Merge pull request #883 from googlefonts/improve-html-formatter

Improve HTML formatter & activate it in new packager

Lib/gftools/packager.py

@@ -23,7 +23,7 @@
 from gftools.scripts.add_font import main as add_font
 from gftools.tags import GFTags
 from gftools.util import google_fonts as fonts
-from gftools.utils import download_file, is_google_fonts_repo
+from gftools.utils import download_file, is_google_fonts_repo, format_html
 
 log = logging.getLogger("gftools.packager")
 LOG_FORMAT = "%(message)s"
@@ -307,6 +307,12 @@ def package_family(
                 os.remove(fp)
         shutil.copytree(tmp_dir, family_path, dirs_exist_ok=True)
         save_metadata(family_path / "METADATA.pb", metadata)
+        # Format HTML
+        desc_file = family_path / "DESCRIPTION.en_us.html"
+        with open(desc_file, encoding="utf-8") as fin:
+            description = format_html(fin.read())
+        with open(desc_file, "w", encoding="utf-8") as fout:
+            fout.write(description)
     return True
 
 

Lib/gftools/utils.py

@@ -278,6 +278,33 @@ def _html_custom_formatter(string):
     string = string.replace(". ", f".\n{whitespace}")
     string = string.replace("! ", f"!\n{whitespace}")
     string = string.replace("? ", f"?\n{whitespace}")
+    # Read into list
+    strings = string.split("\n")
+    # Cycle through list to find abbreviations
+    for i in range(1, len(strings)):
+        this_line = strings[i-1]
+        next_line = strings[i]
+        if this_line == "":
+            continue
+        if (
+            re.search(r"i\.?e\.$", this_line)  # ie.
+            or re.search(r"e\.?g\.$", this_line)  # eg.
+            or (
+                re.search(r"etc[\.|\?|!]$", this_line)
+                and next_line[1] == next_line[1].lower()
+            )  # etc.
+            or (
+                re.search(r"\W\w{1,2}[\.|\?|!]$", this_line)
+                and this_line[-2] == this_line[-2].upper()
+                and next_line[1] == next_line[1].lower()
+            )  # H.R. Giger
+        ):
+            strings[i-1] = strings[i-1] + strings[i]
+            strings[i] = ""
+    # Join back together
+    string = "\n".join(strings)
+    # Remove double lines
+    string = string.replace("\n\n", "\n")
     return string
 
 

tests/test_utils.py

@@ -16,3 +16,43 @@ def test_remove_url_prefix(url, want):
     from gftools.utils import remove_url_prefix
     got = remove_url_prefix(url)
     assert got == want
+
+
+def test_format_html():
+    from gftools.utils import format_html
+
+    input = """<p>
+First sentence. Second sentence.
+Sentence that uses an abbreviation, e.g. "for example".    Sentence that uses an abbreviation, eg. "for example".
+Sentence that uses another abbreviation, i.e. "for example".    Sentence that uses another abbreviation, ie. "for example".
+Sentence that ends in etc. Another sentence after it.
+Sentence that uses etc. but then doesn't end.
+The characters of the film were designed by H.R. Giger. His alien characters became iconic throughout pop culture.
+The characters of the film were designed by H.R. Giger, a Swiss sculptural artist. His alien characters became iconic throughout pop culture.
+He was referred to H.R. Giger, who headed the H.R. department at the time, then told them they're fired. <-- Can't have it both ways. Legitimate abbreviations at the end of sentences can only be caught if they are known in advance, e.g. etc.
+</p>
+"""
+
+    output = """<p>
+ First sentence.
+ Second sentence.
+ Sentence that uses an abbreviation, e.g. "for example".
+ Sentence that uses an abbreviation, eg. "for example".
+ Sentence that uses another abbreviation, i.e. "for example".
+ Sentence that uses another abbreviation, ie. "for example".
+ Sentence that ends in etc.
+ Another sentence after it.
+ Sentence that uses etc. but then doesn't end.
+ The characters of the film were designed by H.R.
+ Giger.
+ His alien characters became iconic throughout pop culture.
+ The characters of the film were designed by H.R.
+ Giger, a Swiss sculptural artist.
+ His alien characters became iconic throughout pop culture.
+ He was referred to H.R.
+ Giger, who headed the H.R. department at the time, then told them they're fired.
+ <-- Can't have it both ways.
+ Legitimate abbreviations at the end of sentences can only be caught if they are known in advance, e.g. etc.
+</p>
+"""
+    assert format_html(input) == output