp_path, ['NotoSansVithkuqi[wght].ttf'], real_font_bytes
)
found = provider.find_font_with_glyphs('A')
assert found is not None
assert found[0] == 'NotoSansVithkuqi-Regular'
def test_discovered_font_resolves_by_logical_name(self, tmp_path, real_font_bytes):
"""A font found by coverage is afterwards reachable via get_font()."""
provider = self._provider_for(
tmp_path, ['NotoSansCherokee-Regular.ttf'], real_font_bytes
)
name, font = provider.find_font_with_glyphs('A')
assert provider.get_font(name) is font
def test_negative_cache_does_not_block_discovery(self, tmp_path, real_font_bytes):
"""A prior failed get_font() must not hide a later coverage match."""
provider = self._provider_for(
tmp_path, ['NotoSansCherokee-Regular.ttf'], real_font_bytes
)
assert provider.get_font('NotoSansCherokee-Regular') is None # not listed
name, font = provider.find_font_with_glyphs('A')
assert provider.get_font(name) is font
def test_skips_bold_and_italic_styles(self, tmp_path, real_font_bytes):
"""Only Regular/variable faces are candidates, never Bold or Italic."""
provider = self._provider_for(
tmp_path,
[
'NotoSansCherokee-Bold.ttf',
'NotoSansCherokee-Italic.ttf',
'NotoSans-Italic[wdth,wght].ttf',
],
real_font_bytes,
)
assert provider.find_font_with_glyphs('A') is None
def test_ignores_non_noto_fonts(self, tmp_path, real_font_bytes):
"""Non-Noto system fonts are not enlisted by the coverage search."""
provider = self._provider_for(
tmp_path, ['DejaVuSans.ttf', 'Arial.ttf'], real_font_bytes
)
assert provider.find_font_with_glyphs('A') is None
def test_returns_none_when_no_font_covers_text(self, tmp_path, real_font_bytes):
"""Text no installed font covers yields no match rather than a wrong one."""
provider = self._provider_for(
tmp_path, ['NotoSansCherokee-Regular.ttf'], real_font_bytes
)
# U+13A3 CHEROKEE LETTER O is absent from the Latin font's cmap.
assert provider.find_font_with_glyphs('') is None
def test_empty_text_does_not_match(self, tmp_path, real_font_bytes):
"""Empty text has nothing to cover, so no font is claimed for it."""
provider = self._provider_for(
tmp_path, ['NotoSansCherokee-Regular.ttf'], real_font_bytes
)
assert provider.find_font_with_glyphs('') is None
def test_unloadable_font_file_is_skipped(self, tmp_path, real_font_bytes):
"""A corrupt font file does not abort the search for a usable one."""
(tmp_path / 'NotoSansBroken-Regular.ttf').write_bytes(b'not a font')
provider = self._provider_for(
tmp_path, ['NotoSansCherokee-Regular.ttf'], real_font_bytes
)
found = provider.find_font_with_glyphs('A')
assert found is not None
assert found[0] == 'NotoSansCherokee-Regular'
@pytest.mark.parametrize(
'stem,expected',
[
('NotoSansCherokee-Regular', 'NotoSansCherokee'),
('NotoSansCherokee[wght]', 'NotoSansCherokee'),
('NotoSansArabic[wdth,wght]', 'NotoSansArabic'),
('NotoSansCJKsc-VF', 'NotoSansCJKsc'),
('NotoMusic', 'NotoMusic'),
('NotoSansCJK-Regular', 'NotoSansCJK'),
('NotoSans-Bold', None),
('NotoSans-Italic[wdth,wght]', None),
('NotoSans-SemiCondensedBlackItalic', None),
('DejaVuSans-Regular', None),
],
)
def test_family_base_parsing(self, stem, expected):
"""Filename stems map to family bases, rejecting non-Regular styles."""
assert SystemFontProvider._family_base(stem) == expected
# --- ChainedFontProvider Tests ---
@@ -507,3 +641,49 @@ class TestChainedFontProviderIntegration:
# Chain should have at least as many fonts as builtin
assert chain_fonts >= builtin_fonts
class TestChainedFontProviderCoverageSearch:
"""Test that the chain delegates the coverage search to its members."""
class _Searchable:
"""Provider stub that reports one findable font."""
def __init__(self, result):
self.result = result
self.calls = 0
def get_font(self, name):
return None
def get_available_fonts(self):
return []
def get_fallback_font(self):
raise NotImplementedError
def find_font_with_glyphs(self, text):
self.calls += 1
return self.result
def test_delegates_to_first_provider_that_finds_a_font(self):
"""The first provider with a match wins; later ones are not consulted."""
first = self._Searchable(('NotoSansX-Regular', MagicMock()))
second = self._Searchable(('NotoSansY-Regular', MagicMock()))
chain = ChainedFontProvider([first, second])
assert chain.find_font_with_glyphs('x')[0] == 'NotoSansX-Regular'
assert second.calls == 0
def test_skips_providers_without_the_capability(self):
"""Providers lacking find_font_with_glyphs() are skipped, not fatal."""
legacy = MagicMock(spec=['get_font', 'get_available_fonts'])
searchable = self._Searchable(('NotoSansX-Regular', MagicMock()))
chain = ChainedFontProvider([legacy, searchable])
assert chain.find_font_with_glyphs('x')[0] == 'NotoSansX-Regular'
def test_returns_none_when_nothing_matches(self):
"""No provider matching yields None so the caller can use Occulta."""
chain = ChainedFontProvider([self._Searchable(None)])
assert chain.find_font_with_glyphs('x') is None