BUG: Map suffixed and ligature glyph names in font encodings by michaelkillgta · Pull Request #4310 · py-pdf/pypdf

michaelkillgta · 2026-10-29T21:16:37Z

Closes #4309.

_glyph_name_to_unicode now follows the Adobe Glyph List mapping rules.

Specification references

ISO 32000-2:2020, 9.10.2 "Mapping character codes to Unicode values" (same section number in ISO 32000-1:2008): for a simple font whose /Differences array uses standard character names, map the code to a character name and then "look up the character name in the Adobe Glyph List" to get the Unicode value.
Adobe Glyph List Specification, version 2.9, section 2 "The mapping":
1. Drop all characters from the first period onwards (a.sc -> a, one.oldstyle -> 1).
2. Split the rest into components at underscores (f_i -> f, i).
3. Map each component and concatenate the results. A component is a glyph list name, or uni followed by groups of four hex digits, each in 0000-D7FF or E000-FFFF (uni00660069 -> fi), or u followed by four to six hex digits, in 0000-D7FF or E000-10FFFF. This is why /u1 no longer turns into "\x01" and surrogates are rejected.
Adobe Glyph List Specification, section 3 "Examples": Lcommaaccent_uni20AC0308_u1040C.alternate and uniD801DC0C are used as test cases.

Deviations from the specification

Each of these keeps pypdf's current behavior:

An exact match in pypdf's glyph table is checked first, before and after dropping the suffix. The table also lists some names that contain a period, for example /two.superior, and f_f_i.alt still maps to ﬃ.
Lowercase hex digits are accepted. The specification only allows uppercase, but main already decodes names like /uni00e9.
The specification maps a component that cannot be mapped to an empty string. Here the whole name returns None instead, so callers keep the raw glyph name as before (/f_unknownglyph, /.sc).

The same helper is used for CFF and Type 1 font program encodings, so those get the same handling.

Tests

New parametrized test in tests/test_cmap.py. It fails on main and passes here.
Every new line and branch in _glyph_name_to_unicode and _glyph_component_code_points is covered: pytest tests/test_cmap.py --cov=pypdf._cmap --cov-branch.
The full test suite, ruff check pypdf tests and mypy pypdf pass locally.

Signed-off-by: michaelkillgta <269775689+michaelkillgta@users.noreply.github.com>

codecov · 2026-10-29T21:23:39Z

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 98.63%. Comparing base (819f5ed) to head (78d7de0).
⚠️ Report is 8 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #4310      +/-   ##
==========================================
+ Coverage   98.55%   98.63%   +0.08%     
==========================================
  Files          59       59              
  Lines       11992    11976      -16     
  Branches     2268     2275       +7     
==========================================
- Hits        11819    11813       -6     
+ Misses         93       86       -7     
+ Partials       80       77       -3

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

stefan6419846

Please check the coverage and provide exact references to the official specification.

Signed-off-by: michaelkillgta <269775689+michaelkillgta@users.noreply.github.com>

stefan6419846 · 2026-10-30T21:00:58Z

-            code_point = glyph_name[2:]
-        else:
+    # Adobe Glyph List Specification, section 2, step 1.
+    base_name = glyph_name.lstrip("/").split(".", 1)[0]


Suggested change

base_name = glyph_name.lstrip("/").split(".", 1)[0]

base_name = glyph_name.removeprefix("/").split(".", maxsplit=1)[0]

stefan6419846 · 2026-10-30T21:02:10Z

+    base_name = glyph_name.lstrip("/").split(".", 1)[0]
+    if not base_name:
+        return None
+    if f"/{base_name}" in adobe_glyphs:


Consider using the Walrus operator with a dedicated variable to avoid repeated conversion.

stefan6419846 · 2026-10-30T21:02:49Z

+        return None
+    if f"/{base_name}" in adobe_glyphs:
+        return adobe_glyphs[f"/{base_name}"]
+    result = ""


Prefer a list and "".join over repeated string concatenation.

stefan6419846 · 2026-10-30T21:03:41Z

+    See the Adobe Glyph List Specification, version 2.9, section 2, step 3.
+    Surrogates (D800-DFFF) and values above 10FFFF are rejected.
+    """
+    if component.startswith("uni") and len(component) > 3 and len(component) % 4 == 3:


You are calculating len(component) quite often. Consider using a dedicated variable.

stefan6419846 · 2026-10-30T21:03:55Z

+    """
+    Map a ``uniXXXX[XXXX...]`` or ``uXXXX[XX]`` glyph name component.
+
+    See the Adobe Glyph List Specification, version 2.9, section 2, step 3.


Do you have a link to the specification?

stefan6419846 · 2026-10-30T21:04:37Z

+        code_points = [component[1:]]
+    else:
+        return None
+    result = ""


See above for concatenating.

stefan6419846 · 2026-10-30T21:05:00Z

+        return None
+    result = ""
+    for code_point in code_points:
+        if not all(c in "0123456789ABCDEFabcdef" for c in code_point):


Suggested change

if not all(c in "0123456789ABCDEFabcdef" for c in code_point):

if not all(c in string.hexdigits for c in code_point):

stefan6419846 · 2026-10-30T21:05:37Z

+        if not all(c in "0123456789ABCDEFabcdef" for c in code_point):
+            return None
+        value = int(code_point, 16)
+        if 0xD800 <= value <= 0xDFFF or value > 0x10FFFF:


Please define these range limits as named constants for better readability/documentation.

stefan6419846 · 2026-10-30T21:05:58Z

+    return result
+
+
+def _glyph_component_code_points(component: str) -> str | None:


Function names should start with an action.

stefan6419846 · 2026-10-30T21:07:27Z


+@pytest.mark.parametrize(
+    ("glyph_name", "expected"),
+    [


Are you able to provide meaningful parameter IDs to document what each line is trying to ensure?

BUG: Map suffixed and ligature glyph names in font encodings

ed3f03d

Signed-off-by: michaelkillgta <269775689+michaelkillgta@users.noreply.github.com>

stefan6419846 requested changes Oct 30, 2026

View reviewed changes

thomamartin mentioned this pull request Oct 30, 2026

Ligature issue when converting PDF to text #1351

Open

TST: Cover all glyph name paths and cite the specifications

78d7de0

Signed-off-by: michaelkillgta <269775689+michaelkillgta@users.noreply.github.com>

stefan6419846 requested changes Oct 30, 2026

View reviewed changes

Navigation Menu

Repository navigation

BUG: Map suffixed and ligature glyph names in font encodings - #4310

BUG: Map suffixed and ligature glyph names in font encodings#4310
michaelkillgta wants to merge 2 commits into
py-pdf:mainfrom
michaelkillgta:fix/agl-glyph-name-components

michaelkillgta commented Oct 29, 2026 •

edited

Loading

Uh oh!

codecov Bot commented Oct 29, 2026 •

edited

Loading

Uh oh!

stefan6419846 left a comment

Uh oh!

stefan6419846 Oct 30, 2026

Uh oh!

stefan6419846 Oct 30, 2026

Uh oh!

stefan6419846 Oct 30, 2026

Uh oh!

stefan6419846 Oct 30, 2026

Uh oh!

stefan6419846 Oct 30, 2026

Uh oh!

stefan6419846 Oct 30, 2026

Uh oh!

stefan6419846 Oct 30, 2026

Uh oh!

stefan6419846 Oct 30, 2026

Uh oh!

stefan6419846 Oct 30, 2026

Uh oh!

stefan6419846 Oct 30, 2026

Uh oh!

This branch has not been deployed

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

	base_name = glyph_name.lstrip("/").split(".", 1)[0]
	base_name = glyph_name.removeprefix("/").split(".", maxsplit=1)[0]

	if not all(c in "0123456789ABCDEFabcdef" for c in code_point):
	if not all(c in string.hexdigits for c in code_point):

		return result


		def _glyph_component_code_points(component: str) -> str \| None:

Conversation

michaelkillgta commented Oct 29, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Specification references

Deviations from the specification

Tests

Uh oh!

codecov Bot commented Oct 29, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

stefan6419846 left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

This branch has not been deployed

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

michaelkillgta commented Oct 29, 2026 •

edited

Loading

codecov Bot commented Oct 29, 2026 •

edited

Loading