spirit/doc/qi/directive.qbk
2018-12-18 20:49:25 +00:00

965 lines
26 KiB
Plaintext

[/==============================================================================
Copyright (C) 2001-2011 Joel de Guzman
Copyright (C) 2001-2011 Hartmut Kaiser
Copyright (C) 2011 Bryce Lelbach
Copyright (C) 2016 Frank Hein, maxence business consulting gmbh
Distributed under the Boost Software License, Version 1.0. (See accompanying
file LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt)
===============================================================================/]
[section:directive Parser Directives]
This module includes different directives usable to augment and parameterize
other parsers. It includes the `no_case`, `lexeme`, `omit`, `raw`, `repeat`,
`matches`, `no_skip`, `skip`, `hold`, `as<T>`, `as_string` and
`as_wstring` directives.
[heading Module Header]
// forwards to <boost/spirit/home/qi/directive.hpp>
#include <boost/spirit/include/qi_directive.hpp>
Also, see __include_structure__.
[/------------------------------------------------------------------------------]
[section:lexeme Parser Directive Inhibiting Skipping (`lexeme[]`)]
[heading Description]
The `lexeme` directive makes its subject a primitive. In a logical point
of view, lexemes (and primitives) are minimal atomic units (e.g. words,
numbers, identifiers, etc). These are the things that you'd normally put
in the lexer (hinting at the term "lexeme"), but in a lexer-less world,
you put these in a lexeme. Seeing its subject as a primitive, the
`lexeme` directive does an initial pre-skip (as all primitives do) and
turns off white space skipping.
At the phrase level, the parser ignores white spaces, possibly including
comments. Use `lexeme` in situations where you want to work at the
character level instead of the phrase level. Parsers can be made to work
at the character level by enclosing the pertinent parts inside the
`lexeme` directive. For example, here's a rule that parses integers:
integer = lexeme[ -(lit('+') | '-') >> +digit ];
The `lexeme` directive instructs its subject parser to work on the
character level. Without it, the `integer` rule would have allowed
erroneous embedded white spaces in inputs such as `"1 2 345"` which will
be parsed as `"12345"`.
[note Keep in mind that `lexeme[]` pre-skips spaces. If this is not
desired, use the [qi_no_skip `no_skip`] directive instead.]
[heading Header]
// forwards to <boost/spirit/home/qi/directive/lexeme.hpp>
#include <boost/spirit/include/qi_lexeme.hpp>
Also, see __include_structure__.
[heading Namespace]
[table
[[Name]]
[[`boost::spirit::lexeme // alias: boost::spirit::qi::lexeme` ]]
]
[heading Model of]
[:__unary_parser_concept__]
[variablelist Notation
[[`a`] [A __parser_concept__.]]
]
[heading Expression Semantics]
Semantics of an expression is defined only where it differs from, or is
not defined in __unary_parser_concept__.
[table
[[Expression] [Semantics]]
[[`lexeme[a]`] [Pre-skip and turn off white space skipping for the
subject parser, `a` (and all its children).]]
]
[heading Attributes]
See __qi_comp_attr_notation__.
[table
[[Expression] [Attribute]]
[[`lexeme[a]`]
[``a: A --> lexeme[a]: A
a: Unused --> lexeme[a]: Unused``]]
]
[heading Complexity]
[:The complexity is defined by the complexity of the subject parser, `a`]
[heading Example]
[note The test harness for the example(s) below is presented in the
__qi_basics_examples__ section.]
Some using declarations:
[reference_using_declarations_lexeme]
Simple usage of `lexeme[]`:
[reference_lexeme]
[endsect]
[/------------------------------------------------------------------------------]
[section:no_skip Parser Directive Inhibiting Skipping Without Pre-skip (`no_skip[]`)]
[heading Description]
The `no_skip[]` directive turns off white space skipping. The difference to
__qi_lexeme__ is that it does not do pre-skipping in any case. Otherwise it is
completely equivalent to the __qi_lexeme__ directive.
[heading Header]
// forwards to <boost/spirit/home/qi/directive/no_skip.hpp>
#include <boost/spirit/include/qi_no_skip.hpp>
Also, see __include_structure__.
[heading Namespace]
[table
[[Name]]
[[`boost::spirit::no_skip // alias: boost::spirit::qi::no_skip` ]]
]
[heading Model of]
[:__unary_parser_concept__]
[variablelist Notation
[[`a`] [A __parser_concept__.]]
]
[heading Expression Semantics]
Semantics of an expression is defined only where it differs from, or is
not defined in __unary_parser_concept__.
[table
[[Expression] [Semantics]]
[[`no_skip[a]`] [Turns off white space skipping for the
subject parser, `a` (and all its children). This
directive does not pre-skips.]]
]
[heading Attributes]
See __qi_comp_attr_notation__.
[table
[[Expression] [Attribute]]
[[`no_skip[a]`]
[``a: A --> no_skip[a]: A
a: Unused --> no_skip[a]: Unused``]]
]
[heading Complexity]
[:The complexity is defined by the complexity of the subject parser, `a`]
[heading Example]
[note The test harness for the example(s) below is presented in the
__qi_basics_examples__ section.]
Some using declarations:
[reference_using_declarations_no_skip]
Simple usage of `no_skip[]`:
[reference_no_skip]
[endsect]
[/------------------------------------------------------------------------------]
[section:no_case Parser Directive Inhibiting Case Sensitivity (`no_case[]`)]
[heading Description]
The `no_case[]` directive does not consume any input. The actual
matching is done by its subject parser. It's purpose is to force
matching of the subject parser (and all its children) to be case
insensitive.
[heading Header]
// forwards to <boost/spirit/home/qi/directive/no_case.hpp>
#include <boost/spirit/include/qi_no_case.hpp>
Also, see __include_structure__.
[heading Namespace]
[table
[[Name]]
[[`ns::no_case`]]
]
In the table above, `ns` represents a __char_encoding_namespace__.
[heading Model of]
The model of `no_case` is the model of its subject parser.
[variablelist Notation
[[`a`] [A __parser_concept__.]]
[[`ns`] [A __char_encoding_namespace__.]]
]
[heading Expression Semantics]
Semantics of an expression is defined only where it differs from, or is
not defined in the subject's concept.
[table
[[Expression] [Semantics]]
[[`ns::no_case[a]`] [Force matching of the subject parser, `a`
(and all its children) to be case insensitive]]
]
[heading Attributes]
See __qi_comp_attr_notation__.
[table
[[Expression] [Attribute]]
[[`ns::no_case[a]`]
[``a: A --> ns::no_case[a]: A
a: Unused --> ns::no_case[a]: Unused``]]
]
[heading Complexity]
[:The complexity is defined by the complexity of the subject parser, `a`]
[heading Example]
[note The test harness for the example(s) below is presented in the
__qi_basics_examples__ section.]
Some using declarations:
[reference_using_declarations_no_case]
Simple usage of `no_case[]`:
[reference_no_case]
A more sophisticated use case of `no_case[]` in conjunction with a symbol
table (see __qi_symbols__ for more details):
[reference_symbols_with_no_case]
[endsect]
[/------------------------------------------------------------------------------]
[section:omit Parser Directive Ignoring Attribute (`omit[]`)]
[heading Description]
The `omit[]` ignores the attribute of its subject parser replacing it
with __unused__.
[heading Header]
// forwards to <boost/spirit/home/qi/directive/omit.hpp>
#include <boost/spirit/include/qi_omit.hpp>
Also, see __include_structure__.
[heading Namespace]
[table
[[Name]]
[[`boost::spirit::omit // alias: boost::spirit::qi::omit` ]]
]
[heading Model of]
[:__unary_parser_concept__]
[variablelist Notation
[[`a`] [A __parser_concept__.]]
]
[heading Expression Semantics]
Semantics of an expression is defined only where it differs from, or is
not defined in __unary_parser_concept__.
[table
[[Expression] [Semantics]]
[[`omit[a]`] [Ignore the attribute of the subject parser, `a`]]
]
[heading Attributes]
[table
[[Expression] [Attribute]]
[[`omit[a]`] [__unused_type__]]
]
[heading Complexity]
[:The complexity is defined by the complexity of the subject parser, `a`]
[heading Example]
[note The test harness for the example(s) below is presented in the
__qi_basics_examples__ section.]
Some using declarations:
[reference_using_declarations_omit]
[reference_omit]
[endsect]
[/------------------------------------------------------------------------------]
[section:raw Directive for Transduction Parsing (`raw[]`)]
[heading Description]
The `raw[]` disregards the attribute of its subject parser, instead
exposing the half-open range `[first, last)` pointing to the matched
characters from the input stream. The `raw[]` directive brings back the
classic Spirit transduction (un-attributed) behavior for a subject
parser.
[heading Header]
// forwards to <boost/spirit/home/qi/directive/raw.hpp>
#include <boost/spirit/include/qi_raw.hpp>
Also, see __include_structure__.
[heading Namespace]
[table
[[Name]]
[[`boost::spirit::raw // alias: boost::spirit::qi::raw` ]]
]
[heading Model of]
[:__unary_parser_concept__]
[variablelist Notation
[[`a`] [A __parser_concept__.]]
[[`Iter`] [A __fwditer__ type.]]
]
[heading Expression Semantics]
Semantics of an expression is defined only where it differs from, or is
not defined in __unary_parser_concept__.
[table
[[Expression] [Semantics]]
[[`raw[a]`] [Disregard the attribute of the subject parser, `a`.
Expose instead the half-open range `[first, last)`
pointing to the matched characters from the input stream.]]
]
[heading Attributes]
See __qi_comp_attr_notation__.
[table
[[Expression] [Attribute]]
[[`raw[a]`]
[``a: A --> raw[a]: boost::iterator_range<Iter>
a: Unused --> raw[a]: Unused``]]
]
[note See __boost_iterator_range__.]
[heading Complexity]
[:The complexity is defined by the complexity of the subject parser, `a`]
[heading Example]
[note The test harness for the example(s) below is presented in the
__qi_basics_examples__ section.]
Some using declarations:
[reference_using_declarations_raw]
[reference_raw]
[endsect]
[/------------------------------------------------------------------------------]
[section:repeat Repetition Parser Directive (`repeat[]`)]
[heading Description]
The `repeat[]` provides a more powerful and flexible mechanism for
repeating a parser. There are grammars that are impractical and
cumbersome, if not impossible, for the basic EBNF iteration syntax
(__qi_kleene__ and the __qi_plus__) to specify. Examples:
* A file name may have a maximum of 255 characters only.
* A specific bitmap file format has exactly 4096 RGB color information.
* A 256 bit binary string (1..256 1s or 0s).
[heading Header]
// forwards to <boost/spirit/home/qi/directive/repeat.hpp>
#include <boost/spirit/include/qi_repeat.hpp>
Also, see __include_structure__.
[heading Namespace]
[table
[[Name]]
[[`boost::spirit::repeat // alias: boost::spirit::qi::repeat` ]]
[[`boost::spirit::inf // alias: boost::spirit::qi::inf` ]]
]
[heading Model of]
[:__unary_parser_concept__]
[variablelist Notation
[[`a`] [A __parser_concept__.]]
[[`n`, `min`, `max`] [An `int` anything that can be converted to an
`int`, or a __qi_lazy_argument__ that evaluates to
anything that can be converted to an `int`.]]
]
[heading Expression Semantics]
Semantics of an expression is defined only where it differs from, or is
not defined in __unary_parser_concept__.
[table
[[Expression] [Semantics]]
[[`repeat[a]`] [Repeat `a` zero or more times. Same as __qi_kleene__.]]
[[`repeat(n)[a]`] [Repeat `a` exactly `n` times.]]
[[`repeat(min, max)[a]`] [Repeat `a` at least `min` times and at most `max` times.]]
[[`repeat(min, inf)[a]`] [Repeat `a` at least `min` or more (continuing until `a`
fails or the input is consumed).]]
]
[heading Attributes]
See __qi_comp_attr_notation__.
[table
[[Expression] [Attribute]]
[[`repeat[a]`]
[``a: A --> repeat[a]: vector<A>
a: Unused --> repeat[a]: Unused``]]
[[`repeat(n)[a]`]
[``a: A --> repeat(n)[a]: vector<A>
a: Unused --> repeat(n)[a]: Unused``]]
[[`repeat(min, max)[a]`]
[``a: A --> repeat(min, max)[a]: vector<A>
a: Unused --> repeat(min, max)[a]: Unused``]]
[[`repeat(min, inf)[a]`]
[``a: A --> repeat(min, inf)[a]: vector<A>
a: Unused --> repeat(min, inf)[a]: Unused``]]
]
[heading Complexity]
[:The overall complexity is defined by the complexity of its subject
parser. The complexity of `repeat` itself is O(N), where N is the number
of repetitions to execute.]
[heading Example]
[note The test harness for the example(s) below is presented in the
__qi_basics_examples__ section.]
Using the repeat directive, we can now write our examples above.
Some using declarations:
[reference_using_declarations_repeat]
[reference_repeat]
The Loop parsers can be dynamic. Consider the parsing of a binary file
of Pascal-style length prefixed string, where the first byte determines
the length of the incoming string. Here's a sample input:
[:__pascal_string__]
[reference_repeat_pascal]
[endsect]
[/------------------------------------------------------------------------------]
[section:matches Directive Testing if Parser Succeeded (`matches[]`)]
[heading Description]
The `matches[]` directive executes the embedded parser and returns whether it
succeeded matching.
[heading Header]
// forwards to <boost/spirit/home/qi/directive/matches.hpp>
#include <boost/spirit/include/qi_matches.hpp>
Also, see __include_structure__.
[heading Namespace]
[table
[[Name]]
[[`boost::spirit::matches // alias: boost::spirit::qi::matches` ]]
]
[heading Model of]
[:__unary_parser_concept__]
[variablelist Notation
[[`a`] [A __parser_concept__.]]
]
[heading Expression Semantics]
Semantics of an expression is defined only where it differs from, or is
not defined in __unary_parser_concept__.
[table
[[Expression] [Semantics]]
[[`matches[a]`] [Execute the subject parser `a`, and return as its
attribute whether it succeeded. The directive itself
does always succeed.]]
]
[heading Attributes]
[table
[[Expression] [Attribute]]
[[`matches[a]`] [`bool`]]
]
[heading Complexity]
[:The complexity is defined by the complexity of the subject parser, `a`]
[heading Example]
[note The test harness for the example(s) below is presented in the
__qi_basics_examples__ section.]
Some using declarations:
[reference_using_declarations_matches]
[reference_matches]
[endsect]
[/------------------------------------------------------------------------------]
[section:skip Parser Directive Re-Establishing Skipping (`skip[]`)]
[heading Description]
The `skip` directive is the inverse of __qi_lexeme__ or [qi_no_skip `no_skip`].
While the __qi_lexeme__ directive turns off white space
skipping, the `skip` directive turns it on again. This is simply done by
wrapping the parts inside the `skip` directive:
skip[a]
It is also possible to supply a skip parser to the `skip` directive:
skip(p)[a] // Use `p` as a skipper for parsing `a`
This makes it possible to:
* Perform localized phrase level parsing while doing character level parsing.
* Replace the current skipper anywhere with an entirely different
skipper while doing phrase level parsing.
[heading Header]
// forwards to <boost/spirit/home/qi/directive/skip.hpp>
#include <boost/spirit/include/qi_skip.hpp>
Also, see __include_structure__.
[heading Namespace]
[table
[[Name]]
[[`boost::spirit::skip // alias: boost::spirit::qi::skip` ]]
]
[heading Model of]
[:__unary_parser_concept__]
[variablelist Notation
[[`a`] [A __parser_concept__.]]
]
[heading Expression Semantics]
Semantics of an expression is defined only where it differs from, or is
not defined in __unary_parser_concept__.
[table
[[Expression] [Semantics]]
[[`skip[a]`] [Re-establish the skipper that got inhibited by lexeme or no_skip]]
[[`skip(p)[a]`] [Use `p` as a skipper for parsing `a`]]
]
[heading Attributes]
See __qi_comp_attr_notation__.
[table
[[Expression] [Attribute]]
[[`skip[a]`]
[``a: A --> skip[a]: A
a: Unused --> skip[a]: Unused``]]
[[`skip(p)[a]`]
[``a: A --> skip(p)[a]: A
a: Unused --> skip(p)[a]: Unused``]]
]
[heading Complexity]
[:The complexity is defined by the complexity of the subject parser, `a`]
[heading Example]
[note The test harness for the example(s) below is presented in the
__qi_basics_examples__ section.]
Some using declarations:
[reference_using_declarations_skip]
Simple usage of `skip[]`:
[reference_skip]
[endsect]
[/------------------------------------------------------------------------------]
[section:hold Parser Directive for Attribute Commit/Rollback (`hold[]`)]
[heading Description]
The `hold[]` directive helps managing attributes, mainly for alternative
parsers. It instantiates a new attribute instance for the embedded parser. The
value of that attribute instance is copied to the outer attribute if the
embedded parser succeeds and it is discarded otherwise. Alternative parsers
normally do not rollback changes made to the outer attribute by an failed
alternative. Wrapping those alternatives into a `hold[]` directive ensures that
only the succeeding alternative gets to modify the attribute.
[heading Header]
// forwards to <boost/spirit/home/qi/directive/hold.hpp>
#include <boost/spirit/include/qi_hold.hpp>
Also, see __include_structure__.
[heading Namespace]
[table
[[Name]]
[[`boost::spirit::hold // alias: boost::spirit::qi::hold` ]]
]
[heading Model of]
[:__unary_parser_concept__]
[variablelist Notation
[[`a`] [A __parser_concept__.]]
]
[heading Expression Semantics]
Semantics of an expression is defined only where it differs from, or is
not defined in __unary_parser_concept__.
[table
[[Expression] [Semantics]]
[[`hold[a]`] [Create a new attribute instance while parsing `a`,
copying the result to the outer attribute only after
`a` succeeds.]]
]
[heading Attributes]
See __qi_comp_attr_notation__.
[table
[[Expression] [Attribute]]
[[`hold[a]`]
[``a: A --> hold[a]: A
a: Unused --> hold[a]: Unused``]]
]
[note The `hold[]` directive uses `swap()` to implement the rollback/commit
semantics for the attribute. For this reason the attribute type needs to
to be usable with `boost::swap` (needs to either define a proper overload
for `swap(attribute_type&, attribute_type&)` or expose a member function
`attribute_type::swap(attribute_type&)`.]
[heading Complexity]
[:The complexity is defined by the complexity of the subject parser, `a`]
[heading Example]
[note The test harness for the example(s) below is presented in the
__qi_basics_examples__ section.]
Some using declarations:
[reference_using_declarations_hold]
[reference_hold]
[endsect]
[/------------------------------------------------------------------------------]
[section:as Parser Directives Forcing Atomic Assignment (`as<T>, as_string[], as_wstring[]`)]
[heading Description]
The `as<T>` class forces the atomic assignment of it's subject's synthesized
attribute. Usually, repetitive parsers (such as __qi_kleene__, etc) or
sequences exposing a `vector<A>` will assign elements to the container supplied
as their synthesized attribute by calling __customize_push_back_container__
repeatedly. In some cases, this may be undesirable. The `as<T>` class creates a
directive that will pass a temporary object of type `T` to it's subject. If the
subject parser passes, the temporary object will be assigned to the directive's
supplied attribute with a single call to __customize_assign_to__. If the
subject parser fails, the directive's attribute is not mutated.
[note `T` is required to be a container type. If __customize_is_container__
does not return true for `T`, a compile-time error will occur.]
[note The `as<T>` implicitly causes commit/rollback semantics
similar in nature to the __qi_hold__ directive.]
[caution The __customize_assign_to__ customization point may end up
using __customize_push_back_container__ to assign the temporary object to the
supplied attribute by default, depending on the types involved. Use the
interface described in __sec_customization_points__ to manipulate the semantics
of this assignment operation.]
[heading Header]
// forwards to <boost/spirit/home/qi/directive/as.hpp>
#include <boost/spirit/include/qi_as.hpp>
Also, see __include_structure__.
[heading Namespace]
[table
[[Name]]
[[`boost::spirit::as // alias: boost::spirit::qi::as` ]]
[[`boost::spirit::as_string // alias: boost::spirit::qi::as_string` ]]
[[`boost::spirit::as_wstring // alias: boost::spirit::qi::as_wstring` ]]
]
[heading Synopsis]
template <typename T>
struct as;
[heading Template parameters]
[table
[[Parameter] [Description] [Default]]
[[`T`] [A container type. [none]]
]
[heading Model of]
[:__unary_parser_concept__]
[variablelist Notation
[[`a`] [A __parser_concept__.]]
[[`t`] [A container of type `T`.]]
[[`attr`] [The attribute supplied to the directive.]]
]
[heading Expression Semantics]
Semantics of an expression is defined only where it differs from, or is
not defined in __unary_parser_concept__.
[table
[[Expression] [Semantics]]
[[`as<T>()[a]`] [Create a temporary object of `t` of type `T`,
and invoke the subject parser `a`, supplying
`t` as an attribute. If the subject parser
passes, assign `t` to `attr`.]]
[[`as_string[a]`] [Equivalent to `as<std::string>()[a]`]]
[[`as_wstring[a]`] [Equivalent to `as<std::wstring>()[a]`]]
]
[heading Attributes]
See __qi_comp_attr_notation__.
[table
[[Expression] [Attribute]]
[[`as<T>()[a]`] [`a: A --> as<T>()[a]: T`]]
]
[heading Complexity]
[:The complexity is defined by the complexity of the subject parser, `a`, and
the complexity of the assignment of the container `t` to the supplied
attribute `attr`.]
[heading Example]
[note The test harness for the example(s) below is presented in the
__qi_basics_examples__ section.]
Some using declarations:
[reference_using_declarations_as]
Simple usage of `as<T>`, `as_string` and `as_wstring`:
[reference_as]
[endsect]
[/------------------------------------------------------------------------------]
[section:expect Expectation Directive (`expect[]`)]
[heading Description]
There are occasions in which it is expected that the input must match a
particular parser or the input is invalid. Such cases generally arise
after matching a portion of a grammar, such that the context is fully
known. In such a situation, failure to match should result in an
exception. For example, when parsing an e-mail address, a name, an "@"
and a domain name must be matched or the address is invalid.
The expect directive requires that the argument parser matches
the input or an exception is emitted. Using on_error(), that exception
can be handled by calling a handler with the context at which the
parsing failed can be reported.
The expect directive parses an operand parser expression which may be a
single parser or a complex parser expression like a sequence.
Single parser:
expect[a]
Parser expression:
expect[a >> b >> ...]
In the latter case while the plain __qi_sequence__ simply returns a no-match
(returns `false`) when one of the elements fail, the expect directive
throws an __qi_expectation_failure__`<Iter>` if any of the parsers
(even the first parser of a sequence) fails to match.
[note Spirit provides two ways to handle expectation failures by throwing an expectation exception.
Use the __qi_expect__ if you do not need an exception to be thrown when the first parser of a sequence fails. ]
[heading Header]
// forwards to <boost/spirit/home/qi/directive/expect.hpp>
#include <boost/spirit/include/qi_expect.hpp>
Also, see __include_structure__.
[heading Model of]
[:__unary_parser_concept__]
[variablelist Notation
[[`a`] [A __parser_concept__]]
[[`Iter`] [A __fwditer__ type]]
]
[heading Expectation Failure]
When the operand parser fails to match an `expectation_failure<Iter>` is thrown:
template <typename Iter>
struct expectation_failure : std::runtime_error
{
Iter first; // [first, last) iterator pointing
Iter last; // to the error position in the input.
__info__ what_; // Information about the nature of the error.
};
[heading Expression Semantics]
Semantics of an expression is defined only where it differs from, or is not
defined in __unary_parser_concept__.
[table
[[Expression] [Semantics]]
[[`expect[a]`] [Match `a`. If `a` fails, throw an `expectation_failure<Iter>`]]
]
[heading Attributes]
See __qi_comp_attr_notation__.
[table
[[Expression] [Attribute]]
[[`expect[a]`]
[``a: A --> expect[a]: A
a: Unused --> expect[a] : Unused``]]
]
[heading Complexity]
[:The overall complexity of the expectation parser is defined by the
complexity of it's argument parser. The complexity of the expect directive itself is O(1).]
[heading Example]
[note The test harness for the example(s) below is presented in the
__qi_basics_examples__ section.]
Some using declarations:
[reference_using_declarations_expectd]
[reference_expectd]
[endsect]
[endsect]