pegomancy 1.2.0

Yet another parsing thingy

pegomancy

Yet another parsing thingy

Installation

This project is written using Python 3.8. It can be installed using setuptools or pip.

Usage

As a tool

Pegomancy provides a pegomant executable that can be used to generate Python code to parse data according to a given grammar specification.

usage: pegomant [-h] [-c CLASS_NAME] [-o OUTPUT_FILE] grammar_file

positional arguments:
  grammar_file

optional arguments:
  -h, --help            show this help message and exit
  -c CLASS_NAME, --class_name CLASS_NAME
  -o OUTPUT_FILE, --output-file OUTPUT_FILE

As a library

The pegomancy module can be used as a library to fully control grammars and how code is generated.

Grammar syntax

Pegomancy grammars look like regular PEG grammars, with a dash of sugar syntax. Here is an example grammar specification that can be used to parse arithmetic expressions:

integer: r"[0-9]+"

expr: left:expr op:'+' right:term
    | left:expr op:'-' right:term
    | term

term: left:term op:'*' right:atom
    | left:term op:'/' right:atom
    | atom

atom: integer | '(' expr ')'

In case the syntax isn't familiar, let's provide a bit of information.

Rules

A grammar is made of one or several rules. Each rule is specified with the following syntax:

rule_name: expression

In the grammar above, integer, expr, term and atom are the rules.

Expressions

Atoms

Atoms are the most primitive constructs used in an expression: they match a simple portion of the source text.

In the example grammar given above, we have different kinds of atoms:

• '(', ')', '+' (and others), each matching a raw string of text
• r"[0-9]+", matching a regular expression
• integer (in the atom rule), matching what the integer rule matches

Note that atoms can be concatenated: '(' expr ')' will match an opening parenthesis, then what the expr rule matches, then a closing parenthesis.

Items

Items are more complex expressions and introduce modifiers to repeat or make expressions optional:

• the * operator can be used to allow repeating an expression zero or more times
• the + operator can be used to allow repeating an expression one or more times
• the ? operator can be used to make an expression optional

Items are unnamed by default, but can be named using the : operator, as in op:'+', which gives the name op to the '+' atom.

Alternatives

Some rules might allow multiple possibilities: for example, the atom rule in the above grammar can match either an integer or a parenthesized expression. The notion of alternative is expressed in the grammar using the | operator.

Parse results

Default AST

By default, parsers generated with pegomant will produce AST nodes that are either:

• a single value, if the matched expression has only one component
• a list, if the matched expression has multiple components
• a dictionary, if the matched expression has named items (unnamed items are discarded from the result)

Customizing the AST

The default AST can be enough, but in some cases it is useful to transform it into a custom data structure. When using a Pegomancy parser, you can specify an object whose methods will be invoked when a rule matches some input.

That object must be given as the rule_handler parameter when initializing the parser.

Here is a possible rule handler for the example grammar given in previous sections.

class RuleHandler:
    def integer(self, node):
        return int(node)

    def expr(self, node):
        if isinstance(node, dict):
            if node["op"] == "+":
                return node["left"] + node["right"]
            else:
                return node["left"] - node["right"]
        else:
            return node

    def term(self, node):
        if isinstance(node, dict):
            if node["op"] == "*":
                return node["left"] * node["right"]
            else:
                return node["left"] / node["right"]
        else:
            return node

    def atom(self, node):
        if isinstance(node, list):
            return node[1]
        return node

The methods can raise a ParseError to indicate a parse failure for the rule being handled.