Improve fallback font export tests

Working on matplotlib#28765, I realized that we don't need an external font to
test font fallback, as `cmr10`, and related, fonts have fairly low
coverage compared to `DejaVu Sans`. Thus font fallback can easily be
tested by starting with `cmr10`, and pulling additional characters from
`DejaVu Sans`.

In addition, this change increases unique characters in the figure to
587, which triggers some code paths for tables that are limited to 256
characters.

Author: QuLogic

lib/matplotlib/testing/__init__.py

@@ -1,13 +1,15 @@
 """
 Helper functions for testing.
 """
-from pathlib import Path
-from tempfile import TemporaryDirectory
+import itertools
 import locale
 import logging
 import os
+from pathlib import Path
+import string
 import subprocess
 import sys
+from tempfile import TemporaryDirectory
 
 import matplotlib as mpl
 from matplotlib import _api
@@ -232,3 +234,45 @@ def is_ci_environment():
             return True
 
     return False
+
+
+def _gen_multi_font_text():
+    """
+    Generate text intended for use with multiple fonts to exercise font fallbacks.
+
+    Returns
+    -------
+    fonts : list of str
+        The names of the fonts used to render the test string, sorted by intended
+        priority. This should be set as the font family for the Figure or Text artist.
+    text : str
+        The test string.
+    """
+    # These fonts are serif and sans-serif, and would not normally be combined, but that
+    # should make it easier to see which glyph is from which font.
+    fonts = ['cmr10', 'DejaVu Sans']
+    # cmr10 does not contain accented characters, so they should fall back to DejaVu
+    # Sans. However, some accented capital A versions *are* in cmr10 with non-standard
+    # glyph shapes, so don't test those (otherwise this Latin1 supplement group would
+    # start at 0xA0.)
+    start = 0xC5
+    latin1_supplement = [chr(x) for x in range(start, 0xFF+1)]
+    latin_extended_A = [chr(x) for x in range(0x100, 0x17F+1)]
+    latin_extended_B = [chr(x) for x in range(0x180, 0x24F+1)]
+    ipa_extensions = [chr(x) for x in range(0x250, 0x2AF+1)]
+    count = itertools.count(start - 0xA0)
+    non_basic_characters = '\n'.join(
+        ''.join(line)
+        for _, line in itertools.groupby(  # Replace with itertools.batched for Py3.12+.
+            [*latin1_supplement, *latin_extended_A, *latin_extended_B, *ipa_extensions],
+            key=lambda x: next(count) // 32)  # 32 characters per line.
+    )
+    test_str = f"""There are basic characters
+{string.ascii_uppercase} {string.ascii_lowercase}
+{string.digits} {string.punctuation}
+and accented characters
+{non_basic_characters}
+in between!"""
+    # The resulting string contains 587 unique characters. Some file formats use 8-bit
+    # tables, which the large number of characters exercises twice over.
+    return fonts, test_str

lib/matplotlib/testing/__init__.pyi

@@ -52,3 +52,4 @@ def ipython_in_subprocess(
     all_expected_backends: dict[tuple[int, int], str],
 ) -> None: ...
 def is_ci_environment() -> bool: ...
+def _gen_multi_font_text() -> tuple[list[str], str]: ...