etassia 0.1.0

data/lib/etassia/svg/tspan.rb

@@ -0,0 +1,48 @@
+# frozen_string_literal: true
+
+require "lutaml/model"
+require_relative "element"
+
+module Etassia
+  module Svg
+    # SVG Tspan element <tspan>
+    #
+    # One styled run inside a <text> element: a bold or italic word, or the
+    # start of a new line after a hard line break. Only the attributes a
+    # markdown run actually needs — position resets (`x`, `y`) and per-run
+    # styling (`font-weight`, `font-style`, and for creole runs `font-family`,
+    # `font-size` and `text-decoration`; `fill` is a base attribute).
+    #
+    # `dy` is not an attribute: the :metanorma profile rejects it on a
+    # tspan. A run starting a new line sets `line_shift` instead, and the
+    # parent Text turns it into an absolute `y` (see Text#place_lines).
+    # Everything else (font family/size, fill, text-anchor) is inherited
+    # from the parent <text> element and never repeated here.
+    class Tspan < Element
+      attribute :x, :float
+      attribute :y, :float
+      # Lines below the previous placed run; never written to the SVG.
+      attribute :line_shift, :integer
+      attribute :font_weight, :string
+      attribute :font_style, :string
+      attribute :font_family, :string
+      attribute :font_size, :string
+      attribute :text_decoration, :string
+      attribute :content, :string, collection: true
+
+      writes_attributes :x, :y, :font_weight, :font_style, :font_family,
+                        :font_size, :text_decoration
+
+      protected
+
+      # `content` is declared `collection: true`, so join rather than
+      # interpolate.
+      #
+      # @return [String] XML string
+      def element_markup
+        escaped_content = Escaping.escape_text(Array(content).join)
+        "<tspan#{build_attributes}>#{escaped_content}</tspan>"
+      end
+    end
+  end
+end

data/lib/etassia/svg.rb

@@ -0,0 +1,38 @@
+# frozen_string_literal: true
+
+require_relative "svg/escaping"
+require_relative "svg/numbers"
+require_relative "svg/path_geometry"
+require_relative "svg/style"
+require_relative "svg/element"
+require_relative "svg/document"
+require_relative "svg/group"
+require_relative "svg/rect"
+require_relative "svg/circle"
+require_relative "svg/line"
+require_relative "svg/polygon"
+require_relative "svg/arrowhead"
+require_relative "svg/path"
+require_relative "svg/polyline"
+require_relative "svg/tspan"
+require_relative "svg/text"
+require_relative "svg/ellipse"
+
+module Etassia
+  # The SVG document model. Everything Etassia writes into an SVG goes
+  # through these classes.
+  module Svg
+    # The svg_conform profile Etassia's output is built to satisfy.
+    #
+    # Requires a viewBox, rejects @import/<link>/external images, and
+    # enforces the SVG Tiny 1.2 element/attribute table -- but does not
+    # by itself guarantee self-containment: an external
+    # <?xml-stylesheet?> PI still validates clean under it. Nothing
+    # Etassia emits produces one; spec/etassia/svg_conformance_spec.rb holds
+    # that line -- don't drop it assuming the profile alone covers it.
+    #
+    # :svg_1_2_rfc restricts colour/fonts and would erase themes -- not
+    # a drop-in equivalent.
+    CONFORMANCE_PROFILE = :metanorma
+  end
+end

data/lib/etassia/text_measurement/arial_advances.rb

@@ -0,0 +1,122 @@
+# frozen_string_literal: true
+
+# GENERATED by scripts/generate_text_advances.rb; do not edit by hand.
+#
+# Horizontal advance widths in per-mille em, read from the hmtx table of
+# Liberation Sans Regular (SIL OFL 1.1), which is metric-compatible with
+# Arial and Helvetica. Cross-checked against Arial where it has the glyph:
+# a codepoint where the two differ by more than 1/1000 em is
+# absent here, not guessed.
+# Advances are numeric facts about a glyph, not a copy of any font.
+module Etassia
+  class TextMeasurement
+    module ArialAdvances
+      # A line "U+XXXX: a b c" starts a run at that codepoint; every number
+      # is the advance of the next consecutive codepoint, and a line without
+      # a marker continues the previous run.
+      DATA = <<~RUNS
+        U+0020: 278 278 355 556 556 889 667 191 333 333 389 584 278 333 278 278
+        556 556 556 556 556 556 556 556 556 556 278 278 584 584 584 556
+        1015 667 667 722 722 667 611 778 722 278 500 667 556 833 722 778
+        667 778 722 667 611 722 667 944 667 667 611 278 278 278 469 556
+        333 556 556 500 556 556 278 556 556 222 222 500 222 833 556 556
+        556 556 333 500 278 556 500 722 500 500 500 334 260 334 584
+        U+00A0: 278 333 556 556 556 556 260 556 333 737 370 556 584 333 737 552
+        400 549 333 333 333 576 537 333 333 333 365 556 834 834 834 611
+        667 667 667 667 667 667 1000 722 667 667 667 667 278 278 278 278
+        722 722 778 778 778 778 778 584 778 722 722 722 722 667 667 611
+        556 556 556 556 556 556 889 500 556 556 556 556 278 278 278 278
+        556 556 556 556 556 556 556 549 611 556 556 556 556 500 556 500
+        667 556 667 556 667 556 722 500 722 500 722 500 722 500 722 615
+        722 556 667 556 667 556 667 556 667 556 667 556 778 556 778 556
+        778 556 778 556 722 556 722 556 278 278 278 278 278 278 278 222
+        278 278 735 444 500 222 667 500 500 556 222 556 222 556 292 556
+        334 556 222 722 556 722 556 722 556 604 723 556 778 556 778 556
+        778 556 1000 944 722 333 722 333 722 333 667 500 667 500 667 500
+        667 500 611 278 611 375 611 278 722 556 722 556 722 556 722 556
+        722 556 722 556 944 722 667 500 667 611 500 611 500 611 500 222
+        556 758 656 556 656 556 722 722 500 722 810 656 556 557 667 752
+        604 611 556 778 624 881 222 278 667 500 222 500 891 722 556 778
+        857 656 868 667 754 556 667 667 500 618 380 278 611 278 611 854
+        669 748 722 772 500 611 500 611 611 545 545 556 556 458 487 556
+        260 413 584 278 1333 1222 1049 1062 833 451 1222 944 771 667 556 278
+        222 778 556 722 556 722 556 722 556 722 556 722 556 556 667 556
+        667 556 1000 889 778 556 778 556 667 500 778 556 778 556 611 545
+        222 1333 1222 1049 778 556 1034 618 722 556 667 556 1000 889 778 611
+        667 556 667 556 667 556 667 556 278 278 278 278 778 556 778 556
+        722 333 722 333 722 556 722 556 667 500 611 278 545 437 722 556
+        706 682 604 565 611 500 667 556 667 556 778 556 778 556 778 556
+        778 556 667 500 349 685 367 222 887 887 667 722 500 556 611 500
+        500 581 453 667 722 668 667 556 500 222 737 556 722 333 667 500
+        U+0374: 333 333
+        U+037A: 333 500 500 500 278
+        U+0384: 333 333 667 278 784 838 384
+        U+038C: 774
+        U+038E: 855 752 222 667 667 551 668 667 611 722 778 278 667 668 833 722
+        650 778 722 667
+        U+03A3: 618 611 667 798 667 835 748 278 667 578 446 556 222 547 578 575
+        500 557 446 441 556 556 222 500 500 576 500 448 556 690 569 482
+        617 395 547 648 525 713 781 222 547 556 547 781
+        U+03D0: 575 547 772 958 772 560 781 601 778 556 722 500 611 404 625 529
+        756 577 891 833 674 556 674 500 667 667 609 596 737 554 464 410
+        601 573 500 222 778 442 442 667 556 722 833 688 569 722 722 722
+        667 667 865 542 719 667 278 278 500 1057 1010 854 583 719 635 719
+        667 656 667 542 677 667 923 604 719 719 583 656 833 722 778 719
+        667 722 611 635 760 667 740 667 917 938 792 885 656 719 1010 722
+        556 573 531 365 583 556 669 458 559 559 438 583 688 552 556 542
+        556 500 458 500 823 500 573 521 802 823 625 719 521 510 750 542
+        556 556 556 365 510 500 222 278 222 906 813 556 438 559 500 552
+        1338 624 778 613 950 713 668 500 897 695 829 685 1053 867 604 458
+        796 688 778 556 803 631 803 631
+        U+047A: 833 612 1191 852 1338 624 722 500 503 0 0 0 0 0 0 0
+        719 559 656 521 667 556 489 411 542 365 670 549 923 669 604 458
+        583 438 583 438 583 438 742 536 722 552 879 648 1137 870 753 521
+        722 500 611 458 556 500 556 500 667 500 925 691 667 521 667 521
+        667 556 861 666 861 666 278 923 669 667 551 656 583 722 552 722
+        552 667 521 833 688 222 667 556 667 556 1000 889 667 556 752 556
+        752 556 923 669 604 458 604 545 719 559 719 559 778 556 778 556
+        778 556 719 510 635 500 635 500 635 500 667 521 542 365 885 719
+        542 365 667 500 667 500
+        U+2000: 500 1000 500 1000 333 250 167 556 278 200 83
+        U+2010: 333
+        U+2012: 556 556 1000 1000 413 552 222 222 222 222 333 333 333 333 556 556
+        350
+        U+2026: 1000
+        U+202F: 200 1000
+        U+2032: 188 354 354
+        U+2039: 333 333
+        U+203C: 500
+        U+203E: 333
+        U+2044: 167
+        U+205E: 278
+        U+20A0: 556 722 722 556 556 833 722 1094 1164 944
+        U+20AB: 513 556 667 611 1000 521 667 778 667 556 722
+        U+20BF: 740
+        U+2105: 885
+        U+2113: 323
+        U+2116: 1073 737
+        U+2122: 1000
+        U+2126: 768
+        U+212E: 600
+        U+214D: 943 490
+        U+2190: 1000 500 1000 500 1000 500
+        U+21A8: 500
+        U+21D4: 1000
+      RUNS
+
+      # codepoint => advance in per-mille em
+      TABLE = begin
+        codepoint = 0
+        DATA.each_line.with_object({}) do |line, table|
+          start, advances =
+            line.include?(":") ? line.split(":", 2) : [nil, line]
+          codepoint = start.delete_prefix("U+").to_i(16) if start
+          advances.split.each do |advance|
+            table[codepoint] = advance.to_i
+            codepoint += 1
+          end
+        end.freeze
+      end
+    end
+  end
+end

data/lib/etassia/text_measurement/dejavu_advances.rb

@@ -0,0 +1,111 @@
+# frozen_string_literal: true
+
+# GENERATED by scripts/generate_dejavu_advances.rb; do not edit by hand.
+#
+# Horizontal advance widths of DejaVu Sans Regular and Bold, in design
+# units of a 2048-unit em, for Basic Latin and Latin-1 through Latin
+# Extended-B. Advances are numeric facts about a glyph, not a copy of any
+# font. Same layout as ArialAdvances::DATA; REGULAR and BOLD hold them
+# in per-mille em.
+module Etassia
+  class TextMeasurement
+    module DejaVuAdvances
+      REGULAR_DATA = <<~RUNS
+        U+0020: 651 821 942 1716 1303 1946 1597 563 799 799 1024 1716 651 739 651 690
+        1303 1303 1303 1303 1303 1303 1303 1303 1303 1303 690 690 1716 1716 1716 1087
+        2048 1401 1405 1430 1577 1294 1178 1587 1540 604 604 1343 1141 1767 1532 1612
+        1235 1612 1423 1300 1251 1499 1401 2025 1403 1251 1403 799 690 799 1716 1024
+        1024 1255 1300 1126 1300 1260 721 1300 1298 569 569 1186 569 1995 1298 1253
+        1300 1300 842 1067 803 1298 1212 1675 1212 1212 1075 1303 690 1303 1716
+        U+00A0: 651 821 1303 1303 1303 1303 690 1024 1024 2048 965 1253 1716 739 2048 1024
+        1024 1716 821 821 1024 1303 1303 651 1024 821 965 1253 1985 1985 1985 1087
+        1401 1401 1401 1401 1401 1401 1995 1430 1294 1294 1294 1294 604 604 604 604
+        1587 1532 1612 1612 1612 1612 1612 1716 1612 1499 1499 1499 1499 1251 1239 1290
+        1255 1255 1255 1255 1255 1255 2011 1126 1260 1260 1260 1260 569 569 569 569
+        1253 1298 1253 1253 1253 1253 1253 1716 1253 1298 1298 1298 1298 1212 1300 1212
+        1401 1255 1401 1255 1401 1255 1430 1126 1430 1126 1430 1126 1430 1126 1577 1300
+        1587 1300 1294 1260 1294 1260 1294 1260 1294 1260 1294 1260 1587 1300 1587 1300
+        1587 1300 1587 1300 1540 1298 1876 1423 604 569 604 569 604 569 604 569
+        604 569 1208 1138 604 569 1343 1186 1186 1141 569 1141 569 1141 768 1141
+        700 1151 582 1532 1298 1532 1298 1532 1298 1666 1532 1298 1612 1253 1612 1253
+        1612 1253 2191 2095 1423 842 1423 842 1423 842 1300 1067 1300 1067 1300 1067
+        1300 1067 1251 803 1251 803 1251 803 1499 1298 1499 1298 1499 1298 1499 1298
+        1499 1298 1499 1298 2025 1675 1251 1212 1251 1403 1075 1403 1075 1403 1075 721
+        1300 1505 1405 1300 1405 1300 1440 1430 1126 1587 1677 1405 1300 1253 1294 1612
+        1258 1178 721 1587 1406 2015 724 604 1527 1186 569 1212 1995 1532 1298 1612
+        1870 1253 1943 1555 1335 1300 1423 1300 1067 1294 688 803 1251 803 1251 1757
+        1298 1565 1476 1523 1496 1403 1075 1364 1364 1183 1075 1303 1364 1183 1045 1300
+        604 1008 940 605 2912 2660 2364 1711 1611 935 1907 1892 1633 1401 1255 604
+        569 1612 1253 1499 1298 1499 1298 1499 1298 1499 1298 1499 1298 1260 1401 1255
+        1401 1255 1995 2011 1587 1300 1587 1300 1343 1186 1612 1253 1612 1253 1364 1183
+        569 2912 2660 2364 1587 1300 2279 1397 1532 1298 1401 1255 1995 2011 1612 1253
+        1401 1255 1401 1255 1294 1260 1294 1260 604 569 604 569 1612 1253 1612 1253
+        1423 842 1423 842 1499 1298 1499 1298 1300 1067 1251 803 1284 1068 1540 1298
+        1506 1716 1430 1250 1403 1075 1401 1255 1294 1260 1612 1253 1612 1253 1612 1253
+        1612 1253 1251 1212 972 1726 977 569 2044 2044 1401 1430 1126 1141 1251 1067
+        1075 1235 981 1405 1499 1401 1294 1260 604 569 1600 1300 1423 842 1251 1212
+      RUNS
+
+      BOLD_DATA = <<~RUNS
+        U+0020: 713 934 1067 1716 1425 2052 1786 627 936 936 1071 1716 778 850 778 748
+        1425 1425 1425 1425 1425 1425 1425 1425 1425 1425 819 819 1716 1716 1716 1188
+        2048 1585 1561 1503 1700 1399 1399 1681 1714 762 762 1587 1305 2038 1714 1741
+        1501 1741 1577 1475 1397 1663 1585 2259 1579 1483 1485 936 748 936 1716 1024
+        1024 1382 1466 1214 1466 1389 891 1466 1458 702 702 1362 702 2134 1458 1407
+        1466 1466 1010 1219 979 1458 1335 1892 1321 1335 1192 1458 748 1458 1716
+        U+00A0: 713 934 1425 1425 1303 1425 748 1024 1024 2048 1155 1323 1716 850 2048 1024
+        1024 1716 897 897 1024 1507 1303 778 1024 897 1155 1323 2120 2120 2120 1188
+        1585 1585 1585 1585 1585 1585 2222 1503 1399 1399 1399 1399 762 762 762 762
+        1716 1714 1741 1741 1741 1741 1741 1716 1741 1663 1663 1663 1663 1483 1511 1473
+        1382 1382 1382 1382 1382 1382 2146 1214 1389 1389 1389 1389 702 702 702 702
+        1407 1458 1407 1407 1407 1407 1407 1716 1407 1458 1458 1458 1458 1335 1466 1335
+        1585 1382 1585 1382 1585 1382 1503 1214 1503 1214 1503 1214 1503 1214 1700 1466
+        1716 1466 1399 1389 1399 1389 1399 1389 1399 1389 1399 1389 1681 1466 1681 1466
+        1681 1466 1681 1466 1714 1458 1994 1618 762 702 762 702 762 702 762 702
+        762 702 1524 1404 762 702 1587 1362 1362 1305 702 1305 702 1305 982 1305
+        1140 1315 760 1714 1458 1714 1458 1714 1458 2013 1714 1458 1741 1407 1741 1407
+        1741 1407 2390 2241 1577 1010 1577 1010 1577 1010 1475 1219 1475 1219 1475 1219
+        1475 1219 1397 979 1397 979 1397 979 1663 1458 1663 1458 1663 1458 1663 1458
+        1663 1458 1663 1458 2259 1892 1483 1335 1483 1485 1192 1485 1192 1485 1192 891
+        1466 1661 1561 1466 1561 1466 1503 1503 1214 1716 1800 1550 1466 1408 1399 1739
+        1425 1399 891 1681 1624 2140 892 797 1587 1362 738 1212 2134 1714 1458 1741
+        1789 1407 2217 1868 1601 1466 1577 1475 1219 1399 1130 979 1447 979 1397 1711
+        1458 1741 1666 1633 1594 1485 1192 1582 1582 1312 1192 1425 1582 1312 1173 1466
+        762 1349 1114 762 3185 2892 2658 2067 2007 1404 2476 2416 2160 1585 1382 762
+        702 1741 1407 1663 1458 1663 1458 1663 1458 1663 1458 1663 1458 1389 1585 1382
+        1585 1382 2222 2146 1681 1466 1681 1466 1587 1362 1741 1407 1741 1407 1582 1192
+        702 3185 2892 2658 1681 1466 2639 1612 1714 1458 1585 1382 2222 2146 1741 1407
+        1585 1382 1585 1382 1399 1389 1399 1389 762 702 762 702 1741 1407 1741 1407
+        1577 1010 1577 1010 1663 1458 1663 1458 1475 1219 1397 979 1414 1244 1714 1458
+        1714 1771 1657 1349 1485 1192 1585 1382 1399 1389 1741 1407 1741 1407 1741 1407
+        1741 1407 1483 1335 1007 1775 1048 702 2228 2228 1585 1503 1214 1305 1397 1219
+        1192 1601 1258 1561 1663 1585 1399 1389 762 702 1762 1620 1577 1010 1483 1335
+      RUNS
+
+      UNITS_PER_EM = 2048.0
+
+      # codepoint => advance in per-mille em
+      def self.parse(data)
+        codepoint = 0
+        data.each_line.with_object({}) do |line, table|
+          start, values = split(line)
+          codepoint = start.delete_prefix("U+").to_i(16) if start
+          values.each do |value|
+            table[codepoint] = value.to_i * 1000 / UNITS_PER_EM
+            codepoint += 1
+          end
+        end.freeze
+      end
+
+      # A run's marker, if the line starts one, and its advances.
+      def self.split(line)
+        start, values = line.include?(":") ? line.split(":", 2) : [nil, line]
+        [start, values.split]
+      end
+
+      REGULAR = parse(REGULAR_DATA)
+      BOLD = parse(BOLD_DATA)
+      private_class_method :parse, :split
+    end
+  end
+end

data/lib/etassia/text_measurement/times_advances.rb

@@ -0,0 +1,64 @@
+# frozen_string_literal: true
+
+# GENERATED by scripts/generate_times_advances.rb; do not edit by hand.
+#
+# Horizontal advance widths in per-mille em of Times New Roman Regular,
+# for Basic Latin, Latin-1 through Latin Extended-B and U+FB02. Advances
+# are numeric facts about a glyph, not a copy of any font.
+module Etassia
+  class TextMeasurement
+    module TimesAdvances
+      # Same layout as ArialAdvances::DATA.
+      DATA = <<~RUNS
+        U+0020: 250 333 408 500 500 833 778 180 333 333 500 564 250 333 250 278
+        500 500 500 500 500 500 500 500 500 500 278 278 564 564 564 444
+        921 722 667 667 722 611 556 722 722 333 389 722 611 889 722 722
+        556 722 667 556 611 722 722 944 722 722 611 333 278 333 469 500
+        333 444 500 444 500 444 333 500 500 278 278 500 278 778 500 500
+        500 500 333 389 278 500 500 722 500 500 444 480 200 480 541
+        U+00A0: 250 333 500 500 500 500 200 500 333 760 276 500 564 333 760 500
+        400 549 300 300 333 576 453 333 333 300 310 500 750 750 750 444
+        722 722 722 722 722 722 889 667 611 611 611 611 333 333 333 333
+        722 722 722 722 722 722 722 564 722 722 722 722 722 722 556 500
+        444 444 444 444 444 444 667 444 444 444 444 444 278 278 278 278
+        500 500 500 500 500 500 500 549 500 500 500 500 500 500 500 500
+        722 444 722 444 722 444 667 444 667 444 667 444 667 444 722 646
+        722 500 611 444 611 444 611 444 611 444 611 444 722 500 722 500
+        722 500 722 500 722 500 722 500 333 278 333 278 333 278 333 278
+        333 278 708 552 389 278 722 500 500 611 278 611 278 611 406 611
+        344 611 278 722 500 722 500 722 500 604 702 495 722 500 722 500
+        722 500 889 722 667 333 667 333 667 333 556 389 556 389 556 389
+        556 389 611 278 611 427 611 278 722 500 722 500 722 500 722 500
+        722 500 722 500 944 722 722 500 722 611 444 611 444 611 444 278
+        500 760 574 500 574 500 667 667 444 722 815 574 500 471 611 722
+        501 556 500 722 722 770 253 333 722 500 278 485 816 722 500 722
+        722 529 925 691 650 500 556 556 389 582 344 278 611 278 611 776
+        542 743 722 781 500 611 444 539 539 444 444 500 500 443 422 500
+        200 281 249 333 1333 1166 944 1000 889 556 1111 1000 778 722 444 333
+        278 722 500 722 500 722 500 722 500 722 500 722 500 444 722 444
+        722 444 889 667 722 500 722 500 722 500 722 500 722 500 539 444
+        278 1333 1166 944 722 500 950 560 722 500 722 444 889 667 722 500
+        722 444 722 444 611 444 611 444 333 278 333 278 722 500 722 500
+        667 333 667 333 722 500 722 500 556 389 611 278 563 395 722 500
+        650 500 604 500 611 444 722 444 611 444 722 500 722 500 722 500
+        722 500 722 500 278 500 317 278 774 772 722 667 500 611 611 389
+        444 534 397 667 722 725 611 444 389 278 702 500 667 333 722 500
+        U+FB02: 556
+      RUNS
+
+      # codepoint => advance in per-mille em
+      TABLE = begin
+        codepoint = 0
+        DATA.each_line.with_object({}) do |line, table|
+          start, values =
+            line.include?(":") ? line.split(":", 2) : [nil, line]
+          codepoint = start.delete_prefix("U+").to_i(16) if start
+          values.split.each do |value|
+            table[codepoint] = value.to_i
+            codepoint += 1
+          end
+        end.freeze
+      end
+    end
+  end
+end

data/lib/etassia/text_measurement/trebuchet_advances.rb

@@ -0,0 +1,53 @@
+# frozen_string_literal: true
+
+# GENERATED by scripts/generate_trebuchet_advances.rb; do not edit by hand.
+#
+# Horizontal advance widths in per-mille em of Trebuchet MS Regular,
+# for Basic Latin and Latin-1 through Latin Extended-B. Advances are
+# numeric facts about a glyph, not a copy of any font.
+module Etassia
+  class TextMeasurement
+    module TrebuchetAdvances
+      # Same layout as ArialAdvances::DATA.
+      DATA = <<~RUNS
+        U+0020: 301 367 325 524 524 600 706 160 367 367 367 524 367 367 367 524
+        524 524 524 524 524 524 524 524 524 524 367 367 524 524 524 367
+        771 590 566 598 613 536 525 676 654 278 477 576 506 709 638 674
+        558 676 582 481 581 648 587 852 557 570 550 367 355 367 524 524
+        524 525 557 495 557 545 370 502 546 285 367 504 295 830 546 537
+        557 557 389 405 396 546 490 744 501 493 475 367 524 367 524
+        U+00A0: 301 367 524 524 524 570 524 454 524 713 367 524 524 367 713 524
+        524 524 451 454 524 546 524 367 524 451 367 524 814 814 814 367
+        590 590 590 590 590 590 867 598 536 536 536 536 278 278 278 278
+        613 638 674 674 674 674 674 524 657 648 648 648 648 570 556 546
+        525 525 525 525 525 525 873 495 545 545 545 545 285 285 285 285
+        549 546 537 537 537 537 537 524 545 546 546 546 546 493 553 493
+        590 525 590 525 590 525 598 495 598 495 598 495 598 495 613 691
+        613 557 536 545 536 545 536 545 536 545 536 545 676 502 676 502
+        676 502 676 502 654 546 681 548 278 285 278 285 278 285 278 285
+        278 285 727 602 477 367 576 504 545 506 295 506 295 506 295 506
+        330 506 298 638 546 638 546 638 546 635 653 546 674 537 674 537
+        674 537 993 924 582 389 582 389 582 389 481 405 481 405 481 405
+        481 405 581 396 581 496 581 396 648 546 648 546 648 546 648 546
+        648 546 648 546 852 744 570 493 570 550 475 550 475 550 475 331
+        U+0192: 388
+        U+01FA: 590 525 867 873 657 545
+        U+0218: 481 405 581 396
+      RUNS
+
+      # codepoint => advance in per-mille em
+      TABLE = begin
+        codepoint = 0
+        DATA.each_line.with_object({}) do |line, table|
+          start, values =
+            line.include?(":") ? line.split(":", 2) : [nil, line]
+          codepoint = start.delete_prefix("U+").to_i(16) if start
+          values.split.each do |value|
+            table[codepoint] = value.to_i
+            codepoint += 1
+          end
+        end.freeze
+      end
+    end
+  end
+end

data/lib/etassia/text_measurement/trebuchet_kerning.rb

@@ -0,0 +1,29 @@
+# frozen_string_literal: true
+
+module Etassia
+  class TextMeasurement
+    # Trebuchet MS kern adjustments used by the remaining Mindmap corpus.
+    # Values are design units from the regular face's kern table.
+    module TrebuchetKerning
+      UNITS_PER_EM = 2048.0
+      PAIRS = {
+        " A" => -113,
+        " T" => -37,
+        "Pe" => -96,
+        "Po" => -96,
+        "Re" => -83,
+        "Ro" => -83,
+        "To" => -255,
+      }.freeze
+
+      module_function
+
+      def adjustment(text, font_size:)
+        units = text.each_char.each_cons(2).sum do |pair|
+          PAIRS.fetch(pair.join, 0)
+        end
+        units * font_size / UNITS_PER_EM
+      end
+    end
+  end
+end

data/lib/etassia/text_measurement.rb

@@ -0,0 +1,198 @@
+# frozen_string_literal: true
+
+require_relative "text_measurement/arial_advances"
+require_relative "text_measurement/dejavu_advances"
+require_relative "text_measurement/trebuchet_advances"
+
+module Etassia
+  # Estimates text dimensions for layout, in pure Ruby: no font file is read
+  # at runtime. Width is the sum of per-glyph advances of the default theme's
+  # stack ("Arial, Helvetica, sans-serif"), from the generated ArialAdvances.
+  #
+  # BOUNDS: every codepoint in ArialAdvances (exact in Liberation Sans, and in
+  # Arial wherever Arial has the glyph; macOS Arial lacks U+0487, U+2010,
+  # U+202F, U+20BF and U+21D4); CJK and fullwidth at 1.0 em, the four Arabic
+  # ligatures in OUTLIERS, and zero-width marks (all measured in Chrome).
+  # ESTIMATES: emoji, and any pictographic character followed by U+FE0F, at
+  # 1.5 em; everything else at 1.0 em. A ZWJ sequence counts each emoji.
+  # DOES NOT BOUND: a font the viewer substitutes for a glyph Arial lacks,
+  # other stacks (DejaVu, Verdana run up to 1.19 times wider), bold, italic,
+  # kerning (it can widen a pair as well as narrow it), shaping beyond
+  # OUTLIERS. Callers that need room use Renderer::Base#reserved_text_width.
+  #
+  # @example Basic usage
+  #   TextMeasurement.measure("Hello", font_size: 14)
+  #   # => { width: 31.892, height: 14.0 }
+  #
+  # @example With dimension overrides
+  #   TextMeasurement.measure("Text", font_size: 12, width: 100, height: 20)
+  #   # => { width: 100, height: 20 }
+  class TextMeasurement
+    # Text height as ratio of font size
+    HEIGHT_RATIO = 1.0
+
+    # Per-mille em for a codepoint no rule below knows.
+    DEFAULT_ADVANCE = 1000
+    EMOJI_ADVANCE = 1500
+    MONOSPACE_ADVANCE = 600
+    # DejaVu Sans Mono: 1233 units of 2048 per glyph.
+    DEJAVU_MONO_ADVANCE = 1233 * 1000.0 / 2048
+    SPACE_ADVANCE = ArialAdvances::TABLE.fetch(0x20)
+
+    # SVG draws one <text> as ONE line: a tab, newline or carriage return is
+    # collapsed to a space. A CRLF is counted twice (an over-estimate).
+    SVG_WHITESPACE = /[\t\n\r]/
+    ASCII_WHITESPACE = [9, 10, 13].freeze
+
+    # Codepoints whose width comes from ligature substitution, so no
+    # per-character rule describes them. Chrome measured 1.234, 1.025, 1.221
+    # and 6.489 em; these round up.
+    OUTLIERS = {
+      0xFDFA => 1250, 0xFDFB => 1050, 0xFDFC => 1250, 0xFDFD => 6500
+    }.freeze
+
+    # The faces a caller can ask for. A codepoint a face lacks falls back to
+    # the default one.
+    FACE_TABLES = {
+      arial: ArialAdvances::TABLE,
+      trebuchet: TrebuchetAdvances::TABLE,
+      dejavu: DejaVuAdvances::REGULAR,
+      dejavu_bold: DejaVuAdvances::BOLD,
+    }.freeze
+
+    ZERO_WIDTH = /[\p{Mn}\p{Me}\p{Cf}\p{Cc}]/
+    EMOJI = /\p{Extended_Pictographic}/
+    # Ruby 4 moved to Unicode 17, where U+266D no longer matches
+    # Extended_Pictographic. Keep the measured width independent of the
+    # Unicode tables bundled with the running Ruby.
+    PINNED_PICTOGRAPHS = [0x266D].freeze
+
+    # A pictographic character plus U+FE0F asks for the emoji glyph, even for
+    # one the table has as text (U+00A9 is 0.737 em as text, 1.0 em as emoji
+    # in Chrome). Any other character is a cluster of its own.
+    CLUSTER = /#{EMOJI}\uFE0F|./m
+
+    # Measures the approximate dimensions of text.
+    #
+    # @param text [String] the text to measure, as one line: a caller that
+    #   draws several lines measures each line on its own
+    # @param font_size [Numeric] the font size in points
+    # @param width [Numeric, nil] optional width override
+    # @param height [Numeric, nil] optional height override
+    # @param monospace [Boolean] measure for a monospace font-family
+    # @return [Hash] hash with :width and :height keys
+    def self.measure(text, font_size:, width: nil, height: nil,
+                     monospace: false)
+      {
+        width: width || calculate_width(text, font_size, monospace, :arial),
+        height: height || calculate_height(font_size),
+      }
+    end
+
+    # Measures text in a face other than the default, with the default
+    # standing in for a codepoint the face lacks.
+    #
+    # @param text [String] the text to measure, as one line
+    # @param font_size [Numeric] the font size in points
+    # @param face [Symbol] :arial, :trebuchet, :dejavu or :dejavu_bold
+    # @return [Hash] hash with :width and :height keys
+    def self.measure_in(face, text, font_size:)
+      {
+        width: calculate_width(text, font_size, false, face),
+        height: calculate_height(font_size),
+      }
+    end
+
+    # Measures text in the `monospace` family PlantUML's JAR resolves to,
+    # DejaVu Sans Mono: every glyph advances 1233/2048 em (a combining mark
+    # or control character advances nothing).
+    #
+    # @param text [String] the text to measure, as one line
+    # @param font_size [Numeric] the font size in points
+    # @return [Hash] hash with :width and :height keys
+    def self.measure_mono(text, font_size:)
+      cells = normalize(text).scan(CLUSTER).count do |cluster|
+        !cluster.match?(ZERO_WIDTH)
+      end
+      { width: cells * DEJAVU_MONO_ADVANCE * font_size / 1000.0,
+        height: calculate_height(font_size) }
+    end
+
+    # @return [Float] the width of the text drawn as one line
+    def self.calculate_width(text, font_size, monospace, face)
+      normalized = normalize(text)
+      if normalized.ascii_only?
+        return ascii_width(normalized, monospace, face) * font_size / 1000.0
+      end
+
+      total = normalized.scan(CLUSTER).sum do |cluster|
+        cluster.length == 2 ? EMOJI_ADVANCE : advance(cluster, monospace, face)
+      end
+      total * font_size / 1000.0
+    end
+
+    def self.ascii_width(text, monospace, face)
+      text.each_byte.sum { |code| ascii_advance(code, monospace, face) }
+    end
+
+    def self.ascii_advance(code, monospace, face)
+      return whitespace_advance(monospace) if ASCII_WHITESPACE.include?(code)
+      return 0 if code < 32 || code == 127
+
+      table = FACE_TABLES.fetch(face)[code] || ArialAdvances::TABLE[code]
+      table_advance(table, monospace)
+    end
+
+    # @return [Integer] the advance of one character in per-mille em
+    def self.advance(char, monospace, face)
+      return whitespace_advance(monospace) if char.match?(SVG_WHITESPACE)
+      return 0 if char.match?(ZERO_WIDTH)
+
+      code = char.ord
+      table = FACE_TABLES.fetch(face)[code] || ArialAdvances::TABLE[code]
+      return table_advance(table, monospace) if table
+
+      OUTLIERS.fetch(code) { fallback_advance(char, monospace) }
+    end
+
+    # Monospace keeps a zero advance (combining marks) at zero.
+    def self.table_advance(table, monospace)
+      monospace && table.positive? ? MONOSPACE_ADVANCE : table
+    end
+
+    def self.whitespace_advance(monospace)
+      monospace ? MONOSPACE_ADVANCE : SPACE_ADVANCE
+    end
+
+    def self.fallback_advance(char, monospace)
+      return EMOJI_ADVANCE if PINNED_PICTOGRAPHS.include?(char.ord) ||
+        char.match?(EMOJI)
+
+      monospace ? MONOSPACE_ADVANCE : DEFAULT_ADVANCE
+    end
+
+    # Never raises on an undecodable string: a layout must not fail on bytes.
+    # The final re-tag is needed for UTF-8 too: encode from CESU-8 and the
+    # UTF8-DoCoMo/KDDI/SoftBank family can leave a stray continuation byte
+    # that still reports valid, and only a re-scan of the bytes sees it.
+    def self.normalize(text)
+      string = text.to_s
+      unless string.encoding == Encoding::UTF_8
+        string = string.encode(Encoding::UTF_8, invalid: :replace,
+                                                undef: :replace)
+      end
+      string.b.force_encoding(Encoding::UTF_8).scrub
+    rescue EncodingError
+      string.b.encode(Encoding::UTF_8, invalid: :replace, undef: :replace)
+    end
+
+    # @return [Float] the calculated height
+    def self.calculate_height(font_size)
+      font_size * HEIGHT_RATIO
+    end
+
+    private_class_method :calculate_width, :calculate_height, :advance,
+                         :fallback_advance, :table_advance, :whitespace_advance,
+                         :normalize, :ascii_width, :ascii_advance
+  end
+end