internal package Foswiki::Infix::Parser

See PublishedAPI for packages intended to be used by Plugin and Contrib authors, or browse all packages.
See also Developing plugins, Developer's Bible, Technical Overview

internal package Foswiki::Infix::Parser

A simple LL(1) parser that parses infix expressions with nonary, unary and binary operators specified using an operator table.

The parser works by examining each token in the input stream from left to right, and constructs parse nodes as soon as they are identified. The parser doesn't dictate the type of the parse nodes, instead using a factory to generate them. the output from the parser is a parse tree built using nodes generated by the node factory.

Escapes are supported in strings, using backslash.

ClassMethod new($node_factory, \%options) → $parser_object

Creates a new infix parser. Operators must be added for it to be useful.

The tokeniser matches tokens in the following order: operators, quotes (" and '), numbers, words, brackets. If you have any overlaps (e.g. an operator '<' and a bracket operator '<<') then the first choice will match.

$node_factory needs to be ( the name of a package | an object ) that supports the following two functions:
  • newLeaf($val, $type) - create a terminal. $type will be:
    1. if the terminal matched the words specification (see below).
    2. if it is a number matched the numbers specification (see below)
    3. if it is a quoted string
  • newNode($op, @params) - create a new operator node. @params is a variable-length list of parameters, left to right. $op is a reference to the operator hash in the \@opers list.
These functions should throw Error::Simple in the event of errors. Foswiki::Infix::Node is such a class, ripe for subclassing.

The remaining parameters are named, and specify options that affect the behaviour of the parser:
  1. words=>qr// - should be an RE specifying legal words (unquoted terminals that are not operators i.e. names and numbers). By default this is \w+. It's ok if operator names match this RE; operators always have precedence over names.
  2. numbers=>qr// - should be an RE specifying legal numbers (unquoted terminals that are not operators or words). By default this is qr/[+-]?(?:\d+\.\d+|\d+\.|\.\d+|\d+)(?:[eE][+-]?\d+)?/, which matches integers and floating-point numbers. Number matching always takes precedence over word matching (i.e. "1xy" will be parsed as a number followed by a word. A typical usage of this option is when you only want to recognise integers, in which case you would set this to numbers => qr/\d+/.
Strings should always be surrounded by 'single-quotes'. Single quotes in values may be escaped using backslash (\).

ObjectMethod addOperator($oper)

Add an operator to the parser.

$oper is an object that implements the Foswiki::Infix::OP interface.

ObjectMethod parse($string) → $parseTree

Parses $string, calling newLeaf and newNode in the client class as necessary to create a parse tree. Returns the result of calling newNode on the root of the parse.

Throws Foswiki::Infix::Error in the event of parse errors.

ObjectMethod onCloseExpr($@opands)

Designed to be overridden by subclasses that need to perform an action on the operand stack (such as pushing) when a sub-expression is closed. Also called when the root expression is closed. The default is a no-op.

Topic revision: r1 - 03 May 2016, UnknownUser
This site is powered by FoswikiCopyright © by the contributing authors. All material on this site is the property of the contributing authors.
Ideas, requests, problems regarding sms4science? Send feedback