Skip to content

Home

Jump to bottom Edit New page
Bastiaan Veelo edited this page Apr 14, 2020 · 43 revisions

PEGGED

Pegged is a parsing expression grammar (PEG) generator implemented in the D programming language.

The idea is to give the generator a PEG, with the syntax presented in the reference article . From this grammar definition a set of related parsers will be created, to be used at runtime or compile time.

Usage

To use Pegged, just call the grammar function with a PEG and mix it in. For example:

import pegged.grammar;

mixin(grammar(`
Arithmetic:
    Term     < Factor (Add / Sub)*
    Add      < "+" Factor
    Sub      < "-" Factor
    Factor   < Primary (Mul / Div)*
    Mul      < "*" Primary
    Div      < "/" Primary
    Primary  < Parens / Neg / Number / Variable
    Parens   < "(" Term ")"
    Neg      < "-" Primary
    Number   < ~([0-9]+)
    Variable <- identifier
`));

This creates the Arithmetic grammar, with the Expr, Add, Factor (and so on) rules for basic arithmetic expressions with operator precedence ('*' and '/' bind stronger than '+' or '-'). identifier is a pre-defined parser recognizing your basic C-style identifier (first a letter or underscore, then digits, letters or underscores). In the rest of this document, I'll call 'rule' a Parser <- Parsing Expression expression and I'll use 'grammar' to designate the entire group of rules given to grammar.

To use a grammar, call it with a string. It will return a parse tree containing the calls to the different rules:

// Parsing at compile-time:
enum parseTree1 = Arithmetic("1 + 2 - (3*x-5)*6");

pragma(msg, parseTree1.matches);
assert(parseTree1.matches == ["1", "+", "2", "-", "(", "3", "*", "x", "-", "5", ")", "*", "6"]);
writeln(parseTree1);

// And at runtime too:
auto parseTree2 = Arithmetic(" 0 + 123 - 456 ");
assert(parseTree2.matches == ["0", "+", "123", "-", "456"]);

Even for such a simple grammar and such a simple expression, the resulting parse tree is a bit long to be shown here. See Parse Result.

By default, the grammars do not silently consume spaces, as this is the standard behaviour for PEGs. There is an opt-out though, with the simple < arrow instead of <- (you can see it in the previous example).

How to get Pegged

Pegged is a github project, hosted at https://github.com/PhilippeSigaud/Pegged

To get it:

$ git clone https://github.com/PhilippeSigaud/Pegged

The /docs directory contains an empty /wiki directory, linked to the github wiki as a git submodule. Here is how to get it:

$ cd <pegged directory>
$ git submodule init
$ git submodule update

This should give you a /docs/wiki directory full of markdown files, right from the online wiki.

Tutorial and docs

Here is a growing tutorial. All the wiki pages are also present (as Markdown files) in the docs directory.

As a Library

The only modules you need are pegged.peg, pegged.grammar and pegged.parser. To compile Pegged as a library, do:

>  dmd -lib -oflibpegged.a pegged/peg.d pegged/grammar.d pegged/parser.d

(FWIW, I use rdmd personally)

Features

  • The complete set of operators described here are implemented, with the 'traditional' PEG syntax. See Peg Basics.
  • Pegged can parse its input at compile time and generate a complete parse tree at compile time. In a word: compile-time string (read: D code) transformation and generation. See Generating Code for example.
  • You can parse at runtime also, you lucky you. (Using the Parse Tree)
  • Use a standard and readable PEG syntax as a DSL, not a bunch of templates that hide the parser in noise.
  • But you can use expression templates if you want, as parsers are all available as such. Pegged is implemented as an expression template, and what's good for the library writer is sure OK for the user too. (How Pegged Works)
  • Some useful additional operators are there too: a way to discard matches (thus dumping them from the parse tree), to push captures on a stack, to accept matches that are equal to another match: see Extended PEG Syntax.
  • Adding new parsers is easy. See User-Defined Parsers to see how to do that.
  • Grammars are composable: you can put different mixin(grammar(rules)); in a module and then grammars and rules can refer to one another. That way, you can have utility grammars providing their functionalities to other grammars. See Grammar Composition.
  • That's why Pegged comes with some pre-defined grammars (JSON, C, XML, CSV, D, the PEG grammar itself, etc). See Grammar Examples.
  • Grammars can be dumped in a file to create a module. Use the asModule(string moduleName, string gram) function in pegged.grammar to do that. See Grammars as D Modules.

More advanced features, outside the standard PEG perimeter are there to bring more power in the mix:

  • Full support for Left-Recursion in grammars.
  • Parameterized rules: "List(E, Sep) <- E (Sep E)*" is possible. The previous rule defines a parameterized parser taking two other parsers (namely, E and Sep) to match a Sep-separated list of E's. Entire grammars can be parameterized, too. See Parameterized Rules to see what's possible.
  • Semantic actions can be added to any rule in a grammar. Once a rule has matched, its associated action is called on the rule output and passed as final result to other parsers further up the grammar. Do what you want to the parse tree. If the passed actions are delegates, they can access external variables. See Semantic Actions.

Limitations

  • Pegged accepts only strings for the time being (wstrings or dstrings were OK as input some time ago, but I recently dropped them).
  • Error reporting is the same as for any in-my-own-free-time / just-one-guy parsing project (read: perfectible).
  • I recently dropped named captures and the = and @ operators.

Long-Term Goals (the Right to Dream)

  • As a long-term goal, parsing structures, as presented in OMeta. Yeah, I know, but I find that wonderfully interesting. Rules could match not only strings by any D type inner structure (matching a struct if it contains an int member named foo and a bar method, etc).
  • Hence, a pattern-matcher. if you used Haskell or ML, you know what I'm talking about.
  • As a longer-term goal: implementing the complete D grammar and see if that flies.
  • As an even longer-term goal: macros in D. Think Lisp and talking to God.

References

Articles:

D Code:

Other languages:

License

Pegged is released with the Boost license (like most D projects). See here for more details.

Add a custom footer
Add a custom sidebar
Clone this wiki locally