The framework predefines some parser primitives. These are the most basic building blocks that the client uses to build more complex parsers. These primitive parsers are template classes, making them very flexible.
All of these primitive parsers are classes which can be instantiated directly or through a templatized helper function. Generally, the helper function is far simpler to deal with as it involves less typing.
We've seen the character literal parser before through the generator function ch_p which is not really a parser but, rather, a parser generator. Class chlit<CharT> is the actual template class behind the character literal parser. To instantiate a chlit object, you must explicitly provide the character type, CharT, as a template parameter which determines the type of the character. This type typically corresponds to the input type, usually char or wchar_t. The following expression creates a temporary parser object which will recognize the single letter 'X'.
Using chlit's generator function ch_p simplifies the usage of the chlit<> class (this is true of most Spirit parser classes, for that matter, since most have corresponding generator functions). It is more convenient to call the function because the compiler will deduce the template type through argument deduction for us. The example above could be expressed less verbosely using the ch_p helper function, .
ch_p('X') // equivalent to chlit<char>('X') object
Whenever you see an invocation of the parser generator function, it is equivalent to the parser itself. Therefore, we often call ch_p a character parser, even if, technically speaking, it is a function that generates a character parser.
The following grammar snippet shows these forms in action:
// a rule can "store" a parser object. They're covered // later, but for now just consider a rule as an opaque type rule<> r1, r2, r3; chlit<char> x('X'); // declare a parser named x r1 = chlit<char>('X'); // explicit declaration r2 = x; // using x r3 = ch_p('X') // using the generator
Matches a single character literal. chlit has a single template type parameter which defaults to char (i.e. chlit<> is equivalent to chlit<char>). This type parameter is the character type that chlit will deal with when parsing. As mentioned, the function generator version deduces the template type parameters from the actual function arguments. The chlit class constructor accepts a single parameter: the character it will match the input against. Examples:
r1 = chlit<>('X'); r2 = chlit<wchar_t>(L'X'); r3 = ch_p('X');
Going back to our original example:
group = '(' >> expr >> ')'; expr1 = integer | group; expr2 = expr1 >> *(('*' >> expr1) | ('/' >> expr1)); expr = expr2 >> *(('+' >> expr2) | ('-' >> expr2));
the character literals '(', ')', '+', '-', '*' and '/' in the grammar declaration are chlit objects that are implicitly created behind the scenes.
The reason this works is from two special templatized overloads of operator>> that takes a (char, ParserT), or (ParserT, char). These functions convert the character into a chlit object.
One may prefer to declare these explicitly as:
chlit<> plus('+'); chlit<> minus('-'); chlit<> times('*'); chlit<> divide('/'); chlit<> oppar('('); chlit<> clpar(')');
A range of characters is created from a low/high character pair. Such a parser matches a single character that is in the range, including both endpoints. Like chlit, range has a single template type parameter which defaults to char. The range class constructor accepts two parameters: the character range (from and to, inclusive) it will match the input against. The function generator version is range_p. Examples:
range<>('A','Z') // matches 'A'..'Z' range_p('a','z') // matches 'a'..'z'
Note, the first character must be "before" the second, according to the underlying character encoding characters. The range, like chlit is a single character parser.
Character mapping to is inherently platform dependent. It is not guaranteed in the standard for example that 'A' < 'Z', however, in many occassions, we are well aware of the character set we are using such as ASCII, ISO-8859-1 or Unicode. Take care though when porting to another platform.
This parser matches a string literal. strlit has a single template type parameter: an iterator type. Internally, strlit holds a begin/end iterator pair pointing to a string or a container of characters. The strlit attempts to match the current input stream with this string. The template type parameter defaults to char const*. strlit has two constructors. The first accepts a null-terminated character pointer. This constructor may be used to build strlits from quoted string literals. The second constructor takes in a first/last iterator pair. The function generator version is str_p. Examples:
strlit<>("Hello World") str_p("Hello World") std::string msg("Hello World"); strlit<std::string::const_iterator>(msg.begin(), msg.end());
and phrase level parsing
Typical parsers regard the processing of characters (symbols that form words or lexemes) and phrases (words that form sentences) as separate domains. Entities such as reserved words, operators, literal strings, numerical constants, etc., which constitute the terminals of a grammar are usually extracted first in a separate lexical analysis stage.
At this point, as evident in the examples we have so far, it is important to note that, contrary to standard practice, the Spirit framework handles parsing tasks at both the character level as well as the phrase level. One may consider that a lexical analyzer is seamlessly integrated in the Spirit framework.
Although the Spirit parser library does not need a separate lexical analyzer, there is no reason why we cannot have one. One can always have as many parser layers as needed. In theory, one may create a preprocessor, a lexical analyzer and a parser proper, all using the same framework.
Matches a character sequence. chseq has the same template type parameters and constructor parameters as strlit. The function generator version is chseq_p. Examples:
strlit is an implicit lexeme. That is, it works solely on the character level. chseq, strlit's twin, on the other hand, can work on both the character and phrase levels. What this simply means is that it can ignore white spaces in between the string characters. For example:
ABCDEFG A B C D E F G AB CD EFG
The framework also predefines the full repertoire of single character parsers. Unlike the ch_p and the rest of the generator functions we've seen above, these parsers are actual instantiations.
|Single character parsers|
|anychar_p||Matches any single character (including the null terminator: '\0')|
|alnum_p||Matches alpha-numeric characters|
|alpha_p||Matches alphabetic characters|
|blank_p||Matches spaces or tabs|
|cntrl_p||Matches control characters|
|digit_p||Matches numeric digits|
|graph_p||Matches non-space printing characters|
|lower_p||Matches lower case letters|
|print_p||Matches printable characters|
|punct_p||Matches punctuation symbols|
|space_p||Matches spaces, tabs, returns, and newlines|
|upper_p||Matches upper case letters|
|xdigit_p||Matches hexadecimal digits|
Single character parsers such as the chlit, range, anychar_p, alnum_p etc. can be negated. For example:
matches any character except 'x'. Double negation of a character parser cancels out the negation. ~~alpha_p is equivalent to alpha_p.
Matches the end of line (CR/LF and combinations thereof).
Never matches anything and always fails.
Matches the end of input (returns a sucessful match with 0 length when the input is exhausted)
Not strictly a primitive parser, epsilon_p, eps_p are parsers that match the null string and return a match of zero length:
epsilon_p // always returns a zero-length match
The epsilon also operates as a parser generator. In this role, they take an argument that is a 0-ary function/functor or another parser. They construct parsers that will report either an empty (zero length) match or a failure.
A failure will be reported when the function/functor result evaluates to false or when the contained parser reports a failure. Otherwise an empty match will be reported.
Operator ~ is defined for parsers constructed by epsilon_p/eps_p. It performs negation by complementing the results reported. ~~eps_p(x) is identical to eps_p(x).
epsilon_p('0') >> oct_p // note that '0' is actually a ch_p('0')
Epsilon here is used as a syntactic predicate. oct_p is parsed only if we see a leading '0'. Wrapping the leading '0' inside an epsilon makes the parser not consume anything from the input. If a '0' is seen, epsilon_p reports a successful match with zero length. We shall learn more about oct_p when we get to the section on numerics. Suffice it to say that it is a primitive parser that parses octal numbers.
Copyright © 1998-2003 Joel de Guzman
Copyright © 2003 Martin Wille
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)