Skip to content

Building parsers

The peg tag accepts any valid Pegase expression and always returns a Parser instance.

Pegase parsers follow the combinator paradigm: simple parsers are combined to form more complex parsers. You can read more about it in the Parser section. In the following sections are the different expressions you can use as building blocks. To see the exact grammar of pegs (called metagrammar), please refer to the Metagrammar page.


Any character

.

Matches any character.

Children: []

Precedence: 0


End of input

$

Matches the end of the input. This expression is syntactic sugar for !. @token("end of input"). Indeed, the end of the input has been reached when there's no character left, thus when "not any character" (!.) can be evaluated as true.

Children: []

Precedence: 0


Epsilon

ε

Matches the empty string. Strictly equivalent to '' and always a success.

Children: []

Precedence: 0


Cut operator

^

In an ordered choice expression, the cut operator commits to the current alternative to prevent exploring any further in case it fails. Example : 'x' ^ a | b will not try b if 'x' was found but a failed. Although it's called an operator, it's really a Parser on its own. Used outside an ordered choice expression, it's a no-op. Read more in Cut operator.

Children: []

Precedence: 0


Sub-parsers

(expr)
('a' 'b' 'c')
${jsParser}

Delegates the parsing to a sub-parser. This happens when wrapping a peg expression in parentheses, or by injecting an external Parser instance via tag argument.

Children: Forwarded from the sub-parser.

Precedence: 0


Back references

>id<

Back reference to an earlier string capture. This will matches the string literal captured as id. Read more in Semantic action and dataflow.

Children: []

Precedence: 0


Nullary directives

@@dir
@@dir(x, y)
@${jsFunction}
@ => 'label'

This is syntactic sugar for directives applied to the empty literal parser: '' @dir, '' ${jsFunction}, etc. Handy if the directive you want to invoke doesn't care about a wrapped parser. Read more in Directives.

Children: Directives generate new parsers. So children depends on whatever parser is generated.

Precedence: 0


Non-terminals

identifier
identifier('a', 'c' | 'd')
identifier('a',, 'd')

Matches the non-terminal. A non-terminal can refer to a grammar rule or to a rule parameter. Since rules can be parametrized, non-terminals can have parameters. Parameters can be omitted at any place if the corresponding rule defines a default value for them.

Children: Forwarded from the non-terminal (identifier in the example)

Precedence: 0


String parsers

"lit"
'test'
''
${jsString}
42
${jsNumber}

Matches the string literal. When a JS string value is inserted as a tag argument into the a peg expression, it's cast into a non-emissive string literal parser. String literals that represent numbers can directly be written as numbers, without quotes. Inserting a JS number via tag argument also creates a string literal parser.

Children: When double-quotes are used, the parsed substring is emitted as a single child. [] otherwise.

Precedence: 0


Character classes

[abc]
[0-9]
[a-zA-Z]
[^u]
[^a-zA-Z]

There are two types of character classes: non-negated and negated (^). Non-negated classes match one character in the given character class. Negated classes match one character not in the given class. Internally, characters classes are instantiated as RegExp instances and thus end up as regex parsers.

Children: []

Precedence: 0


Metacharacters

\n
\s
\w
\xAF
\uA6F1

Matches the escaped metacharacter. The same metacharacters available in RegExp expressions are available in Pegase, meaning \s matches any whitespace, \S any non-whitespace, \uA6F1 matches the unicode character A6F1, etc. (See RegExp documentation for a complete list of supported metacharacters).

Children: []

Precedence: 0


Regex parsers

/ab/
/\w+/
/(\d+):(\d+)/
${jsRegExp}

Matches the regex. It can either be a regex literal (same syntax as JS' RegExp), or an actual RegExp instance injected as tag argument. Named capturing groups are transformed into Pegase captures. Read more in Working with RegExp.

Children: The regex's capturing groups

Precedence: 0


Repetitions

a?
a+
a*
a{4}
a{4, 15}
a{4,}
a{${jsNumber}, 15}

Invokes a sub-parser (a) greedily as many times as specified. Different kind of quantifiers are available:

Expression form Meaning
a? Optional a (zero or one)
a+ Matches one or more a
a* Matches any sequence of a (zero or more)
a{n} Matches a exactly n times. n can be a number literal or a JS number injected via tag argument.
a{n, m} Matches a between n and m times. Number literals and tag arguments are allowed for n and m.
a{n,} Matches a at least n times. As above, n is a number literal or a tag argument.

Children: Forwarded and concatenated from the wrapped expression

Precedence: 1


Predicates

&a
!a

In case of a positive predicate ($), a is matched without consuming any input. For negative predicates (!), the expression fails if a succeeds and vice-versa.

Children: []

Precedence: 2


Captures

<id>a
<>id
<...id>a
<...>id

Captures associate a parser's value or children to an identifier (id in the example). The value of a parser is its single child. If there is no child, or multiple ones, value is undefined. To captures the whole children array, use .... When the wrapped expression is a non-terminal of the same name than the capture, the capture's name can be omitted.

Captures are accumulated in scopes, can be overwritten and be read in semantic actions and custom parsers. Rule definitions and ordered choice alternatives create new capture scopes. Read more in Semantic action and dataflow.

Children: Forwarded from the wrapped expression

Precedence: 3


Synchronizations

...a

Skips input character by character until a is matched. This can be used to implement synchronization to recover from errors and is equivalent to (!a .)* a. Write ...&a if you want to sync to a without consuming a. See Failure recovery.

Children: Forwarded from the wrapped expression

Precedence: 4


Repetitions with separators

a % b
a %? b
a %{3} b
a %{3,} b
a %{3, ${jsNumber}} b

Matches a sequence of a separated by b. The % operator can be quantified explicitly using the quantifiers described in Repetitions. a % b is equivalent to a (b a)*, a %? b to a (b a)?, etc.

Children: Forwarded and concatenated from the subexpressions

Precedence: 5


Subtractions

a - b

Matches a but not b (fails if b succeeds). Strictly equivalent to !b a.

Children: Forwarded from expression a

Precedence: 6


Sequences

a b
a b c

Matches a followed by b.

Children: Forwarded and concatenated from expression a and b

Precedence: 7


Directives (includes semantic actions)

a @dir
a @dir(x, y)
a @dir(x, ${y})
a @dir1 @dir2 @dir3

a ${jsFunction}
a => 'label'

Applies the directive(s) to the parser a. Directives are functions that take a parser and return a new parser. They can take additional arguments and can be chained. These arguments can include number, string, regex literals, characters classes, metacharacters, non-terminals and tag arguments.

Injecting a JS function right after a peg expression is syntactic sugar for the @action directive, which builds an action parser (a parser that calls a custom function on success). So a ${jsFunction} and a @action(${jsFunction}) are strictly equivalent.

The a => 'label' syntax is syntactic sugar for the @node directive and is equivalent to a @node('label'). The label can be specified via a string tag argument if necessary.

Directives are a powerful and central tool within Pegase. Read more about it in Directives.

Children: Directives generate new parsers. So children depends on whatever parser is generated. For example: when using the @action directive (or shorthand), one can emit a single child by returning a non-undefined value from the function, or emit multiple children by calling the $emit hook inside the function. Not returning anything and not calling $emit will forward the wrapped expression's children. When using the @node directive (or shorthand), the generated Node is emitted as a single child (see AST and visitors).

Precedence: 8


Ordered choices

a | b
| a | b | c

Ordered choices (or alternatives) try their subexpression one by one in order and succeed if a subexpression succeeds. Fails otherwise. Please note that you can add a leading bar for aesthetic purposes.

Children: Forwarded from the successful subexpression (a or b in the example)

Precedence: 9


Rules and grammars

id: a
$id: a
id @directive: a
id(p1, p2): a
id(p1, p2) @directive: a
id(p = \d): a
id(p1, p2 = 'a', p3): a

These expressions create rules. Think of rules as alias to peg expressions. Once defined, they can be invoked via non-terminals (see Non-terminals). Rules can be stacked to form grammars. They can also be parametrized and the parameters can take default values. Since non-terminal invocations can skip any parameter, default parameters values don't need to appear at the last parameter positions.

If directives are specified just before :, they are applied to the whole right-side expression.

Adding $ at the beginning of a rule name acts as syntactic sugar. It applies an implicit @token directive whose display name in failure reports will be the rule name transformed to space case. Example: $myToken: a is equivalent to $myToken @token("my token"): a.

Grammars can be nested by using parentheses:

r1: 'a' (
  nested1: 'u' nested2
  nested2: 'v'
) r2
r2: 'b'

Nesting grammars can be used to parametrize entire grammars, or as a separation of concern method.

Children: Forwarded from the topmost rule

Precedence: 10

Back to top