Support mathjax (v3) math rendering

xkcd-script/generator/README.md

@@ -0,0 +1,25 @@
+# xkcd-script font generation pipeline
+
+Each script runs in order inside the `fontbuilder` Docker image; `run.sh` orchestrates them and accepts an optional starting step (`./run.sh 5` skips pt1–pt4).
+
+## Stages
+
+| # | Script | What it does |
+|---|---|---|
+| 1 | `pt1_character_extraction.py` | Extract character strokes from `handwriting_minimal.png` (scikit-image). |
+| 2 | `pt2_character_classification.py` | Cluster strokes into lines (k-means, fixed seed). |
+| 3 | `pt3_ppm_to_svg.py` | Convert per-character PPM → SVG via `potrace`. |
+| 4 | `pt4_additional_sources.py` | Trace extra glyphs from comic panels and `extras/`. |
+| 5 | `pt5_svg_to_font.py` | Import SVG glyphs into a FontForge SFD; apply stroke normalisation, weight nudges, math-symbol imports. |
+| 6 | `pt6_derived_chars.py` | Build derived/composed glyphs: diacritics, ligatures, Greek, IPA, combining marks, math cmap aliases (U+1D400 block via altuni). |
+| 7 | `pt7_font_properties.py` | Apply kerning, GPOS anchors, pin CFF hints and OS/2 metrics. Output is `xkcd-script-pt7.sfd` — the **base** font used for everything downstream. |
+| 8 | `pt8_derivatives.py` | Orchestrator. Runs each `pt8X_*.py` derivative step in turn. |
+| 9 | `pt9_gen_reprod_font.py` | Scrub the SFD for reproducibility, freeze CFF charstrings, generate committed binaries (otf/ttf/woff). |
+
+## Derivatives (pt8)
+
+`pt7` produces a single kitchen-sink base SFD with everything — Latin, Greek, math symbols and aliases, ligatures, combining marks. Each `pt8X_<name>.py` reads that base and either writes its own derivative SFD or extracts data from it to splice elsewhere; `pt8_derivatives.py` runs them with `runpy`.
+
+Today there is no live derivative font: the sole entry, `pt8a_mathjax3.py`, only extracts extensible-glyph outline data into `../xkcd-mathjax3.js`. The display-sized large operators that used to live in a separate mathjax3 WOFF are now stylistic alternates in the base font (`ss01`).
+
+Because derivatives can only subtract or overlay what pt7 already has, **everything plausibly useful belongs in pt7**. Don't pre-strip pt7 for size — that loses the subtractive option.

xkcd-script/generator/pt2_character_classification.py

@@ -53,7 +53,7 @@
 U N A U T H O R I T A T I V E N E S S   L EA T H E R B A R K   I N T R A CO L I C   M I CR OCH E L I A
 O F F S I D ER   G LA S S W EE D   R O TT O L O   A LB E R T I T E   H ER M A T O RR H A C H I S
 O R G A N O M E T A LL I C   S E G R E G A T I ON I S T   U N E V A N G E L I C   CA M PS TO O L
-+ - x * ! ? # @ $ % ¦ & ^ _ — – - ( ) [ ] { } / \ < > ÷ ± √ Σ
++ - × * ! ? # @ $ % ¦ & ^ _ — – - ( ) [ ] { } / \ < > ÷ ± √ Σ
 1 2 3 4 5 6 7 8 9 0 ∫ = ≈ ≠ ~ ≤ ≥ |> <| 🎂 . , ; : “ H I ” ’ ‘ C A N ' T ' "
 É Ò Å Ü ≪ ≫ ‽ Ē Ő “ ”
 """.strip()
@@ -74,7 +74,7 @@ def merge(img1, img1_bbox, img2, img2_bbox):
     shape = bbox[3] - bbox[1], bbox[2] - bbox[0], 3
     img1_slice = [slice(img1_bbox[1] - bbox[1], img1_bbox[3] - bbox[1]),
                   slice(img1_bbox[0] - bbox[0], img1_bbox[2] - bbox[0])]
-    
+
     img2_slice = [slice(img2_bbox[1] - bbox[1], img2_bbox[3] - bbox[1]),
                   slice(img2_bbox[0] - bbox[0], img2_bbox[2] - bbox[0])]
 
@@ -102,6 +102,18 @@ def merge(img1, img1_bbox, img2, img2_bbox):
 
 
 
+_LIGATURE_SOURCES = [
+    (5, 4, 5, 'TH'),   # UNAUTHORITATIVENESS: T + H
+    (5, 31, 32, 'TR'),  # INTRACOLIC: T + R
+    (6, 24, 25, 'TI'),  # ALBERTITE: T + I
+]
+for line_no, left_no, right_no, lig in _LIGATURE_SOURCES:
+    _, left_bbox, left_img = characters_by_line[line_no][left_no]
+    _, right_bbox, right_img = characters_by_line[line_no][right_no]
+    lig_img, lig_bbox = merge(left_img, left_bbox, right_img, right_bbox)
+    characters_by_line[line_no].append([lig, lig_bbox, lig_img])
+
+
 import skimage.io
 
 if not os.path.isdir('../generated/characters'):

xkcd-script/generator/pt4_additional_sources.py

@@ -68,17 +68,23 @@ def _clean_potrace_svg(raw_svg_path, clean_svg_path):
     os.remove(clean_svg_path + '.sfd')
 
 
-def extract_symbol(arr, y0, y1, x0, x1, name, exclude=None):
+def extract_symbol(arr, y0, y1, x0, x1, name, exclude=None, pad=0):
     """Crop glyph region, upsample, binarise, run potrace, clean, save SVG.
 
     exclude: optional list of (y0, y1, x0, x1) regions in full-image coordinates
              to blank out (set to background) before potrace, for removing
              artefacts that cannot be separated by tightening the main crop.
+    pad: white-pixel border added around the crop before upsampling.  Use
+         when the source PNG is tightly cropped and ink touches the edge —
+         potrace otherwise produces edge artefacts where contours run into
+         the canvas boundary.
     """
     crop = arr[y0:y1, x0:x1].copy()
     if exclude:
         for ey0, ey1, ex0, ex1 in exclude:
             crop[ey0 - y0:ey1 - y0, ex0 - x0:ex1 - x0] = 255
+    if pad:
+        crop = np.pad(crop, pad, mode='constant', constant_values=255)
     big = Image.fromarray(crop).resize(
         (crop.shape[1] * UPSAMPLE, crop.shape[0] * UPSAMPLE),
         Image.BILINEAR)
@@ -163,14 +169,25 @@ def extract_symbol(arr, y0, y1, x0, x1, name, exclude=None):
     ('AElig', '2763_linguistics_gossip_2x__AE'), # Æ U+00C6 source
     ('cedilla', '2034_equations_2x__cedilla.hand-tweaked'),  # hook cedilla mark source
     ('epsilon', '2034_equations_2x__epsilon'),               # ε U+03B5 source
+    ('Lambda', '2034_equations_2x__Lambda'),                 # Λ U+039B source
+    ('rounded_d', '2520_symbols_2x__rounded_d'),             # ∂ U+2202 source
+    ('infinity', '2343_mathematical_symbol_fight_2x__infinity'),              # ∞ U+221E source
+    ('right_double_arrow', '2343_mathematical_symbol_fight_2x__right_double_arrow'),  # ⇒ U+21D2 source
+    ('right_half_arrow', '2343_mathematical_symbol_fight_2x__right_half_arrow'),      # ⇀ U+21C0 source
+    ('right_lim_arrow', '2343_mathematical_symbol_fight_2x__right_lim_arrow'),        # → U+2192 source
+    ('triangle', '2343_mathematical_symbol_fight_2x__triangle'),              # △ U+25B3 source
+    ('circled_times', '2034_equations_2x__circled_times'),                    # ⊗ U+2297 source
+    ('sqrt_vertical', 'sqrt_vertical'),                                       # √ tall surd, unencoded glyph `radical.tall`
+    ('braceleft_tall', '2435_geothmetic_meandian_2x__brace__shortened'),      # { tall brace for MathJax stretchy assembly
+    ('parenleft_tall', '2059_modified_bayes_theorem_2x__lparen'),             # ( tall paren for MathJax stretchy assembly (\binom, \left(, pmatrix); right paren is mirrored in pt5
 ]
 
 print('Extracting hand-drawn extras...')
 for name, filename in EXTRAS:
     src_path = os.path.join(EXTRAS_DIR, f'{filename}.png')
     arr_extra = np.array(Image.open(src_path).convert('L'))
     h, w = arr_extra.shape
-    extract_symbol(arr_extra, 0, h, 0, w, name)
+    extract_symbol(arr_extra, 0, h, 0, w, name, pad=10)
 
 
 # ---------------------------------------------------------------------------