Reference¶

compynator.core
compynator.niceties

compynator.core¶

The essentials of all parser combinators.

The six basic regex definitions are mapped according to:

Regex	Compynator
empty set	`Fail`
epsilon	`Empty`
character	`Terminal`
concatenation	`+`
alternative	`\|` or `^`
Kleene star	`repeat`

Monadic properties are Succeed for unit, Parser.then for bind, and optionally Fail for zero.

compynator.core.Empty = Parser<Succeed@140076859025104>¶: An empty string. This always succeeds.

class compynator.core.Failure(parser, message, remain='', cause_or_causes=None)¶

A collection of zero ``Result``s.

This class is used to propagate parse failures and incrementally construct a Success. We usually start out with an instance of this class, then add more Result objects to it to produce a Success.

>>> parser = 'a' + Terminal('b') + 'c'
>>> ret = Failure(parser, 'Parser fails.', None)
>>> s = ret.add_all(parser('abc'))
>>> isinstance(s, Success)
True

add(result)¶: Returns a ResultSet (could be self) with result.

add_all(results)¶: Returns a ResultSet (could be self) with all results.

class compynator.core.ParseContext(options=None)¶: Internal book-keeping data structure.

class compynator.core.ParseOptions(max_recursion)¶

property max_recursion¶: Alias for field number 0

class compynator.core.Parser(parser_function)¶

Callable that takes a sequence of tokens & returns a ResultSet.

The specific types of inputs and outputs are not known. However, inputs usually are strings. The requirements for inputs are:

they must have __len__
they are indexable and slicable

A parser must return a collection of Result``s. Their ``value elements can be any type but their remain elements must be a slice of the input tokens.

This class can be used as a decorator:

@Parser
def head(tokens):
    if len(tokens) >= 1:
        return Success(result=Result(tokens[0], tokens[1:]))
    return Failure(head, 'Unable to obtain more tokens', tokens)

In that example, head is a parser that returns the first element of the sequence of tokens. Then head can be chained (then, +) with other parsers, filtered (where, value), or composed together to be more useful.

call(callback)¶

Simple wrapper around filter to always call callback.

In ambiguous grammar (like the example below), there might be repeated results if call makes up a part of the variable. Please note the difference in two definitions of the same production rule.

>>> count = 0
>>> def cb(r):
...     global count
...     count += 1
>>> empty = Succeed('')
>>> s = ((Terminal('s') + (lambda _: s)) ^ empty).call(cb)
>>> r = s('ss')
>>> assert len(r) == 3
>>> count
6
>>> count = 0
>>> s = (Terminal('s') + (lambda _: s)) ^ empty
>>> r = s.call(cb)('ss')
>>> assert len(r) == 3
>>> count
3

filter(callback, take_if=True)¶

Executes callback on a successful parse and filters results.

callback must take a Result. Every possible result of a rule will be passed to callback.

If truth value as returned by callback is the same as take_if, that Result object is included.

NOTE: The ordering between filter and memoize is important and may result in callback not being invoked.

memoize()¶

Memoizes parsed results of self.

The memoization allows for ambiguous grammar to be processed efficiently. See the paper Parser Combinators for Ambiguous Left Recursive Grammars.

This modifier is recommended when the unbiased __xor__ operator is used, or when left recursion is in the grammar:

>>> empty = Succeed('')
>>> s = ((Terminal('s') + (lambda _: s) + (lambda _: s)) ^
...      empty).memoize()
>>> len(s('s' * 20))
21

Without the memoize modifier in the above example, it would take a very long time to parse.

parse(tokens)¶: Parses the input tokens under the default context.

parse_with_context(tokens, context)¶: Parses input tokens under the context of context.

repeat(lower=0, upper=None, reducer=<built-in function concat>, value='', take_all=False)¶

Repeatedly parses [lower, upper] occurrences.

If upper is None, there is no upper bound. The reducer is used to join the results together similar to how it is used in then. The zeroth parse result (parser is not invoked yet) is a Success of value. The first reduction is between zeroth and first results. If take_all, then all results are returned. If not take_all, then only the greediest results are returned.

>>> p = Terminal('a').repeat()
>>> set(p(''))
{Result(value='', remain='')}
>>> set(p('b'))
{Result(value='', remain='b')}
>>> set(p('a'))
{Result(value='a', remain='')}
>>> set(p('aa'))
{Result(value='aa', remain='')}

skip(binder)¶: Similar to then, but the reducer takes the first value.

then(binder, reducer=<function Parser.<lambda>>)¶

Chains self and parser(s) returned by binder via reducer.

This is the bind function in monadic sense. binder is a callable that takes in a Result.value and returns a Parser object. This parser is then applied on Result.remain.

binder can also be a Parser object. In this case, binder is used directly as the second parser.

If not, binder will be converted into a Terminal(str(binder)).

reducer takes two arguments, the first is Result.value of this parser, and the second is the Result.value of the second parser. The result of reducer makes up the final result of the composed parser.

The default reducer only takes the second Result.value.

In code, this looks like:

ret = Fail(tokens)
for value, remain self(tokens):
    next_parser = binder(value)
    for next_value, next_remain in next_parser(remain):
        final_value = reducer(value, next_value)
        ret = ret.add(Result(final_value, next_remain))

value(converter_or_value)¶

Converts Result.value into a different value.

converter_or_value can be a callable, or an object. If it is a callable, it takes Result.value and returns a converted value. If it is a value, that value is used.

For example:

>>> digit = One.where(lambda c: '0' <= c <= '9')
>>> set(digit('8bc'))
{Result(value='8', remain='bc')}
>>> digit_as_int = digit.value(int)
>>> set(digit_as_int('8bc'))
{Result(value=8, remain='bc')}

where(predicate)¶

Selects results whose values pass predicate.

predicate is a callable that takes Result.value and returns True if that Result should be included. This is a convenient wrapper around filter.

For example:

>>> digit = One.where(lambda c: '0' <= c <= '9')
>>> set(digit('abc'))
set()
>>> set(digit('8bc'))
{Result(value='8', remain='bc')}

class compynator.core.Result¶

Holds the parsed results.

Each result is a 2-tuple of value and remaining unparsed sequence of tokens.

NOTE: The input tokens are assumed to be immutable and len(remain) is sufficient to tell if two ``Result.remain``s are equal.

class compynator.core.ResultSet¶

A sized iterable collection of Result.

To incrementally construct a result set, first start with a Failure, then add more Result via add or add_all.

add(result)¶: Returns a ResultSet (could be self) with result.

add_all(results)¶: Returns a ResultSet (could be self) with all results.

class compynator.core.Succeed(value)¶

Always returns a parsed result of value regardless of input.

For example:

>>> s = Succeed(10)
>>> set(s('abc'))
{Result(value=10, remain='abc')}
>>> set(s('def'))
{Result(value=10, remain='def')}

parse_with_context(tokens, context)¶: Parses input tokens under the context of context.

class compynator.core.Success(*args, result=None, results=None)¶

A collection of Result in a successful parse.

A Success must have at least one Result. The constructor can take either keyword argument result or results, but not both at the same time.

add(result)¶: Returns a ResultSet (could be self) with result.

add_all(results)¶: Returns a ResultSet (could be self) with all results.

class compynator.core.Terminal(terminal)¶

Matches terminal to the beginning of input tokens.

>>> t = Terminal('t')
>>> set(t(''))
set()
>>> set(t('t'))
{Result(value='t', remain='')}

parse_with_context(tokens, context)¶: Parses input tokens under the context of context.

compynator.core.default_parse_context(tokens)¶: Returns ParseContext for tokens.

compynator.niceties¶

compynator.niceties.Alnum = Parser<_Or@140076859338000>¶: Exactly one ASCII letter or digit.

compynator.niceties.Alpha = Parser<_Or@140076859337872>¶: Exactly one letter a-zA-Z

class compynator.niceties.Collect(*parsers)¶

A combinator that runs through all parsers in sequence and collects their results in a collection of many flattened collections.

This is best described with examples:

>>> a, b, c = [Terminal(x) for x in 'abc']
>>> p = Collect(a, b, c)
>>> set(p('adc'))
set()
>>> p('adc')
Failure('Failed to collect.', 'adc', [Failure("Parser index 1: Expecting terminal 'b'.", 'dc', ())])
>>> set(p('abc'))
{Result(value=('a', 'b', 'c'), remain='')}
>>> a = a.repeat(0, 2, take_all=True)
>>> p = Collect(a, a, a)
>>> rs = p('a')
>>> len(rs)
4
>>> Result(value=('', '', ''), remain='a') in rs
True
>>> Result(value=('', '', 'a'), remain='') in rs
True
>>> Result(value=('', 'a', ''), remain='') in rs
True
>>> Result(value=('a', '', ''), remain='') in rs
True
>>> len(p('aa'))  # -/-/-, -/-/a, -/a/-, a/-/-,         ...               # -/-/aa, -/a/a, -/aa/-, a/-/a, a/a/-, aa/-/-
10

Note that the final ResultSet could grow exponentially.

parse_with_context(tokens, context)¶: Parses input tokens under the context of context.

compynator.niceties.Digit = Parser<_Filter@140076859081552>¶: Exactly one decimal digit.

class compynator.niceties.Forward¶

A forward declaration of a rule.

This is useful in case the rule is defined recursively. For example, the BNF rule exp ::= (exp '-' exp) | 'o' could be defined as followed:

>>> exp = Forward()
>>> exp.is_(((exp + '-' + exp) ^ 'o').memoize())
>>> set(exp('o'))
{Result(value='o', remain='')}
>>> sorted(exp('o-o'))
[Result(value='o', remain='-o'), Result(value='o-o', remain='')]

A forward declaration of Parser is the same as referring to that parser in a lambda:

>>> exp = (Succeed(None).then(lambda _: exp + '-' + exp) ^ 'o').memoize()
>>> set(exp('o'))
{Result(value='o', remain='')}
>>> sorted(exp('o-o'))
[Result(value='o', remain='-o'), Result(value='o-o', remain='')]

A RuntimeError will be raised if a Forward has not called is_, or if that method is called more than once.

>>> exp = Forward()
>>> exp('abc')
Traceback (most recent call last):
  File "<stdin>", line 1, in ?
RuntimeError: A forward declaration has no definition.
>>> exp.is_('abc')
>>> exp.is_('abc')
Traceback (most recent call last):
  File "<stdin>", line 1, in ?
RuntimeError: Already defined.

is_(parser)¶

Defines a forward declaration.

If parser is not typed Parser, its string representation will be made into a Terminal.

This method must be called exactly once for each Forward object. A RuntimeError will be raised if it is called more than once.

parse_with_context(tokens, context)¶: Parses input tokens under the context of context.

compynator.niceties.HexDigit = Parser<_Or@140076859081808>¶: Exactly one hexadecimal digit.

class compynator.niceties.ITerminal(terminal)¶

Case insensitive terminal.

parse_with_context(tokens, context)¶: Parses input tokens under the context of context.

class compynator.niceties.Lookahead(parser, take_if=True, value='')¶

Tries parser but does not consume input.

If the truth value of the parse result is take_if, a Success of value is returned. Otherwise, a Failure is returned.

parse_with_context(tokens, context)¶: Parses input tokens under the context of context.

compynator.niceties.Lower = Parser<_Filter@140076859336400>¶: Exactly one letter a-z.

compynator.niceties.OctDigit = Parser<_Filter@140076859335056>¶: Exactly one octadecimal digit.

class compynator.niceties.Regex(regex)¶

Regex matcher.

parse_with_context(tokens, context)¶: Parses input tokens under the context of context.

compynator.niceties.Upper = Parser<_Filter@140076859337616>¶: Exactly one letter A-Z