1. Neogeny
  2. apalala
  3. grako



fury license pyversions travis landscape

At least for the people who send me mail about a new language that they're designing, the general advice is: do it to learn about how to write a compiler. Don't have any expectations that anyone will use it, unless you hook up with some sort of organization in a position to push it hard. It's a lottery, and some can buy a lot of the tickets. There are plenty of beautiful languages (more beautiful than C) that didn't catch on. But someone does win the lottery, and doing a language at least teaches you something.

Dennis Ritchie (1941-2011) Creator of the C programming language and of Unix



If you're interested in the technology, and in future developments such as proper support for left recursion, then please follow the Tatsu project.

Grako (for grammar compiler) is a tool that takes grammars in a variation of EBNF as input, and outputs memoizing (Packrat) PEG parsers in Python.

Grako can also compile a grammar stored in a string into a grako.grammars.Grammar object that can be used to parse any given input, much like the re module does with regular expressions.

Grako is different from other PEG parser generators:

  • Generated parsers use Python's very efficient exception-handling system to backtrack. Grako generated parsers simply assert what must be parsed. There are no complicated if-then-else sequences for decision making or backtracking. Memoization allows going over the same input sequence several times in linear time.
  • Positive and negative lookaheads, and the cut element (with its cleaning of the memoization cache) allow for additional, hand-crafted optimizations at the grammar level.
  • Delegation to Python's re module for lexemes allows for (Perl-like) powerful and efficient lexical analysis.
  • The use of Python's context managers considerably reduces the size of the generated parsers for code clarity, and enhanced CPU-cache hits.
  • Include files, rule inheritance, and rule inclusion give Grako grammars considerable expressive power.
  • Automatic generation of Abstract Syntax Trees_ and Object Models, along with Model Walkers and Code Generators make analysis and translation approachable

The parser generator, the run-time support, and the generated parsers have measurably low Cyclomatic complexity. At around 5 KLOC of Python, it is possible to study all its source code in a single session.

The only dependencies are on the Python standard library, yet the regex library will be used if installed, and colorama will be used on trace output if available. pygraphviz is required for generating diagrams.

Grako is feature-complete and currently being used with complex grammars to parse, analyze, and translate hundreds of thousands of lines of input text, including source code in several programming languages.

Table of Contents


Grako was created to address some recurring problems encountered over decades of working with parser generation tools:

  • Some programming languages allow the use of keywords as identifiers, or have different meanings for symbols depending on context (Ruby). A parser needs control of lexical analysis to be able to handle those languages.
  • LL and LR grammars become contaminated with myriads of lookahead statements to deal with ambiguous constructs in the source language. PEG parsers address ambiguity from the onset.
  • Separating the grammar from the code that implements the semantics, and using a variation of a well-known grammar syntax (EBNF) allows for full declarative power in language descriptions. General-purpose programming languages are not up to the task.
  • Semantic actions do not belong in a grammar. They create yet another programming language to deal with when doing parsing and translation: the source language, the grammar language, the semantics language, the generated parser's language, and the translation's target language. Most grammar parsers do not check the syntax of embedded semantic actions, so errors get reported at awkward moments, and against the generated code, not against the grammar.
  • Preprocessing (like dealing with includes, fixed column formats, or structure-through-indentation) belongs in well-designed program code; not in the grammar.
  • It is easy to recruit help with knowledge about a mainstream programming language like Python, but help is hard to find for working with complex grammar-description languages. Grako grammars are in the spirit of a Translators and Interpreters 101 course (if something is hard to explain to a college student, it's probably too complicated, or not well understood).
  • Generated parsers should be easy to read and debug by humans. Looking at the generated source code is sometimes the only way to find problems in a grammar, the semantic actions, or in the parser generator itself. It's inconvenient to trust generated code that one cannot understand.
  • Python is a great language for working with language parsing and translation.

The Generated Parsers

A Grako generated parser consists of the following classes:

  • A MyLanguageBuffer class derived from grako.buffering.Buffer that handles the grammar definitions for whitespace, comments, and case significance.
  • A MyLanguageParser class derived from grako.parsing.Parser which uses a MyLanguageBuffer for traversing input text, and implements the parser using one method for each grammar rule:
        def _somerulename_(self):
  • A MyLanguageSemantics class with one semantic method per grammar rule. Each method receives as its single parameter the Abstract Syntax Tree (AST) built from the rule invocation:
        def somerulename(self, ast):
            return ast
  • A if __name__ == '__main__': definition, so the generated parser can be executed as a Python script.

The methods in the delegate class return the same AST received as parameter, but custom semantic classes can override the methods to have them return anything (for example, a Semantic Graph). The semantics class can be used as a template for the final semantics implementation, which can omit methods for the rules that do not need semantic treatment.

If present, a _default() method will be called in the semantics class when no method matched the rule name:

def _default(self, ast):
    return ast

If present, a _postproc() method will be called in the semantics class after each rule (including the semantics) is processed. This method will receive the current parsing context as parameter:

def _postproc(self, context, ast):

Using the Tool

As a Library

Grako can be used as a library, much like Python's re, by embedding grammars as strings and generating grammar models instead of generating code.

  • grako.compile(grammar, name=None, **kwargs)

    Compiles the grammar and generates a model that can subsequently be used for parsing input with.

  • grako.parse(grammar, input, **kwargs)

    Compiles the grammar and parses the given input producing an AST as result. The result is equivalent to calling model = compile(grammar); model.parse(input). Compiled grammars are cached for efficiency.

  • grako.to_python_sourcecode(grammar, name=None, filename=None, **kwargs)

    Compiles the grammar to the Python sourcecode that implements the parser.

This is an example of how to use Grako as a library:


    start = expression $ ;

        | term '+' ~ expression
        | term '-' ~ expression
        | term

        | factor '*' ~ term
        | factor '/' ~ term
        | factor

        | '(' ~ @:expression ')'
        | number

    number = /\d+/ ;

def main():
    import pprint
    import json
    from grako import parse
    from grako.util import asjson

    ast = parse(GRAMMAR, '3 + 5 * ( 10 - 20 )')
    pprint.pprint(ast, indent=2, width=20)

    json_ast = asjson(ast)
    print(json.dumps(json_ast, indent=2))

if __name__ == '__main__':

And this is the output:

[ '3',
  [ '5',
    [ '10',


Compiling grammars to Python

Grako can be run from the command line:

$ python -m grako


$ scripts/grako

Or just:

$ grako

if Grako was installed using easy_install or pip.

The -h and --help parameters provide full usage information:

$ python -m grako -h
usage: grako [--generate-parser | --draw | --object-model | --pretty]
            [--color] [--trace] [--no-left-recursion] [--name NAME]
            [--no-nameguard] [--outfile FILE] [--object-model-outfile FILE]
            [--whitespace CHARACTERS] [--help] [--version]

Grako (for "grammar compiler") takes a grammar in a variation of EBNF as
input, and outputs a memoizing PEG/Packrat parser in Python.

positional arguments:
GRAMMAR               the filename of the Grako grammar to parse

optional arguments:
--generate-parser     generate parser code from the grammar (default)
--draw, -d            generate a diagram of the grammar (requires --outfile)
--object-model, -g    generate object model from the class names given as
                        rule arguments
--pretty, -p          generate a prettified version of the input grammar

parse-time options:
--color, -c           use color in traces (requires the colorama library)
--trace, -t           produce verbose parsing output

generation options:
--no-left-recursion, -l
                        turns left-recusion support off
--name NAME, -m NAME  Name for the grammar (defaults to GRAMMAR base name)
--no-nameguard, -n    allow tokens that are prefixes of others
--outfile FILE, --output FILE, -o FILE
                        output file (default is stdout)
--object-model-outfile FILE, -G FILE
                        generate object model and save to FILE
                        characters to skip during parsing (use "" to disable)

common options:
--help, -h            show this help message and exit
--version, -v         provide version information and exit

Using the Generated Parser

To use the generated parser, just subclass the base or the abstract parser, create an instance of it, and invoke its parse() method passing the grammar to parse and the starting rule's name as parameter:

from myparser import MyParser

parser = MyParser()
ast = parser.parse('text to parse', rule_name='start')
print(json.dumps(ast, indent=2)) # ASTs are JSON-friendy

The generated parsers' constructors accept named arguments to specify whitespace characters, the regular expression for comments, case sensitivity, verbosity, and more (see below).

To add semantic actions, just pass a semantic delegate to the parse method:

model = parser.parse(text, rule_name='start', semantics=MySemantics())

If special lexical treatment is required (as in 80 column languages), then a descendant of grako.buffering.Buffer can be passed instead of the text:

class MySpecialBuffer(MyLanguageBuffer):

buf = MySpecialBuffer(text)
model = parser.parse(buf, rule_name='start', semantics=MySemantics())

The generated parser's module can also be invoked as a script:

$ python myparser.py inputfile startrule

As a script, the generated parser's module accepts several options:

$ python myparser.py -h
usage: myparser.py [-h] [-c] [-l] [-n] [-t] [-w WHITESPACE] FILE [STARTRULE]

Simple parser for DBD.

positional arguments:
    FILE                  the input file to parse
    STARTRULE             the start rule for parsing

optional arguments:
    -h, --help            show this help message and exit
    -c, --color           use color in traces (requires the colorama library)
    -l, --list            list all rules and exit
    -n, --no-nameguard    disable the 'nameguard' feature
    -t, --trace           output trace information
    -w WHITESPACE, --whitespace WHITESPACE
                        whitespace specification

Grammar Syntax

Grako uses a variant of the standard EBNF syntax. Syntax definitions for VIM and for Sublime Text can be found under the etc/vim and etc/sublime directories in the source code distribution.


A grammar consists of a sequence of one or more rules of the form:

name = <expre> ;

If a name collides with a Python keyword, an underscore (_) will be appended to it on the generated parser.

Rule names that start with an uppercase character:

FRAGMENT = /[a-z]+/ ;

do not advance over whitespace before beginning to parse. This feature becomes handy when defining complex lexical elements, as it allows breaking them into several rules.

The parser returns an AST value for each rule depending on what was parsed:

  • A single value
  • A list of AST
  • A dict-like object for rules with named elements
  • An object, when ModelBuilderSemantics is used
  • None

See the Abstract Syntax Trees and Building Models sections for more details.


The expressions, in reverse order of operator precedence, can be:

e1 | e2

Choice. Match either e1 or e2.

A | be be used before the first option if desired:

    | e1
    | e2
    | e3

e1 e2

Sequence. Match e1 and then match e2.

( e )

Grouping. Match e. For example: ('a' | 'b').

[ e ]

Optionally match e.

{ e } or { e }*

Closure. Match e zero or more times. Note that the AST returned for a closure is always a list.

{ e }+

Positive closure. Match e one or more times. The AST is always a list.


Empty closure. Match nothing and produce an empty list as AST.


The cut expression. Commit to the current option and prevent other options from being considered even if what follows fails to parse.

In this example, other options won't be considered if a parenthesis is parsed:

      '(' ~ @:expre ')'
    | int
    | bool

s%{ e }+

Positive join. Inspired by Python's str.join(), it parses the same as this expression:

e {s ~ e}

yet the result is a single list of the form:

[e, s, e, s, e....]

Use grouping if s is more complex than a token or a pattern:

(s t)%{ e }+

s%{ e } or s%{ e }*

Join. Parses the list of s-separated expressions, or the empty closure.

It is equivalent to:


op<{ e }+

Left join. Like the join expression, but the result is a left-associative tree built with tuple(), in wich the first elelemnt is the separator (op), and the other two elements are the operands.

The expression:


Will parse this input:

1 + 2 + 3 + 4

To this tree:


op>{ e }+

Right join. Like the join expression, but the result is a right-associative tree built with tuple(), in wich the first elelemnt is the separator (op), and the other two elements are the operands.

The expression:


Will parse this input:

1 + 2 + 3 + 4

To this tree:


s.{ e }+

Positive gather. Like positive join, but the separator is not included in the resulting AST.

s.{ e } or s.{ e }*

Gather. Like the join, but the separator is not included in the resulting AST.

It is equivalent to:



Positive lookahead. Succeed if e can be parsed, but do not consume any input.


Negative lookahead. Fail if e can be parsed, and do not consume any input.

'text' or "text"

Match the token text within the quotation marks.

Note that if text is alphanumeric, then Grako will check that the character following the token is not alphanumeric. This is done to prevent tokens like IN matching when the text ahead is INITIALIZE. This feature can be turned off by passing nameguard=False to the Parser or the Buffer, or by using a pattern expression (see below) instead of a token expression. Alternatively, the @@nameguard or @@namechars directives may be specified in the grammar:

@@nameguard :: False

or to specify additional characters that should also be considered part of names:

@@namechars :: '$-.'

r'text' or r"text"

Match the token text within the quotation marks, interpreting text like Python's raw string literals.

?"regexp" or ?'regexp'

The pattern expression. Match the Python regular expression regexp at the current text position. Unlike other expressions, this one does not advance over whitespace or comments. For that, place the regexp as the only term in its own rule.

The regex is interpreted as a Python's raw string literal and passed as-is to the Python re module (or to regex, if available), using match() at the current position in the text. The matched text is the AST for the expression.

Consecutive patterns are concatenated to form a single one.

  • /regexp/
Another form of the pattern expression.
  • +/regexp/
Concatenate the given pattern with the preceding one.


Match nothing, but behave as if constant had been parsed.

Constants can be used to inject elements into the concrete and abstract syntax trees, perhaps avoiding having to write a semantic action. For example:

boolean_option = name ['=' (boolean|`true`) ] ;


Invoke the rule named rulename. To help with lexical aspects of grammars, rules with names that begin with an uppercase letter will not advance the input over whitespace or comments.


The include operator. Include the right hand side of rule rulename at this point.

The following set of declarations:

includable = exp1 ;

expanded = exp0 >includable exp2 ;

Has the same effect as defining expanded as:

expanded = exp0 exp1 exp2 ;

Note that the included rule must be defined before the rule that includes it.


The empty expression. Succeed without advancing over input. Its value is None.


The fail expression. This is actually ! applied to (), which always fails.


Add the result of e to the AST using name as key. If name collides with any attribute or method of dict, or is a Python keyword, an underscore (_) will be appended to the name.


Add the result of e to the AST using name as key. Force the entry to be a list even if only one element is added. Collisions with dict attributes or Python keywords are resolved by appending an underscore to name.


The override operator. Make the AST for the complete rule be the AST for e.

The override operator is useful to recover only part of the right hand side of a rule without the need to name it, or add a semantic action.

This is a typical use of the override operator:

subexp = '(' @:expre ')' ;

The AST returned for the subexp rule will be the AST recovered from invoking expre.


Like @:e, but make the AST always be a list.

This operator is convenient in cases such as:

arglist = '(' @+:arg {',' @+:arg}* ')' ;

In which the delimiting tokens are of no interest.


The end of text symbol. Verify that the end of the input text has been reached.

# comment

Python-style comments are also allowed.

When there are no named items in a rule, the AST consists of the elements parsed by the rule, either a single item or a list. This default behavior makes it easier to write simple rules:

number = /[0-9]+/ ;

Without having to write:

number = number:/[0-9]+/ ;

When a rule has named elements, the unnamed ones are excluded from the AST (they are ignored).

Deprecated Expressions

The following expressions are still recognized in grammars, but they are considered deprecated, and will be removed in a future version of Grako.

  • ?/regexp/?
Another form of the pattern expression that can be used when there are slashes (/) in the pattern. Use the ?"regexp" or ?'regexp' forms instead.
  • (* comment *)
Comments may appear anywhere in the text. Use the Python-style comments instead.

Rules with Arguments

Grako allows rules to specify Python-style arguments:

addition(Add, op='+')
    addend '+' addend

The arguments values are fixed at grammar-compilation time.

An alternative syntax is available if no keyword parameters are required:

addition::Add, '+'
    addend '+' addend

Semantic methods must be ready to receive any arguments declared in the corresponding rule:

def addition(self, ast, name, op=None):

When working with rule arguments, it is good to define a _default() method that is ready to take any combination of standard and keyword arguments:

def _default(self, ast, *args, **kwargs):

Based Rules

Rules may extend previously defined rules using the < operator. The base rule must be defined previously in the grammar.

The following set of declarations:

base::Param = exp1 ;

extended < base = exp2 ;

Has the same effect as defining extended as:

extended::Param = exp1 exp2 ;

Parameters from the base rule are copied to the new rule if the new rule doesn't define its own. Repeated inheritance should be possible, but it hasn't been tested.

Rule Overrides

A grammar rule may be redefined by using the @override decorator:

start = ab $;

ab = 'xyz' ;

ab = @:'a' {@:'b'} ;

When combined with the #include directive, rule overrides can be used to create a modified grammar without altering the original.

Abstract Syntax Trees (ASTs)

By default, and AST is either a list (for closures and rules without named elements), or dict-derived object that contains one item for every named element in the grammar rule. Items can be accessed through the standard dict syntax (ast['key']), or as attributes (ast.key).

AST entries are single values if only one item was associated with a name, or lists if more than one item was matched. There's a provision in the grammar syntax (the +: operator) to force an AST entry to be a list even if only one element was matched. The value for named elements that were not found during the parse (perhaps because they are optional) is None.

When the parseinfo=True keyword argument has been passed to the Parser constructor, a parseinfo element is added to AST nodes that are dict-like. The element contains a collections.namedtuple with the parse information for the node:

ParseInfo = namedtuple(

With the help of the Buffer.line_info() method, it is possible to recover the line, column, and original text parsed for the node. Note that when ParseInfo is generated, the Buffer used during parsing is kept in memory for the lifetime of the AST.

Generation of parseinfo can also be controlled using the @@parseinfo :: True grammar directive.

Grammar Name

The prefix to be used in classes generated by Grako can be passed to the command-line tool using the -m option:

$ grako -m MyLanguage mygrammar.ebnf

will generate:

class MyLanguageParser(Parser):

The name can also be specified within the grammar using the @@grammar directive:

@@grammar :: MyLanguage


By default, Grako generated parsers skip the usual whitespace characters with the regular expression r'\s+' using the re.UNICODE flag (or with the Pattern_White_Space property if the regex module is available), but you can change that behavior by passing a whitespace parameter to your parser.

For example, the following will skip over tab (\t) and space characters, but not so with other typical whitespace characters such as newline (\n):

parser = MyParser(text, whitespace='\t ')

The character string is converted into a regular expression character set before starting to parse.

You can also provide a regular expression directly instead of a string. The following is equivalent to the above example:

parser = MyParser(text, whitespace=re.compile(r'[\t ]+'))

Note that the regular expression must be pre-compiled to let Grako distinguish it from plain string.

If you do not define any whitespace characters, then you will have to handle whitespace in your grammar rules (as it's often done in PEG parsers):

parser = MyParser(text, whitespace='')

Whitespace may also be specified within the grammar using the @@whitespace directive, although any of the above methods will overwrite the setting in the grammar:

@@whitespace :: /[\t ]+/

Case Sensitivity

If the source language is case insensitive, it can be specified in the parser by using the ignorecase parameter:

parser = MyParser(text, ignorecase=True)

You may also specify case insensitivity within the grammar using the @@ignorecase directive:

@@ignorecase :: True

The change will affect both token and pattern matching.


Parsers will skip over comments specified as a regular expression using the comments_re parameter:

parser = MyParser(text, comments_re="\(\*.*?\*\)")

For more complex comment handling, you can override the Buffer.eat_comments() method.

For flexibility, it is possible to specify a pattern for end-of-line comments separately:

parser = MyParser(

Both patterns may also be specified within a grammar using the @@comments and @@eol_comments directives:

@@comments :: /\(\*.*?\*\)/
@@eol_comments :: /#.*?$/

Reserved Words and Keywords

Some languages must reserve the use of certain tokens as valid identifiers because the tokens are used to mark particular constructs in the language. Those reserved tokens are known as Reserved Words or Keywords

Grako provides support for preventing the use of keywords as identifiers though the @@ keyword directive,and the @ name decorator.

A grammar may specify reserved tokens providing a list of them in one or more @@ keyword directives:

@@keyword :: if endif
@@keyword :: else elseif

The @ name decorator checks that the result of a grammar rule does not match a token defined as a keyword:

identifier = /(?!\d)\w+/ ;

There are situations in which a token is reserved only in a very specific context. In those cases, a negative lookahead will prevent the use of the token:

statements = {!'END' statement}+ ;

Semantic Actions

There are no constructs for semantic actions in Grako grammars. This is on purpose, because semantic actions obscure the declarative nature of grammars and provide for poor modularization from the parser-execution perspective.

Semantic actions are defined in a class, and applied by passing an object of the class to the parse() method of the parser as the semantics= parameter. Grako will invoke the method that matches the name of the grammar rule every time the rule parses. The argument to the method will be the AST constructed from the right-hand-side of the rule:

class MySemantics(object):
    def some_rule_name(self, ast):
        return ''.join(ast)

    def _default(self, ast):

If there's no method matching the rule's name, Grako will try to invoke a _default() method if it's defined:

def _default(self, ast):

Nothing will happen if neither the per-rule method nor _default() are defined.

The per-rule methods in classes implementing the semantics provide enough opportunity to do rule post-processing operations, like verifications (for inadequate use of keywords as identifiers), or AST transformation:

class MyLanguageSemantics(object):
    def identifier(self, ast):
        if my_lange_module.is_keyword(ast):
            raise FailedSemantics('"%s" is a keyword' % str(ast))
        return ast

For finer-grained control it is enough to declare more rules, as the impact on the parsing times will be minimal.

If preprocessing is required at some point, it is enough to place invocations of empty rules where appropriate:

myrule = first_part preproc {second_part} ;

preproc = () ;

The abstract parser will honor as a semantic action a method declared as:

def preproc(self, ast):

Include Directive

Grako grammars support file inclusion through the include directive:

#include :: "filename"

The resolution of the filename is relative to the directory/folder of the source. Absolute paths and ../ navigations are honored.

The functionality required for implementing includes is available to all Grako-generated parsers through the Buffer class; see the EBNFBuffer class in the grako.parser module for an example.

Building Models

Naming elements in grammar rules makes the parser discard uninteresting parts of the input, like punctuation, to produce an Abstract Syntax Tree (AST) that reflects the semantic structure of what was parsed. But an AST doesn't carry information about the rule that generated it, so navigating the trees may be difficult.

Grako defines the grako.model.ModelBuilderSemantics semantics class which helps construct object models from abtract syntax trees:

from grako.model import ModelBuilderSemantics

parser = MyParser(semantics=ModelBuilderSemantics())

Then you add the desired node type as first parameter to each grammar rule:

addition::AddOperator = left:mulexpre '+' right:addition ;

ModelBuilderSemantics will synthesize a class AddOperator(Node): class and use it to construct the node. The synthesized class will have one attribute with the same name as the named elements in the rule.

You can also use Python's built-in types as node types, and ModelBuilderSemantics will do the right thing:

integer::int = /[0-9]+/ ;

ModelBuilderSemantics acts as any other semantics class, so its default behavior can be overidden by defining a method to handle the result of any particular grammar rule.

Walking Models

The class grako.model.NodeWalker allows for the easy traversal (walk) a model constructed with a ModelBuilderSemantics instance:

from grako.model import NodeWalker

class MyNodeWalker(NodeWalker):

    def walk_AddOperator(self, node):
        left = self.walk(node.left)
        right = self.walk(node.right)

        print('ADDED', left, right)

model = MyParser(semantics=ModelBuilderSemantics()).parse(input)

walker = MyNodeWalker()

When a method with a name like walk_AddOperator() is defined, it will be called when a node of that type is walked (the pythonic version of the class name may also be used for the walk method: walk_add_operator().

If a walk method for a node class is not found, then a method for the class's bases is searched, so it is possible to write catch-all methods such as:

def walk_Node(self, node):
    print('Reached Node', node)

def walk_str(self, s):
    return s

def walk_object(self, o):
    raise Exception('Unexpected tyle %s walked', type(o).__name__)

Predeclared classes can be passed to ModelBuilderSemantics instances through the types= parameter:

from mymodel import AddOperator, MulOperator

semantics=ModelBuilderSemantics(types=[AddOperator, MulOperator])

ModelBuilderSemantics assumes nothing about types=, so any constructor (a function, or a partial function) can be used.

Model Class Hierarchies

It is possible to specify a a base class for generated model nodes:

    | addition
    | substraction

    left:mulexpre op:'+' right:additive

    left:mulexpre op:'-' right:additive

Grako will generate the base class if it's not already known.

Base classes can be used as the target class in walkers, and in code generators:

class MyNodeWalker(NodeWalker):
    def walk_Operator(self, node):
        left = self.walk(node.left)
        right = self.walk(node.right)
        op = self.walk(node.op)

        print(type(node).__name__, op, left, right)

class Operator(ModelRenderer):
    template = '{left} {op} {right}'

Templates and Translation


As of Grako 3.2.0, code generation is separated from grammar models through grako.codegen.CodeGenerator as to allow for code generation targets different from Python. Still, the use of inline templates and rendering.Renderer hasn't changed. See the regex example for merged modeling and code generation.

Grako doesn't impose a way to create translators with it, but it exposes the facilities it uses to generate the Python source code for parsers.

Translation in Grako is template-based, but instead of defining or using a complex templating engine (yet another language), it relies on the simple but powerful string.Formatter of the Python standard library. The templates are simple strings that, in Grako's style, are inlined with the code.

To generate a parser, Grako constructs an object model of the parsed grammar. A grako.codegen.CodeGenerator instance matches model objects to classes that descend from grako.codegen.ModelRenderer and implement the translation and rendering using string templates. Templates are left-trimmed on whitespace, like Python doc-comments are. This is an example taken from Grako's source code:

class Lookahead(ModelRenderer):
    template = '''\
                with self._if():

Every attribute of the object that doesn't start with an underscore (_) may be used as a template field, and fields can be added or modified by overriding the render_fields(fields) method. Fields themselves are lazily rendered before being expanded by the template, so a field may be an instance of a ModelRenderer descendant.

The rendering module defines a Formatter enhanced to support the rendering of items in an iterable one by one. The syntax to achieve that is:


All of ind, sep, and fmt are optional, but the three colons are not. A field specified that way will be rendered using:

indent(sep.join(fmt % render(v) for v in value), ind)

The extended format can also be used with non-iterables, in which case the rendering will be:

indent(fmt % render(value), ind)

The default multiplier for ind is 4, but that can be overridden using n*m (for example 3*1) in the format.


Using a newline character (\n) as separator will interfere with left trimming and indentation of templates. To use a newline as separator, specify it as \\n, and the renderer will understand the intention.

Left Recursion

Grako provides experimental support for left recursion in PEG grammars. The implementation of left recursion is ongoing; it does not yet handle all cases. The algorithm used is Warth et al's.

Sometimes, while debugging a grammar, it is useful to turn left-recursion support on or off:

parser = MyParser(

Left recursion can also be turned off from within the grammar using the @@left_recursion directive:

@@left_recursion :: False



The file etc/grako.ebnf contains a grammar for the Grako grammar language written in its own grammar language. It is used in the bootstrap test suite to prove that Grako can generate a parser to parse its own language, and the resulting parser is made the bootstrap parser every time Grako is stable (see grako/bootstrap.py for the generated parser).

Grako uses Grako to translate grammars into parsers, so it is a good example of end-to-end translation.


The project examples/regexp contains a regexp-to-EBNF translator and parser generator. The project has no practical use, but it's a complete, end-to-end example of how to implement a translator using Grako.


The project examples/calc implements a calculator for simple expressions, and is written as a tutorial over most of the features provided by Grako.


The project examples/antlr2grako contains a ANTLR to Grako grammar translator. The project is a good example of the use of models and templates in translation. The program, antlr2grako.py generates the Grako grammar on standard output, but because the model used is Grako's own, the same code can be used to directly generate a parser from an ANTLR grammar. Please take a look at the examples README to know about limitations.

Other open-source Examples

  • Christian Ledermann wrote parsewkt a parser for [Well-known text] (WTK) using Grako.
  • Marcus Brinkmann (lambdafu) wrote smc.mw, a parser for a MediaWiki-style language.
  • Marcus Brinkmann (lambdafu) is working on a C++ code generator for Grako called Grako++. Help in the form of testing, test cases, and pull requests is welcome.


Copyright (C) 2017      by Juancarlo Añez
Copyright (C) 2012-2016 by Juancarlo Añez and Thomas Bragg

You may use Grako under the terms of the BSD-style license described in the enclosed LICENSE.txt file. If your project requires different licensing please email.

Contact and Updates

For general Q&A, please use the [grako] tag on StackOverflow.

To discuss Grako and to receive notifications about new releases, please join the low-volume Grako Forum at Google Groups.

You can also follow the latest Grako developments with @GrakoPEG on Twitter.


The following must be mentioned as contributors of thoughts, ideas, code, and funding to the Grako project:

  • Niklaus Wirth was the chief designer of the programming languages Euler, Algol W, Pascal, Modula, Modula-2, Oberon, and Oberon-2. In the last chapter of his 1976 book Algorithms + Data Structures = Programs, Wirth creates a top-down, descent parser with recovery for the Pascal-like, LL(1) programming language PL/0. The structure of the program is that of a PEG parser, though the concept of PEG wasn't formalized until 2004.
  • Bryan Ford introduced PEG (parsing expression grammars) in 2004.
  • Other parser generators like PEG.js by David Majda inspired the work in Grako.
  • William Thompson inspired the use of context managers with his blog post that I knew about through the invaluable Python Weekly newsletter, curated by Rahul Chaudhary
  • Jeff Knupp explains why Grako's use of exceptions is sound, so I don't have to.
  • Terence Parr created ANTLR, probably the most solid and professional parser generator out there. Ter, ANTLR, and the folks on the ANLTR forums helped me shape my ideas about Grako.
  • JavaCC (originally Jack) looks like an abandoned project. It was the first parser generator I used while teaching.
  • Grako is very fast. But dealing with millions of lines of legacy source code in a matter of minutes would be impossible without PyPy, the work of Armin Rigo and the PyPy team.
  • Guido van Rossum created and has lead the development of the Python programming environment for over a decade. A tool like Grako, at under six thousand lines of code, would not have been possible without Python.
  • Kota Mizushima welcomed me to the CSAIL at MIT [PEG and Packrat parsing mailing list], and immediately offered ideas and pointed me to documentation about the implementation of cut in modern parsers. The optimization of memoization information in Grako is thanks to one of his papers.
  • My students at UCAB inspired me to think about how grammar-based parser generation could be made more approachable.
  • Gustavo Lau was my professor of Language Theory at USB, and he was kind enough to be my tutor in a thesis project on programming languages that was more than I could chew. My peers, and then teaching advisers Alberto Torres, and Enzo Chiariotti formed a team with Gustavo to challenge us with programming languages like LATORTA and term exams that went well into the eight hours. And, of course, there was also the pirate patch that should be worn on the left or right eye depending on the LL or LR challenge.
  • Manuel Rey led me through another, unfinished, thesis project that taught me about what languages (spoken languages in general, and programming languages in particular) are about. I learned why languages use declensions, and why, although the underlying words are in English, the structure of the programs we write is more like Japanese.
  • Marcus Brinkmann has kindly submitted patches that have resolved obscure bugs in Grako's implementation, and that have made the tool more user-friendly, specially for newcomers to parsing and translation.
  • Robert Speer cleaned up the nonsense in trying to have Unicode handling be compatible with 2.7.x and 3.x, and figured out the canonical way of honoring escape sequences in grammar tokens without throwing off the encoding.
  • Basel Shishani has been an incredibly throrough peer-reviewer of Grako.
  • Paul Sargent implemented Warth et al's algorithm for supporting direct and indirect left recursion in PEG parsers.
  • Kathryn Long proposed better support for UNICODE in the treatment of whitespace and regular expressions (patterns) in general. Her other contributions have made Grako more congruent, and more user-friendly.
  • David Röthlisberger provided the definitive patch that allows the use of Python keywords as rule names.


The following, among others, have contributted to Grako with features, bug fixes, or suggestions:

basel-shishani , drothlis , franz_g , gapag , gegenschall , gkimbar , jimon , lambdafu , leewz , linkdd , nehz , neumond , pauls , pgebhard , r_speer , siemer , sjbrownBitbucket , starkat , tonico_strasser , vinay.sajip , vmuriart


See the CHANGELOG for details.