Wolfram/Parser | Paclet Repository

Wolfram`Parser`

EBNFParse

EBNFParse [source]
	reads a BNF / EBNF grammar from the string source and returns an Association mapping each rule name to the ParserCombinator it lowers to.

EBNFParse [ File [path]]
	reads the grammar from a file.

Details and Options

▪

A grammar rule has the shape

<name>::=alt1|alt2|…

: a non-terminal name in angle brackets, an arrow, then alternatives separated by

. Each alternative is a whitespace-separated sequence of elements.

▪

In a rule body a non-terminal reference

<name>

lowers to a recursive reference resolved at parse time (through

ParseRecursive

), a bare token lowers to a

ParseLiteral

, and a postfix

<name>*

lowers to a

ParseMany

▪

Whitespace between adjacent elements is consumed automatically, so

12+34

and

12+34

both parse.

▪

Each value in the returned association is the parser for one rule; run it on input with

Parse

▪

Four arrow kinds are recognized:

::=

(syntactic),

:==

(semantic, same surface shape),

::-

(token construction), and

:::

(character class). The

::-

and

:::

bodies compile through a regex-style meta-parser that handles classes (

[a-z]

[^*]

), grouping (

(x|y)

), references (

<name>

), and the postfix operators

▪

A directly left-recursive rule

A::=Ar|b

is rewritten to the equivalent

A::=b(r)*

before lowering; indirect (mutual) left recursion is not rewritten.

▪

Each rule's alternatives are sorted longest-first by element count, so a longer shared-prefix alternative is tried before a shorter one.

▪

The BNF source is itself parsed by a grammar built entirely from

Parse*

combinators — see the tutorial

Parsing BNF Grammars

Option	Default	Description
"PrimitiveOverrides"	`<
"Actions"	`<
"ChoiceMode"	"Auto"	how alternatives combine: "PEG" (first match wins), "Longest" (longest match wins), or "Auto" (longest match only when alternatives have equal element counts)