Improve font management

This commit adds many small improvements in the way fonts are managed.

1. Most of the font deduplication logic has been put in add_font(), instead of
draw_first_line().

2. Font objects now have only one (reproducible) hash, based on the font
content and the face index. Before this commit, we had 3 (!) different hashes.
Fixes #1553.

3. Font fields are calculated during initialization, to avoid useless
parameters storage.

4. Harfbuzz face is not read twice from Pango font when a new Font is added.

5. Font face deduplication is now done using the Pango face pointer, instead of
the Harfbuzz face hash extracted from the Pango font. This could save some
time, especially for very long documents always using the same font face.

6. Font name displayed in PDF now includes weight and style.

weasyprint/document.py

@@ -63,29 +63,39 @@ def _w3c_date_to_pdf(string, attr_name):
 
 
 class Font:
-    def __init__(self, font_hash, file_content, pango_font, index):
-        pango_metrics = pango.pango_font_get_metrics(pango_font, ffi.NULL)
+    def __init__(self, pango_font):
         hb_font = pango.pango_font_get_hb_font(pango_font)
         hb_face = harfbuzz.hb_font_get_face(hb_font)
-        self._font_description = pango.pango_font_describe(pango_font)
-        self.family = ffi.string(pango.pango_font_description_get_family(
-            self._font_description))
-        font_size = pango.pango_font_description_get_size(
-            self._font_description)
+        self.index = harfbuzz.hb_face_get_index(hb_face)
+        hb_blob = ffi.gc(
+            harfbuzz.hb_face_reference_blob(hb_face),
+            harfbuzz.hb_blob_destroy)
+        with ffi.new('unsigned int *') as length:
+            hb_data = harfbuzz.hb_blob_get_data(hb_blob, length)
+            self.file_content = ffi.unpack(hb_data, int(length[0]))
+
+        pango_metrics = pango.pango_font_get_metrics(pango_font, ffi.NULL)
+        description = pango.pango_font_describe(pango_font)
+        font_size = pango.pango_font_description_get_size(description)
+        self.style = pango.pango_font_description_get_style(description)
+        self.family = ffi.string(
+            pango.pango_font_description_get_family(description))
+        digest = hashlib.sha1(self.file_content + bytes(self.index)).digest()
+        self.hash = ''.join(chr(65 + letter % 26) for letter in digest[:6])
+
+        # Name
         description_string = ffi.string(
-            pango.pango_font_description_to_string(self._font_description))
-        sha = hashlib.sha256()
-        sha.update(str(font_hash).encode())
-        sha.update(description_string)
-
-        self.file_content = file_content
-        self.index = index
-        self.file_hash = hash(file_content + bytes(index))
-        self.hash = ''.join(
-            chr(65 + letter % 26) for letter in sha.digest()[:6])
-        self.name = (
-            b'/' + self.hash.encode() + b'+' + self.family.replace(b' ', b''))
-        self.italic_angle = 0  # TODO: this should be different
+            pango.pango_font_description_to_string(description))
+        fields = description_string.split(b' ')
+        if fields and b'=' in fields[-1]:
+            fields.pop()  # Remove variations
+        if fields:
+            fields.pop()  # Remove font size
+        else:
+            fields = [b'Unknown']
+        self.name = b'/' + self.hash.encode() + b'+' + b''.join(fields)
+
+        # Ascent & descent
         if font_size:
             self.ascent = int(
                 pango.pango_font_metrics_get_ascent(pango_metrics) /
@@ -95,6 +105,9 @@ def __init__(self, font_hash, file_content, pango_font, index):
                 font_size * 1000)
         else:
             self.ascent = self.descent = 0
+
+        # Various properties
+        self.italic_angle = 0  # TODO: this should be different
         self.upem = harfbuzz.hb_face_get_upem(hb_face)
         self.png = harfbuzz.hb_ot_color_has_png(hb_face)
         self.svg = harfbuzz.hb_ot_color_has_svg(hb_face)
@@ -104,17 +117,15 @@ def __init__(self, font_hash, file_content, pango_font, index):
         self.widths = {}
         self.cmap = {}
 
-    @property
-    def flags(self):
-        flags = 2 ** (3 - 1)  # Symbolic, custom character set
-        if pango.pango_font_description_get_style(self._font_description):
-            flags += 2 ** (7 - 1)  # Italic
+        # Font flags
+        self.flags = 2 ** (3 - 1)  # Symbolic, custom character set
+        if self.style:
+            self.flags += 2 ** (7 - 1)  # Italic
         if b'Serif' in self.family.split():
-            flags += 2 ** (2 - 1)  # Serif
+            self.flags += 2 ** (2 - 1)  # Serif
         widths = self.widths.values()
         if len(widths) > 1 and len(set(widths)) == 1:
-            flags += 2 ** (1 - 1)  # FixedPitch
-        return flags
+            self.flags += 2 ** (1 - 1)  # FixedPitch
 
 
 class Stream(pydyf.Stream):
@@ -217,13 +228,11 @@ def set_alpha(self, alpha, stroke=False, fill=None):
                     self._states[key] = pydyf.Dictionary({'ca': alpha})
                 super().set_state(key)
 
-    def add_font(self, font_hash, font_content, pango_font, index):
-        self._document.fonts[font_hash] = Font(
-            font_hash, font_content, pango_font, index)
-        return self._document.fonts[font_hash]
-
-    def get_fonts(self):
-        return self._document.fonts
+    def add_font(self, pango_font):
+        face = pango.pango_font_get_face(pango_font)
+        if face not in self._document.fonts:
+            self._document.fonts[face] = Font(pango_font)
+        return self._document.fonts[face]
 
     def add_group(self, bounding_box):
         states = pydyf.Dictionary()
@@ -1323,10 +1332,7 @@ def write_pdf(self, target=None, zoom=1, attachments=None, finisher=None):
         pdf_fonts = pydyf.Dictionary()
         fonts_by_file_hash = {}
         for font in self.fonts.values():
-            if font.file_hash in fonts_by_file_hash:
-                fonts_by_file_hash[font.file_hash].append(font)
-            else:
-                fonts_by_file_hash[font.file_hash] = [font]
+            fonts_by_file_hash.setdefault(font.hash, []).append(font)
         font_references_by_file_hash = {}
         for file_hash, fonts in fonts_by_file_hash.items():
             content = fonts[0].file_content
@@ -1409,7 +1415,7 @@ def write_pdf(self, target=None, zoom=1, attachments=None, finisher=None):
                 'StemV': font.stemv,
                 'StemH': font.stemh,
                 (f'FontFile{"3" if font_type == "otf" else "2"}'):
-                    font_references_by_file_hash[font.file_hash],
+                    font_references_by_file_hash[font.hash],
             })
             if font_type == 'otf':
                 font_descriptor['Subtype'] = '/OpenType'

weasyprint/draw.py

@@ -1087,7 +1087,6 @@ def draw_first_line(stream, textbox, text_overflow, block_ellipsis, x, y,
     stream.text_matrix(*matrix.values)
     last_font = None
     string = ''
-    fonts = stream.get_fonts()
     x_advance = 0
     emojis = []
     for run in runs:
@@ -1101,19 +1100,7 @@ def draw_first_line(stream, textbox, text_overflow, block_ellipsis, x, y,
 
         # Font content
         pango_font = glyph_item.item.analysis.font
-        hb_font = pango.pango_font_get_hb_font(pango_font)
-        hb_face = harfbuzz.hb_font_get_face(hb_font)
-        font_hash = hash(hb_face)
-        if font_hash in fonts:
-            font = fonts[font_hash]
-        else:
-            hb_blob = ffi.gc(
-                harfbuzz.hb_face_reference_blob(hb_face),
-                harfbuzz.hb_blob_destroy)
-            hb_data = harfbuzz.hb_blob_get_data(hb_blob, stream.length)
-            file_content = ffi.unpack(hb_data, int(stream.length[0]))
-            index = harfbuzz.hb_face_get_index(hb_face)
-            font = stream.add_font(font_hash, file_content, pango_font, index)
+        font = stream.add_font(pango_font)
 
         # Positions of the glyphs in the UTF-8 string
         utf8_positions = [offset + clusters[i] for i in range(1, num_glyphs)]
@@ -1178,6 +1165,8 @@ def draw_first_line(stream, textbox, text_overflow, block_ellipsis, x, y,
             previous_utf8_position = utf8_position
 
             if font.svg:
+                hb_font = pango.pango_font_get_hb_font(pango_font)
+                hb_face = harfbuzz.hb_font_get_face(hb_font)
                 hb_blob = ffi.gc(
                     harfbuzz.hb_ot_color_glyph_reference_svg(hb_face, glyph),
                     harfbuzz.hb_blob_destroy)
@@ -1189,6 +1178,7 @@ def draw_first_line(stream, textbox, text_overflow, block_ellipsis, x, y,
                     a = d = font.widths[glyph] / 1000 / font.upem * font_size
                     emojis.append([image, font, a, d, x_advance, 0])
             elif font.png:
+                hb_font = pango.pango_font_get_hb_font(pango_font)
                 hb_blob = ffi.gc(
                     harfbuzz.hb_ot_color_glyph_reference_png(hb_font, glyph),
                     harfbuzz.hb_blob_destroy)

weasyprint/text/ffi.py

@@ -44,6 +44,7 @@
     typedef ... PangoAttrList;
     typedef ... PangoAttrClass;
     typedef ... PangoFont;
+    typedef ... PangoFontFace;
     typedef guint PangoGlyph;
     typedef gint PangoGlyphUnit;
 
@@ -214,6 +215,7 @@
     PangoLayoutLine * pango_layout_get_line_readonly (
         PangoLayout *layout, int line);
 
+    PangoFontFace * pango_font_get_face (PangoFont *font);
     hb_font_t * pango_font_get_hb_font (PangoFont *font);
 
     PangoFontDescription * pango_font_description_new (void);