Support absolute TeX units (#732)

* Support absolute TeX units

* Implement @kohler's comments

* Rewrite unit documentation according to @kohler's comments

README.md

@@ -66,6 +66,25 @@ katex.render("c = \\pm\\sqrt{a^2 + b^2}\\in\\RR", element, {
 
 Math on the page can be automatically rendered using the auto-render extension. See [the Auto-render README](contrib/auto-render/README.md) for more information.
 
+#### Font size and lengths
+
+By default, KaTeX math is rendered in a 1.21× larger font than the surrounding
+context, which makes super- and subscripts easier to read. You can control
+this using CSS, for example:
+
+```css
+.katex { font-size: 1.1em; }
+```
+
+KaTeX supports all TeX units, including absolute units like `cm` and `in`.
+Absolute units are currently scaled relative to the default TeX font size of
+10pt, so that `\kern1cm` produces the same results as `\kern2.845275em`.
+As a result, relative and absolute units are both uniformly scaled relative
+to LaTeX with a 10pt font; for example, the rectangle `\rule{1cm}{1em}` has
+the same aspect ratio in KaTeX as in LaTeX.  However, because most browsers
+default to a larger font size, this typically means that a 1cm kern in KaTeX
+will appear larger than 1cm in browser units.
+
 ## Contributing
 
 See [CONTRIBUTING.md](CONTRIBUTING.md)

src/Parser.js

@@ -4,6 +4,7 @@ import environments from "./environments";
 import MacroExpander from "./MacroExpander";
 import symbols from "./symbols";
 import utils from "./utils";
+import units from "./units";
 import { cjkRegex } from "./unicodeRegexes";
 import ParseNode from "./ParseNode";
 import ParseError from "./ParseError";
@@ -798,11 +799,11 @@ class Parser {
             number: +(match[1] + match[2]), // sign + magnitude, cast to number
             unit: match[3],
         };
-        if (data.unit !== "em" && data.unit !== "ex" && data.unit !== "mu") {
+        if (!units.validUnit(data)) {
             throw new ParseError("Invalid unit: '" + data.unit + "'", res);
         }
         return new ParseFuncOrArgument(
-            new ParseNode("color", data, this.mode),
+            new ParseNode("size", data, this.mode),
             false);
     }
 

src/buildHTML.js

@@ -12,6 +12,7 @@ import Style from "./Style";
 import buildCommon, { makeSpan } from "./buildCommon";
 import delimiter from "./delimiter";
 import domTree from "./domTree";
+import units from "./units";
 import utils from "./utils";
 import stretchy from "./stretchy";
 
@@ -545,43 +546,6 @@ groupTypes.genfrac = function(group, options) {
         options);
 };
 
-/**
- * Parse a `sizeValue`, as parsed by functions.js argType "size", into
- * a CSS em value. `options` gives the current options.
- */
-const calculateSize = function(sizeValue, options) {
-    let scale;
-    // `mu` units scale with scriptstyle/scriptscriptstyle.
-    // Other units always refer to the *textstyle* font in the current size.
-    if (sizeValue.unit === "mu") {
-        scale = options.fontMetrics().cssEmPerMu;
-    } else {
-        let unitOptions;
-        if (options.style.isTight()) {
-            // isTight() means current style is script/scriptscript.
-            unitOptions = options.havingStyle(options.style.text());
-        } else {
-            unitOptions = options;
-        }
-        // TODO: In TeX these units are relative to the quad of the current
-        // *text* font, e.g. cmr10. KaTeX instead uses values from the
-        // comparably-sized *Computer Modern symbol* font. At 10pt, these
-        // match. At 7pt and 5pt, they differ: cmr7=1.138894, cmsy7=1.170641;
-        // cmr5=1.361133, cmsy5=1.472241. Consider $\scriptsize a\kern1emb$.
-        // TeX \showlists shows a kern of 1.13889 * fontsize;
-        // KaTeX shows a kern of 1.171 * fontsize.
-        if (sizeValue.unit === "ex") {
-            scale = unitOptions.fontMetrics().xHeight;
-        } else {
-            scale = unitOptions.fontMetrics().quad;
-        }
-        if (unitOptions !== options) {
-            scale *= unitOptions.sizeMultiplier / options.sizeMultiplier;
-        }
-    }
-    return sizeValue.number * scale;
-};
-
 groupTypes.array = function(group, options) {
     let r;
     let c;
@@ -629,7 +593,7 @@ groupTypes.array = function(group, options) {
 
         let gap = 0;
         if (group.value.rowGaps[r]) {
-            gap = calculateSize(group.value.rowGaps[r].value, options);
+            gap = units.calculateSize(group.value.rowGaps[r].value, options);
             if (gap > 0) { // \@argarraycr
                 gap += arstrutDepth;
                 if (depth < gap) {
@@ -1330,11 +1294,11 @@ groupTypes.rule = function(group, options) {
     // Calculate the shift, width, and height of the rule, and account for units
     let shift = 0;
     if (group.value.shift) {
-        shift = calculateSize(group.value.shift, options);
+        shift = units.calculateSize(group.value.shift, options);
     }
 
-    const width = calculateSize(group.value.width, options);
-    const height = calculateSize(group.value.height, options);
+    const width = units.calculateSize(group.value.width, options);
+    const height = units.calculateSize(group.value.height, options);
 
     // Style the rule to the right size
     rule.style.borderRightWidth = width + "em";
@@ -1358,7 +1322,7 @@ groupTypes.kern = function(group, options) {
     const rule = makeSpan(["mord", "rule"], [], options);
 
     if (group.value.dimension) {
-        const dimension = calculateSize(group.value.dimension, options);
+        const dimension = units.calculateSize(group.value.dimension, options);
         rule.style.marginLeft = dimension + "em";
     }
 

src/macros.js

@@ -15,6 +15,10 @@ defineMacro("\\egroup", "}");
 defineMacro("\\begingroup", "{");
 defineMacro("\\endgroup", "}");
 
+// We don't distinguish between math and nonmath kerns.
+// (In TeX, the mu unit works only with \mkern.)
+defineMacro("\\mkern", "\\kern");
+
 //////////////////////////////////////////////////////////////////////
 // amsmath.sty
 

src/units.js

@@ -0,0 +1,99 @@
+/* eslint no-console:0 */
+
+/**
+ * This file does conversion between units.  In particular, it provides
+ * calculateSize to convert other units into ems.
+ */
+
+import ParseError from "./ParseError";
+
+// This table gives the number of TeX pts in one of each *absolute* TeX unit.
+// Thus, multiplying a length by this number converts the length from units
+// into pts.  Dividing the result by ptPerEm gives the number of ems
+// *assuming* a font size of ptPerEm (normal size, normal style).
+const ptPerUnit = {
+    // https://en.wikibooks.org/wiki/LaTeX/Lengths and
+    // https://tex.stackexchange.com/a/8263
+    "pt": 1,            // TeX point
+    "mm": 7227 / 2540,  // millimeter
+    "cm": 7227 / 254,   // centimeter
+    "in": 72.27,        // inch
+    "bp": 803 / 800,    // big (PostScript) points
+    "pc": 12,           // pica
+    "dd": 1238 / 1157,  // didot
+    "cc": 14856 / 1157, // cicero (12 didot)
+    "nd": 685 / 642,    // new didot
+    "nc": 1370 / 107,   // new cicero (12 new didot)
+    "sp": 1 / 65536,    // scaled point (TeX's internal smallest unit)
+    // https://tex.stackexchange.com/a/41371
+    "px": 803 / 800,    // \pdfpxdimen defaults to 1 bp in pdfTeX and LuaTeX
+};
+
+// Dictionary of relative units, for fast validity testing.
+const relativeUnit = {
+    "ex": true,
+    "em": true,
+    "mu": true,
+};
+
+/**
+ * Determine whether the specified unit (either a string defining the unit
+ * or a "size" parse node containing a unit field) is valid.
+ */
+const validUnit = function(unit) {
+    if (unit.unit) {
+        unit = unit.unit;
+    }
+    return (unit in ptPerUnit || unit in relativeUnit || unit === "ex");
+};
+
+/*
+ * Convert a "size" parse node (with numeric "number" and string "unit" fields,
+ * as parsed by functions.js argType "size") into a CSS em value for the
+ * current style/scale.  `options` gives the current options.
+ */
+const calculateSize = function(sizeValue, options) {
+    let scale;
+    if (sizeValue.unit in ptPerUnit) {
+        // Absolute units
+        scale = ptPerUnit[sizeValue.unit]   // Convert unit to pt
+           / options.fontMetrics().ptPerEm  // Convert pt to CSS em
+           / options.sizeMultiplier;        // Unscale to make absolute units
+    } else if (sizeValue.unit === "mu") {
+        // `mu` units scale with scriptstyle/scriptscriptstyle.
+        scale = options.fontMetrics().cssEmPerMu;
+    } else {
+        // Other relative units always refer to the *textstyle* font
+        // in the current size.
+        let unitOptions;
+        if (options.style.isTight()) {
+            // isTight() means current style is script/scriptscript.
+            unitOptions = options.havingStyle(options.style.text());
+        } else {
+            unitOptions = options;
+        }
+        // TODO: In TeX these units are relative to the quad of the current
+        // *text* font, e.g. cmr10. KaTeX instead uses values from the
+        // comparably-sized *Computer Modern symbol* font. At 10pt, these
+        // match. At 7pt and 5pt, they differ: cmr7=1.138894, cmsy7=1.170641;
+        // cmr5=1.361133, cmsy5=1.472241. Consider $\scriptsize a\kern1emb$.
+        // TeX \showlists shows a kern of 1.13889 * fontsize;
+        // KaTeX shows a kern of 1.171 * fontsize.
+        if (sizeValue.unit === "ex") {
+            scale = unitOptions.fontMetrics().xHeight;
+        } else if (sizeValue.unit === "em") {
+            scale = unitOptions.fontMetrics().quad;
+        } else {
+            throw new ParseError("Invalid unit: '" + sizeValue.unit + "'");
+        }
+        if (unitOptions !== options) {
+            scale *= unitOptions.sizeMultiplier / options.sizeMultiplier;
+        }
+    }
+    return sizeValue.number * scale;
+};
+
+module.exports = {
+    validUnit: validUnit,
+    calculateSize: calculateSize,
+};

test/katex-spec.js

@@ -931,7 +931,7 @@ describe("An overline parser", function() {
 describe("A rule parser", function() {
     const emRule = "\\rule{1em}{2em}";
     const exRule = "\\rule{1ex}{2em}";
-    const badUnitRule = "\\rule{1px}{2em}";
+    const badUnitRule = "\\rule{1au}{2em}";
     const noNumberRule = "\\rule{1em}{em}";
     const incompleteRule = "\\rule{1em}";
     const hardNumberRule = "\\rule{   01.24ex}{2.450   em   }";
@@ -988,7 +988,7 @@ describe("A kern parser", function() {
     const exKern = "\\kern{1ex}";
     const muKern = "\\kern{1mu}";
     const abKern = "a\\kern{1em}b";
-    const badUnitRule = "\\kern{1px}";
+    const badUnitRule = "\\kern{1au}";
     const noNumberRule = "\\kern{em}";
 
     it("should list the correct units", function() {
@@ -1026,7 +1026,7 @@ describe("A non-braced kern parser", function() {
     const abKern1 = "a\\mkern1mub";
     const abKern2 = "a\\kern-1mub";
     const abKern3 = "a\\kern-1mu b";
-    const badUnitRule = "\\kern1px";
+    const badUnitRule = "\\kern1au";
     const noNumberRule = "\\kern em";
 
     it("should list the correct units", function() {

test/screenshotter/ss_data.yaml

@@ -223,6 +223,21 @@ Symbols1: |
 Text: \frac{a}{b}\text{c~ {ab} \ e}+fg
 TextWithMath: \text{for $a < b$ and $ c < d $}.
 Unicode: \begin{matrix}\text{ÀàÇçÉéÏïÖöÛû} \\ \text{БГДЖЗЙЛФЦШЫЮЯ} \\ \text{여보세요} \\ \text{私はバナナです} \end{matrix}
+Units: |
+    \begin{array}{ll}
+      \mathrm H\kern 1em\mathrm H \text{\tiny (1em)}
+      & \mathrm H\kern 1ex\mathrm H \text{\tiny (1ex)} \\
+      \mathrm H{\scriptstyle \kern 1em}\mathrm H \text{\tiny (ss 1em)}
+      & \mathrm H{\scriptstyle \kern 1ex}\mathrm H \text{\tiny (ss 1ex)} \\
+      \mathrm H{\small \kern 1em}\mathrm H \text{\tiny (sm 1em)}
+      & \mathrm H{\small \kern 1ex}\mathrm H \text{\tiny (sm 1ex)} \\
+      \mathrm H\mkern 18mu\mathrm H \text{\tiny (18mu)}
+      & \mathrm H\kern 1cm\mathrm H \text{\tiny (1cm)} \\
+      \mathrm H{\scriptstyle \mkern 18mu}\mathrm H \text{\tiny (ss 18mu)}
+      & \mathrm H{\scriptstyle \kern 1cm}\mathrm H \text{\tiny (ss 1cm)} \\
+      \mathrm H{\small \mkern 18mu}\mathrm H \text{\tiny (sm 18mu)}
+      & \mathrm H{\small \kern 1cm}\mathrm H \text{\tiny (sm 1cm)}
+    \end{array}
 UnsupportedCmds:
     tex: \err\,\frac\fracerr3\,2^\superr_\suberr\,\sqrt\sqrterr
     noThrow: 1