camlmath
tex math to mathml core converter
Star 1
Fork 0
Watch
37 1 3
Files
.github
lib
test
.envrc
.gitignore
.mailmap
.ocamlformat
atelier.toml
camlmath.opam
changes.md
default.nix
dune-project
flake.lock
flake.nix
license.txt
readme.md
readme.md

camlmath #

Binary Cache:

Converts a subset of TeX math to MathML Core markup.

Put this in dune:

(libraries camlmath)

API:

Camlmath.to_mathml : ?display:display -> string -> (string, error) result
Camlmath.parse     : string -> (Ast.node, error) result

to_mathml_exn raises Camlmath_error instead. The argument is the body of a math block with the delimiters already stripped, and display defaults to Block.

Camlmath.to_mathml_exn "\\frac{a}{b}"
(* <math xmlns="..." display="block"><mfrac><mi>a</mi><mi>b</mi></mfrac></math> *)

Supported syntax #

  • \frac{a}{b}, \xrightarrow{label}, \xleftarrow{label}. The optional argument that sets a second label under an arrow is refused rather than misread.
  • \text{...} and \texttt{...}, in text mode where spaces are significant. An unescaped brace groups and vanishes as in TeX, \{ \} \$ \% \& \# \_ are literal, \ is a space and \textbackslash a backslash.
  • {...} groups, _ and ^ in either order, and a single unbraced token as an argument to \frac, the arrows, _ and ^ alike, so \frac12 is one half and x^12 is x to the first followed by a 2, as TeX reads it
  • primes: f' is f^{\prime}, f'' uses the double prime glyph, and a ^ right after a prime joins the same superscript, so f'^2 is f^{\prime 2}
  • digit runs as one <mn>, letters as <mi>
  • ( ) [ ] | pinned against stretching, + - *, and = < > / , ; ! ? .
  • : as the relation U+2236, which the operator dictionary spaces the way TeX spaces a relation, and \colon as the punctuation colon
  • a sign after a relation, an opening fence or punctuation is wrapped with its operand, so x = -1 has no gap after the minus, which is TeX's rule for a binary operator in that position
  • \, \: \; \quad \qquad and the control space
  • \{ \} \vert for a single bar, \| and \Vert for a double one
  • \% \& \# \$ \_, the characters TeX reserves, set as identifiers because TeX gives them no operator spacing
  • \cup \cap \setminus \in \notin \subset \subseteq \times \cdot \pm
  • \emptyset \infty \forall \exists \lnot, ordinary symbols set as identifiers for the same reason
  • \leq \geq \neq with their \le \ge \ne spellings, \equiv \approx \land \lor \vdash \models \ldots \cdots
  • \rightarrow \to \leftarrow \leftrightarrow \Rightarrow \mapsto
  • Greek letters, lowercase with the \var forms, and uppercase set upright the way TeX sets it

Camlmath.Parser.commands returns the same list at runtime.

Failure #

Anything outside the subset will result in an error.

type error =
  | Invalid_utf8 of int
  | Invalid_char of int
  | Unknown_command of string * int
  | Unexpected_char of string * int
  | Unexpected_token of string * int
  | Unclosed_group of int
  | Missing_argument of string * int

Every variant carries a byte offset, so \sqrt{2} reports unknown_command(\sqrt at 0). A character outside printable ASCII is named by its code point, so a typed alpha reports unexpected_char(U+03B1 at 0), and the error strings are ASCII like the markup.

Invalid_char covers what XML cannot carry at all, as a literal or as a reference: most C0 controls, and U+FFFE and U+FFFF. Valid UTF-8 is a wider set than valid XML, so checking the encoding alone is not enough to keep the output parseable. Parser.parse runs both checks itself, so the tree it returns can always be emitted.