To get an overview of how to use Canopy with Python, consider this example of a simplified grammar for URLs:

grammar URL
  url       <-  scheme "://" host pathname search hash?
  scheme    <-  "http" "s"?
  host      <-  hostname port?
  hostname  <-  segment ("." segment)*
  segment   <-  [a-z0-9-]+
  port      <-  ":" [0-9]+
  pathname  <-  "/" [^ ?]*
  search    <-  ("?" query:[^ #]*)?
  hash      <-  "#" [^ ]*

We can compile this grammar into a Python module using canopy:

$ canopy url.peg --lang python

This creates a file called that contains all the parser logic, and it works in Node and in the browser. The --output option can be used to override the default location:

$ canopy url.peg --lang python --output some/dir/url

This will write the generated parser into the file some/dir/

Let’s try our parser out:

import url

tree = url.parse('')

for node in tree.elements:
    print node.offset, node.text

#   prints:

#   0 http
#   4 ://
#   7
#   18 /search
#   25 ?q=hello
#   33 #page=1

This little example shows a few important things:

You invoke the parser by calling the module’s parse() function with a string.

The parse() method returns a tree of nodes.

Each node has three properties:

Walking the parse tree

You can use elements to walk into the structure of the tree:

print tree.elements[4].elements[1].text
# -> 'q=hello'

Or, you can use the labels that Canopy generates, which can make your code clearer:

# -> 'q=hello'

Parsing errors

If you give the parser an input text that does not match the grammar, a url.ParseError is thrown. The error message will list any of the strings or character classes the parser was expecting to find at the furthest position it got to, along with the rule those expectations come from, and it will highlight the line of the input where the syntax error occurs.


# url.ParseError: Line 1: expected one of:
#     - [a-z0-9-] from URL::segment
#      1 |
#                              ^

Implementing actions

Say you have a grammar that uses action annotations, for example:

grammar Maps
  map     <-  "{" string ":" value "}" %make_map
  string  <-  "'" [^']* "'" %make_string
  value   <-  list / number
  list    <-  "[" value ("," value)* "]" %make_list
  number  <-  [0-9]+ %make_number

In Python, you give the action functions to the parser by using the actions keyword argument, which should be an object implementing the named actions:

import maps

class Actions(object):
    def make_map(self, input, start, end, elements):
        return {elements[1]: elements[3]}

    def make_string(self, input, start, end, elements):
        return elements[1].text

    def make_list(self, input, start, end, elements):
        list = [elements[1]]
        for el in elements[2]:
        return list

    def make_number(self, input, start, end, elements):
        return int(input[start:end], 10)

result = maps.parse("{'ints':[1,2,3]}", actions=Actions())

print result
# -> {'ints': [1, 2, 3]}

Extended node types

Say you have a grammar that contains type annotations:

grammar Words
  root  <-  first:"foo" second:"bar" <Extension>

To use this parser, you must pass in an object containing implementations of the named types via the types option. Each defined type contains the methods that will be added to the nodes.

You can import the types from a module:


class Extension(object):
    def convert(self):
        return self.first.text + self.second.text.upper()


import words
import node_types

words.parse('foobar', types=node_types).convert()
# -> 'fooBAR'

Or, you can enclose the extension classes in another class that you pass to the parser:

import words

class Types:
    class Extension(object):
        def convert(self):
            return self.first.text + self.second.text.upper()

words.parse('foobar', types=Types).convert()
# -> 'fooBAR'