Add a FONTSPLITTING_EXTRA parameter to consider non-Unicode glyphs in…

… the splitting. Use this param for STIX.

fonts/OpenTypeMath/Asana-Math/config.py

@@ -22,6 +22,8 @@
 MATHFONT = "Asana-Math.otf"
 MAINFONTS = None
 
+FONTSPLITTING_EXTRA = {}
+
 FONTDATA = {
     "FileVersion": "2.3",
     "Year": "2013",
@@ -30,7 +32,7 @@
     "lineH": .8,
     "lineD": .2,
     "hasStyleChar": True
-    }
+}
 
 RULECHAR = 0x2212
 

fonts/OpenTypeMath/Gyre-Pagella/config.py

@@ -22,6 +22,8 @@
 MATHFONT="texgyrepagella-math.otf"
 MAINFONTS = None
 
+FONTSPLITTING_EXTRA = {}
+
 FONTDATA = {
     "FileVersion": "2.3",
     "Year": "2013",

fonts/OpenTypeMath/Gyre-Termes/config.py

@@ -22,6 +22,8 @@
 MATHFONT="texgyretermes-math.otf"
 MAINFONTS = None
 
+FONTSPLITTING_EXTRA = {}
+
 FONTDATA = {
     "FileVersion": "2.3",
     "Year": "2013",

fonts/OpenTypeMath/Latin-Modern/config.py

@@ -22,6 +22,8 @@
 MATHFONT = "latinmodern-math.otf"
 MAINFONTS = None
 
+FONTSPLITTING_EXTRA = {}
+
 FONTDATA = {
     "FileVersion": "2.3",
     "Year": "2013",

fonts/OpenTypeMath/Neo-Euler/config.py

@@ -22,6 +22,8 @@
 MATHFONT = "euler.otf"
 MAINFONTS = None
 
+FONTSPLITTING_EXTRA = {}
+
 FONTDATA = {
     "FileVersion": "2.3",
     "Year": "2013",

fonts/OpenTypeMath/STIX-Web/config.py

@@ -26,7 +26,80 @@
     "Bold": "STIX-Bold.otf",
     "Italic": "STIX-Italic.otf",
     "BoldItalic": "STIX-BoldItalic.otf"
+}
+
+FONTSPLITTING_EXTRA = {
+    "Alphabets": {
+        (0xE053,0xE057),
+        0xE05D,
+        0xE068,
+        0xE0A2,
+        0xE0B3,
+        (0xE0BD,0xE0BE),
+        (0xE0DD,0xE0E2),
+        (0xE17C,0xE1B3),
+        (0xE2FC,0xE370)
+    },
+    "Marks": {
+        0xE05E,
+        (0xE061,0xE067),
+        (0xE0D8,0xE0DA),
+        (0xE28C,0xE28D)
+    },
+    "Arrows": {
+        (0xE112,0xE115),
+        (0xE11A,0xE11F),
+        (0xE120,0xE137)
+    },
+    "Operators": {
+        (0xE004,0xE02B),
+        (0xE035,0xE052),
+        (0xE059,0xE05C),
+        (0xE06B,0xE098),
+        0xE0BB,
+        0xE0EC,
+        (0xE109,0xE10A),
+        (0xE14C,0xE14F),
+        0xE28E,
+        (0xE292,0xE293),
+        (0xE3C2,0xE3C4)
+    },
+    "Symbols": {
+        (0xE0BF,0xE0D6), # Chemical structures
+        (0xE09A,0xE09B)
+    },
+    "Shapes": {
+        (0xE000,0xE002),
+        0xE099,
+        (0xE0A7,0xE0B2),
+        0xE0D7,
+        0xE0E3,
+        (0xE0E5,0xE0EB),
+        (0xE0ED,0xE0EF),
+        0xE0F1,
+        0xE104,
+        (0xE10C,0xE111),
+        (0xE116,0xE119),
+        (0xE28F,0xE291),
+        (0xE3C7,0xE3C8),
+        ("uni25A9.var", 0xE400)
+   },
+    "Misc": {
+        0xE02F,
+        (0xE0B4,0xE0B9),
+        0xE0BC,
+        (0xE0DB,0xE0DC),
+        (0xE0A4,0xE0A6),
+        (0xE0F2,0xE0F7), # Physics
+        (0xE0F8,0xE0FA),
+        (0xE0FB,0xE103),
+        (0xE105,0xE108),
+        (0xE138,0xE13A)
+    },
+    "Variants": {
+        (0xE261,0xE285) # oldstyle numbers
     }
+}
 
 FONTDATA = {
     "FileVersion": "2.3",
@@ -66,7 +139,7 @@
 TEXOLDSTYLE = "offsetN: 0xE261, remap: {0xE262: 0xE265, 0xE263: 0xE269, \
 0xE264: 0xE26D, 0xE265: 0xE271, 0xE266: 0xE275, 0xE267: 0xE279, \
 0xE268: 0xE27D, 0xE269: 0xE281, 0xE26A: 0xE285}"
-TEXOLDSTYLEFONTS = ["NONUNICODE"]
+TEXOLDSTYLEFONTS = ["VARIANTS"]
 
 SMALLOPFONTS = None
 

fonts/OpenTypeMath/fontSplitting.py

@@ -18,8 +18,8 @@
 
 # This table must be sorted.
 #
-# The script searching in this table assumes it to be sorted. If it does not
-# found a character, it will assume it is "NONUNICODE".
+# fontUtil::computeNormalSizeSplitting assumes the table to be sorted. If it
+# does not found a character, it will assume it is "NONUNICODE".
 #
 
 COPYRIGHT = "Copyright (c) 2013 The MathJax Consortium"
@@ -321,7 +321,7 @@
      (0xA720, 0xA7FF), # Latin Extended-D
      (0xFB00, 0xFB4F) # Alphabetic Presentation Forms
      ],
-    ["Alphabets"
+    ["Alphabets",
      (0x0384, 0x0390), # Greek and Coptic
      (0x03AA, 0x03B0), # Greek and Coptic
      (0x03CA, 0x03CE), # Greek and Coptic
@@ -372,5 +372,6 @@
      (0x2150, 0x218F), # Number Forms
      (0x2460, 0x24FF), # Enclosed Alphanumerics
      (0x2700, 0x27BF)  # Dingbats
-     ]
+     ],
+    ["Variants"] # Used for oldstyle numbers and glyph variants
     ]

fonts/OpenTypeMath/fontUtil.py

@@ -18,6 +18,7 @@
 
 from __future__ import print_function
 
+import sys
 from shutil import copyfile
 import fontforge
 from fontSplitting import FONTSPLITTING, COPYRIGHT
@@ -87,11 +88,18 @@ def saveFont(aFamily, aFont):
     aFont.generate("%s/ttf/%s.ttf" % (aFamily, aFont.fontname))
     aFont.generate("%s/svg/%s.svg" % (aFamily, aFont.fontname))
 
+def hasNonEmptyGlyph(aFont, aGlyphName):
+    # Check that the font has the glyph and that this glyph is not empty.
+    if not(aGlyphName in aFont and
+           aFont[aGlyphName].isWorthOutputting()):
+        return False
+
+    return True
+
 def moveGlyph(aFontFrom, aFontTo, aOldPosition, aNewPosition = None):
     # Ignore glyphs that have already been deleted before
     # (Otherwise, FontForge will copy them as "blank" glyphs)
-    if not(aOldPosition in aFontFrom and
-           aFontFrom[aOldPosition].isWorthOutputting()):
+    if not hasNonEmptyGlyph(aFontFrom, aOldPosition):
         return
 
     # Move the glyph from the font aFontFrom at position aOldPosition
@@ -109,6 +117,20 @@ def moveRange(aFontFrom, aFontTo, aRangeStart, aRangeEnd):
     for codePoint in range(aRangeStart, aRangeEnd+1):
         moveGlyph(aFontFrom, aFontTo, codePoint)
 
+def moveSubset(aFontFrom, aFontTo, aSubset):
+    # move the glyphs in a subset described.
+    for r in aSubset:
+        if type(r) == int:
+            # Single code point: move one glyph.
+            moveGlyph(aFontFrom, aFontTo, r)
+        elif type(r) == tuple:
+            if type(r[0]) == int:
+                # (start, end): move the range of glyphs.
+                moveRange(aFontFrom, aFontTo, r[0], r[1])
+            elif type(r[0]) == str:
+                # (glyphname, newcodepoint): move a non-Unicode glyph
+                moveGlyph(aFontFrom, aFontTo, r[0], r[1])
+
 def getTestString(aFont, aMaxLength):
     s = ""
     # Pick at most 10 glyphs from the font to build a test string
@@ -441,6 +463,9 @@ def computeNormalSizeSplitting(self):
                     size0[codePoint] = name.upper()
                     break
 
+            # FIXME: also search in FONTSPLITTING_EXTRA?
+            # (In theory this should not be needed, since NonUnicode components
+            # are directly copied into SizeMax).
             if not(found):
                 size0[codePoint] = "NONUNICODE"
 
@@ -584,7 +609,7 @@ def moveToPlane0PUA(self, aGlyphName):
         if aGlyphName not in self.mPUAContent:
             # New piece: copy it into the PUA and save the code point.
             if self.mPUAPointer > 0xF8FF:
-                raise BaseException("Too many characters in the PUA. Not supported by the font splitter.")
+                raise BaseException("Too many characters in the Plane 0 PUA. Not supported by the font splitter.")
             codePoint = self.mPUAPointer
             self.mMathFont.selection.select(aGlyphName)
             self.mMathFont.copy()
@@ -607,7 +632,7 @@ def copySizeVariant(self, aIsHorizontal, aSize,
             style = aStyle
         elif aSize == 0:
             if self.isPrivateCharacter(aGlyphName):
-                print("Warning: non-Unicode glyphs %s used for the normal size! Will be copied to the PUA of Size %d..." % (aGlyphName, self.mMaxSize))
+                print("Warning: non-Unicode glyphs %s used for the normal size! Will be copied to the PUA of Size %d..." % (aGlyphName, self.mMaxSize), file=sys.stderr)
                 style = None
                 codePoint = self.moveToPlane0PUA(aGlyphName)
                 size = self.mMaxSize

fonts/OpenTypeMath/splitFont.py

@@ -110,12 +110,11 @@ def boolToString(b):
         for subset in FONTSPLITTING:
             name = subset[0]
             font=fontUtil.newFont(FONTFAMILY, fontFile, config, name, weight)
-            for i in range(1,len(subset)):
-                r = subset[i]
-                if type(r) == int:
-                    fontUtil.moveGlyph(oldfont, font, r)
-                else:
-                    fontUtil.moveRange(oldfont, font, r[0], r[1])
+            fontUtil.moveSubset(oldfont, font, subset)
+
+            if name in config.FONTSPLITTING_EXTRA:
+                fontUtil.moveSubset(oldfont, font,
+                                    config.FONTSPLITTING_EXTRA[name])
 
             if name == "Monospace" and "u1D670" in font:
                 # For the monospace font, ensure that the space has the
@@ -131,8 +130,24 @@ def boolToString(b):
 
         # Save the rest of the glyphs in a NonUnicode font
         font=fontUtil.newFont(FONTFAMILY, fontFile, config, "NonUnicode", weight)
+        PUAPointer = 0xE000
         for g in oldfont.glyphs():
-            fontUtil.moveGlyph(oldfont, font, g.glyphname)
+
+            if g.glyphname == ".notdef":
+                continue
+
+            if PUAPointer > 0xF8FF:
+                raise BaseException("Too many characters in the Plane 0 PUA. Not supported by the font splitter.")
+
+            if fontUtil.hasNonEmptyGlyph(oldfont, PUAPointer):
+                print("Warning: 0x%X is already present in the Plane 0 PUA of the original font. %s will not be moved." % (PUAPointer, g.glyphname),
+                      file=sys.stderr)
+                fontUtil.moveGlyph(oldfont, font, g.glyphname)
+            else:
+                fontUtil.moveGlyph(oldfont, font, g.glyphname, PUAPointer)
+
+            PUAPointer += 1
+
         fontUtil.saveFont(FONTFAMILY, font)
         font.close()
         oldfont.close()