Use libraqm for text in vector outputs

Author: QuLogic

lib/matplotlib/_text_helpers.py

@@ -4,29 +4,23 @@
 
 from __future__ import annotations
 
-import dataclasses
+from collections.abc import Iterator
 
 from . import _api
-from .ft2font import FT2Font, GlyphIndexType, Kerning, LoadFlags
+from .ft2font import FT2Font, CharacterCodeType, LayoutItem, LoadFlags
 
 
-@dataclasses.dataclass(frozen=True)
-class LayoutItem:
-    ft_object: FT2Font
-    char: str
-    glyph_index: GlyphIndexType
-    x: float
-    prev_kern: float
-
-
-def warn_on_missing_glyph(codepoint, fontnames):
+def warn_on_missing_glyph(codepoint: CharacterCodeType, fontnames: str):
     _api.warn_external(
         f"Glyph {codepoint} "
         f"({chr(codepoint).encode('ascii', 'namereplace').decode('ascii')}) "
         f"missing from font(s) {fontnames}.")
 
 
-def layout(string, font, *, features=None, kern_mode=Kerning.DEFAULT, language=None):
+def layout(string: str, font: FT2Font, *,
+           features: tuple[str] | None = None,
+           language: str | tuple[tuple[str, int, int], ...] | None = None
+           ) -> Iterator[LayoutItem]:
     """
     Render *string* with *font*.
 
@@ -41,8 +35,6 @@ def layout(string, font, *, features=None, kern_mode=Kerning.DEFAULT, language=N
         The font.
     features : tuple of str, optional
         The font features to apply to the text.
-    kern_mode : Kerning
-        A FreeType kerning mode.
     language : str, optional
         The language of the text in a format accepted by libraqm, namely `a BCP47
         language code <https://www.w3.org/International/articles/language-tags/>`_.
@@ -51,20 +43,8 @@ def layout(string, font, *, features=None, kern_mode=Kerning.DEFAULT, language=N
     ------
     LayoutItem
     """
-    x = 0
-    prev_glyph_index = None
-    char_to_font = font._get_fontmap(string)  # TODO: Pass in features and language.
-    base_font = font
-    for char in string:
-        # This has done the fallback logic
-        font = char_to_font.get(char, base_font)
-        glyph_index = font.get_char_index(ord(char))
-        kern = (
-            base_font.get_kerning(prev_glyph_index, glyph_index, kern_mode) / 64
-            if prev_glyph_index is not None else 0.
-        )
-        x += kern
-        glyph = font.load_glyph(glyph_index, flags=LoadFlags.NO_HINTING)
-        yield LayoutItem(font, char, glyph_index, x, kern)
-        x += glyph.linearHoriAdvance / 65536
-        prev_glyph_index = glyph_index
+    for raqm_item in font._layout(string, LoadFlags.NO_HINTING,
+                                  features=features, language=language):
+        raqm_item.ft_object.load_glyph(raqm_item.glyph_index,
+                                       flags=LoadFlags.NO_HINTING)
+        yield raqm_item

lib/matplotlib/backends/_backend_pdf_ps.py

@@ -220,8 +220,8 @@ def track(self, font: FT2Font, s: str) -> list[tuple[int, CharacterCodeType]]:
             and the character codes will be returned from the string unchanged.
         """
         return [
-            self.track_glyph(f, ord(c), f.get_char_index(ord(c)))
-            for c, f in font._get_fontmap(s).items()
+            self.track_glyph(raqm_item.ft_object, raqm_item.char, raqm_item.glyph_index)
+            for raqm_item in font._layout(s, ft2font.LoadFlags.NO_HINTING)
         ]
 
     def track_glyph(self, font: FT2Font, chars: str | CharacterCodeType,

lib/matplotlib/backends/backend_pdf.py

@@ -34,7 +34,7 @@
 from matplotlib.figure import Figure
 from matplotlib.font_manager import get_font, fontManager as _fontManager
 from matplotlib._afm import AFM
-from matplotlib.ft2font import FT2Font, FaceFlags, Kerning, LoadFlags, StyleFlags
+from matplotlib.ft2font import FT2Font, FaceFlags, LoadFlags, StyleFlags
 from matplotlib.transforms import Affine2D, BboxBase
 from matplotlib.path import Path
 from matplotlib.dates import UTC
@@ -469,6 +469,7 @@ class Op(Enum):
     textpos = b'Td'
     selectfont = b'Tf'
     textmatrix = b'Tm'
+    textrise = b'Ts'
     show = b'Tj'
     showkern = b'TJ'
     setlinewidth = b'w'
@@ -2285,6 +2286,8 @@ def draw_text(self, gc, x, y, s, prop, angle, ismath=False, mtext=None):
         # If fonttype is neither 3 nor 42, emit the whole string at once
         # without manual kerning.
         if fonttype not in [3, 42]:
+            if not mpl.rcParams['pdf.use14corefonts']:
+                self.file._character_tracker.track(font, s)
             self.file.output(Op.begin_text,
                              self.file.fontName(prop), fontsize, Op.selectfont)
             self._setup_textpos(x, y, angle)
@@ -2305,13 +2308,16 @@ def draw_text(self, gc, x, y, s, prop, angle, ismath=False, mtext=None):
         # kerning between chunks.
         else:
             def output_singlebyte_chunk(kerns_or_chars):
+                if not kerns_or_chars:
+                    return
                 self.file.output(
                     # See pdf spec "Text space details" for the 1000/fontsize
                     # (aka. 1000/T_fs) factor.
                     [(-1000 * next(group) / fontsize) if tp == float  # a kern
                      else self._encode_glyphs(group, fonttype)
                      for tp, group in itertools.groupby(kerns_or_chars, type)],
                     Op.showkern)
+                kerns_or_chars.clear()
             # Do the rotation and global translation as a single matrix
             # concatenation up front
             self.file.output(Op.gsave)
@@ -2326,24 +2332,26 @@ def output_singlebyte_chunk(kerns_or_chars):
             # Emit all the characters in a BT/ET group.
             self.file.output(Op.begin_text)
             for item in _text_helpers.layout(s, font, features=features,
-                                             kern_mode=Kerning.UNFITTED,
                                              language=language):
                 subset, charcode = self.file._character_tracker.track_glyph(
                     item.ft_object, item.char, item.glyph_index)
                 if (item.ft_object, subset) != prev_font:
-                    if singlebyte_chunk:
-                        output_singlebyte_chunk(singlebyte_chunk)
+                    output_singlebyte_chunk(singlebyte_chunk)
                     ft_name = self.file.fontName(item.ft_object.fname, subset)
                     self.file.output(ft_name, fontsize, Op.selectfont)
                     self._setup_textpos(item.x, 0, 0, prev_start_x, 0, 0)
-                    singlebyte_chunk = []
                     prev_font = (item.ft_object, subset)
                     prev_start_x = item.x
+                if item.y:
+                    output_singlebyte_chunk(singlebyte_chunk)
+                    self.file.output(item.y, Op.textrise)
                 if item.prev_kern:
                     singlebyte_chunk.append(item.prev_kern)
                 singlebyte_chunk.append(charcode)
-            if singlebyte_chunk:
-                output_singlebyte_chunk(singlebyte_chunk)
+                if item.y:
+                    output_singlebyte_chunk(singlebyte_chunk)
+                    self.file.output(0, Op.textrise)
+            output_singlebyte_chunk(singlebyte_chunk)
             self.file.output(Op.end_text)
             self.file.output(Op.grestore)
 

lib/matplotlib/backends/backend_ps.py

@@ -776,7 +776,7 @@ def draw_text(self, gc, x, y, s, prop, angle, ismath=False, mtext=None):
         if ismath:
             return self.draw_mathtext(gc, x, y, s, prop, angle)
 
-        stream = []  # list of (ps_name, x, char_name)
+        stream = []  # list of (ps_name, x, y, char_name)
 
         if mpl.rcParams['ps.useafm']:
             font = self._get_font_afm(prop)
@@ -794,7 +794,7 @@ def draw_text(self, gc, x, y, s, prop, angle, ismath=False, mtext=None):
                 kern = font.get_kern_dist_from_name(last_name, name)
                 last_name = name
                 thisx += kern * scale
-                stream.append((ps_name, thisx, name))
+                stream.append((ps_name, thisx, 0, name))
                 thisx += width * scale
 
         else:
@@ -814,14 +814,13 @@ def draw_text(self, gc, x, y, s, prop, angle, ismath=False, mtext=None):
                 ps_name = (item.ft_object.postscript_name
                            .encode("ascii", "replace").decode("ascii"))
                 glyph_name = item.ft_object.get_glyph_name(item.glyph_index)
-                stream.append((f'{ps_name}-{subset}', item.x, glyph_name))
+                stream.append((f'{ps_name}-{subset}', item.x, item.y, glyph_name))
         self.set_color(*gc.get_rgb())
 
-        for ps_name, group in itertools. \
-                groupby(stream, lambda entry: entry[0]):
+        for ps_name, group in itertools.groupby(stream, lambda entry: entry[0]):
             self.set_font(ps_name, prop.get_size_in_points(), False)
-            thetext = "\n".join(f"{x:g} 0 m /{name:s} glyphshow"
-                                for _, x, name in group)
+            thetext = "\n".join(f"{x:g} {y:g} m /{name:s} glyphshow"
+                                for _, x, y, name in group)
             self._pswriter.write(f"""\
 gsave
 {self._get_clip_cmd(gc)}

lib/matplotlib/backends/backend_svg.py

@@ -1048,6 +1048,11 @@ def _draw_text_as_path(self, gc, x, y, s, prop, angle, ismath, mtext=None):
         text2path = self._text2path
         color = rgb2hex(gc.get_rgb())
         fontsize = prop.get_size_in_points()
+        if mtext is not None:
+            features = mtext.get_fontfeatures()
+            language = mtext.get_language()
+        else:
+            features = language = None
 
         style = {}
         if color != '#000000':
@@ -1068,7 +1073,8 @@ def _draw_text_as_path(self, gc, x, y, s, prop, angle, ismath, mtext=None):
         if not ismath:
             font = text2path._get_font(prop)
             glyph_info, glyph_map_new, rects = text2path.get_glyphs_with_font(
-                font, s, glyph_map=glyph_map, return_new_glyphs_only=True)
+                font, s, features=features, language=language,
+                glyph_map=glyph_map, return_new_glyphs_only=True)
             self._update_glyph_map_defs(glyph_map_new)
 
             for glyph_repr, xposition, yposition, scale in glyph_info:

lib/matplotlib/ft2font.pyi

@@ -219,7 +219,6 @@ class FT2Font(Buffer):
     ) -> None: ...
     if sys.version_info[:2] >= (3, 12):
         def __buffer__(self, flags: int) -> memoryview: ...
-    def _get_fontmap(self, string: str) -> dict[str, FT2Font]: ...
     def _layout(
         self,
         text: str,

lib/matplotlib/tests/test_ft2font.py

@@ -972,7 +972,7 @@ def test_fallback_last_resort(recwarn):
            "Glyph 128579 (\\N{UPSIDE-DOWN FACE}) missing from font(s)")
 
 
-def test__get_fontmap():
+def test__layout():
     fonts, test_str = _gen_multi_font_text()
     # Add some glyphs that don't exist in either font to check the Last Resort fallback.
     missing_glyphs = '\n几个汉字'
@@ -981,11 +981,11 @@ def test__get_fontmap():
     ft = fm.get_font(
         fm.fontManager._find_fonts_by_props(fm.FontProperties(family=fonts))
     )
-    fontmap = ft._get_fontmap(test_str)
-    for char, font in fontmap.items():
-        if char in missing_glyphs:
-            assert Path(font.fname).name == 'LastResortHE-Regular.ttf'
-        elif ord(char) > 127:
-            assert Path(font.fname).name == 'DejaVuSans.ttf'
-        else:
-            assert Path(font.fname).name == 'cmr10.ttf'
+    for substr in test_str.split('\n'):
+        for item in ft._layout(substr, ft2font.LoadFlags.DEFAULT):
+            if item.char in missing_glyphs:
+                assert Path(item.ft_object.fname).name == 'LastResortHE-Regular.ttf'
+            elif ord(item.char) > 127:
+                assert Path(item.ft_object.fname).name == 'DejaVuSans.ttf'
+            else:
+                assert Path(item.ft_object.fname).name == 'cmr10.ttf'

lib/matplotlib/tests/test_text.py

@@ -113,7 +113,7 @@ def find_matplotlib_font(**kw):
     ax.set_yticks([])
 
 
-@image_comparison(['complex.png'])
+@image_comparison(['complex'], extensions=['png', 'pdf', 'svg', 'eps'])
 def test_complex_shaping():
     # Raqm is Arabic for writing; note that because Arabic is RTL, the characters here
     # may seem to be in a different order than expected, but libraqm will order them
@@ -1204,7 +1204,8 @@ def test_ytick_rotation_mode():
     plt.subplots_adjust(left=0.4, right=0.6, top=.99, bottom=.01)
 
 
-@image_comparison(baseline_images=['features.png'], remove_text=False, style='mpl20')
+@image_comparison(['features'], remove_text=False, style='mpl20',
+                  extensions=['png', 'pdf', 'svg', 'eps'])
 def test_text_features():
     fig = plt.figure(figsize=(5, 1.5))
     t = fig.text(1, 0.7, 'Default: fi ffi fl st',
@@ -1234,7 +1235,8 @@ def test_text_language_invalid(input, match):
         Text(0, 0, 'foo', language=input)
 
 
-@image_comparison(baseline_images=['language.png'], remove_text=False, style='mpl20')
+@image_comparison(['language'], remove_text=False, style='mpl20',
+                  extensions=['png', 'pdf', 'svg', 'eps'])
 def test_text_language():
     fig = plt.figure(figsize=(5, 3))
 

lib/matplotlib/textpath.py

@@ -147,15 +147,16 @@ def get_glyphs_with_font(self, font, s, glyph_map=None,
             glyph_map_new = glyph_map
 
         xpositions = []
+        ypositions = []
         glyph_reprs = []
         for item in _text_helpers.layout(s, font, features=features, language=language):
             glyph_repr = self._get_glyph_repr(item.ft_object, item.glyph_index)
             glyph_reprs.append(glyph_repr)
             xpositions.append(item.x)
+            ypositions.append(item.y)
             if glyph_repr not in glyph_map:
                 glyph_map_new[glyph_repr] = item.ft_object.get_path()
 
-        ypositions = [0] * len(xpositions)
         sizes = [1.] * len(xpositions)
 
         rects = []

src/ft2font_wrapper.cpp

@@ -623,54 +623,6 @@ PyFT2Font_get_kerning(PyFT2Font *self, FT_UInt left, FT_UInt right,
     return self->get_kerning(left, right, mode);
 }
 
-const char *PyFT2Font_get_fontmap__doc__ = R"""(
-    Get a mapping between characters and the font that includes them.
-
-    .. warning::
-        This API uses the fallback list and is both private and provisional: do not use
-        it directly.
-
-    Parameters
-    ----------
-    text : str
-        The characters for which to find fonts.
-
-    Returns
-    -------
-    dict[str, FT2Font]
-        A dictionary mapping unicode characters to `.FT2Font` objects.
-)""";
-
-static py::dict
-PyFT2Font_get_fontmap(PyFT2Font *self, std::u32string text)
-{
-    std::set<FT_ULong> codepoints;
-
-    py::dict char_to_font;
-    for (auto code : text) {
-        if (!codepoints.insert(code).second) {
-            continue;
-        }
-
-        py::object target_font;
-        int index;
-        if (self->get_char_fallback_index(code, index)) {
-            if (index >= 0) {
-                target_font = self->fallbacks[index];
-            } else {
-                target_font = py::cast(self);
-            }
-        } else {
-            // TODO Handle recursion!
-            target_font = py::cast(self);
-        }
-
-        auto key = py::cast(std::u32string(1, code));
-        char_to_font[key] = target_font;
-    }
-    return char_to_font;
-}
-
 const char *PyFT2Font_set_text__doc__ = R"""(
     Set the text *string* and *angle*.
 
@@ -1705,8 +1657,6 @@ PYBIND11_MODULE(ft2font, m, py::mod_gil_not_used())
              "string"_a, "angle"_a=0.0, "flags"_a=LoadFlags::FORCE_AUTOHINT, py::kw_only(),
              "features"_a=nullptr, "language"_a=nullptr,
              PyFT2Font_set_text__doc__)
-        .def("_get_fontmap", &PyFT2Font_get_fontmap, "string"_a,
-             PyFT2Font_get_fontmap__doc__)
         .def("get_num_glyphs", &PyFT2Font::get_num_glyphs,
              PyFT2Font_get_num_glyphs__doc__)
         .def("load_char", &PyFT2Font_load_char,