The fastest pure-Python PEG parser I can muster
Switch branches/tags
Nothing to show
Pull request Compare This branch is 176 commits behind erikrose:master.
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.



Parsimonious aims to be the fastest PEG parser written in pure Python. It was designed to undergird a MediaWiki parser that wouldn't take 5 seconds or a GB of RAM to do one page.

Beyond speed, secondary goals include...

  • Frugal RAM use
  • Minimalistic, understandable, idiomatic Python code
  • Readable grammars
  • Extensible grammars (TODO: Think about the ^ operator from OMeta. It would let us reference supergrammars' rules without repeating them.)
  • Complete test coverage
  • Separation of concerns. Some Python parsing kits mix recognition with instructions about how to turn the resulting tree into some kind of other representation. This feels limiting to me, since my use case--wikis-- naturally wants to do several different things with a tree: render to HTML, render to text, etc.
  • Good error reporting. I want the parser to work with me as I develop a grammar.

A Little About PEG Parsers

PEG parsers don't draw a distinction between lexing and parsing; everything's done at once. As a result, there is no lookahead limit, as there is with, for instance, Yacc. And, due to both of these properties, PEG grammars are easier to write: they're basically just EBNF. With caching, they take O(grammar size * text length) memory (though I plan to do better), but they run in O(text length) time.

More Technically

PEGs can describe a superset of LL(k) languages, any deterministic LR(k) language, and many others, including some that aren't context-free ( They can also deal with what would be ambiguous languages if described in canonical EBNF. They do this by trading the | alternation operator for the / operator, which works the same except that it makes priority explicit: a / b / c first tries matching a. If that fails, it tries b, and, failing that, moves on to c. Thus, ambiguity is resolved by always yielding the first successful derivation.

Writing Grammars

They're basically just extended BNF syntax:

"some literal"
Used to quote literals. Backslash escaping and Python conventions for "raw" and Unicode strings help support fiddly characters.
Sequences are made out of space- or tab-delimited things. a b c matches spots where those 3 terms appear in that order.
a / b
Alternatives. The first to succeed of a / b / c wins.
a & b
A lookahead assertion followed by a normal, consumed symbol. a & b & c means that a, b, and c all have to match at the current position. c, however, is the only bit that's actually consumed.
An optional expression
Matches if thing isn't found here. Doesn't consume any text.
Zero or more things
One or more things
Regexes have ~ in front and are quoted like literals. Any flags follow the end quotes as single chars. Regexes are good for representing character classes ([a-z0-9]) and optimizing for speed.


bold_text  = bold_open text bold_close
text       = ~"[A-Z 0-9]*"i
bold_open  = '(('
bold_close = '))'

We might implement parentheses in the future for anonymous grouping. For now, just break up complex rules instead.

We shouldn't need to represent Empty; the quantifiers should suffice.

Optimizing Grammars

Don't repeat expressions. If you need a Regex('such-and-such') at some point in your grammar, don't type it twice; make it a rule of its own, and reference it from wherever you need it. You'll get the most out of the caching this way, since cache lookups are by expression object identity (for speed). Even if you have an expression that's very simple, not repeating it will save RAM, as there can, at worst, be a cached int for every char in the text you're parsing. But hmm, maybe I can identify repeat subexpressions automatically and factor that up while building the grammar....

How much should you shove into one Regex, versus how much should you break them up to not repeat yourself? That's a fine balance and worthy of benchmarking. More stuff jammed into a regex will execute faster, because it doesn't have to run any Python between pieces, but a broken-up one will give better cache performance if the individual pieces are re-used elsewhere. If the pieces of a regex aren't used anywhere else, by all means keep the whole thing together.

Quantifiers: bring your ? and * quantifiers up to the highest level you can. Otherwise, lower-level patterns could succeed but be empty and put a bunch of useless nodes in your tree that didn't really match anything.

Dealing With Parse Trees

A parse tree has a node for each expression matched, even if it matched a zero-length string, like "thing"? might do.

TODO: Talk about tree manipulators (if we write any) and about NodeVisitor.

When something goes wrong in your visitor, you get a nice error like this:

[normal traceback here...]
VisitationException: 'Node' object has no attribute 'name'

Parse tree:
<rules "number = ~"[0-9]+"">  <-- *** We were here. ***
    <Node "number = ~"[0-9]+"">
        <rule "number = ~"[0-9]+"">
            <Node "">
            <label "number">
            <Node " ">
                <_ " ">
            <Node "=">
            <Node " ">
                <_ " ">
            <rhs "~"[0-9]+"">
                <term "~"[0-9]+"">
                    <atom "~"[0-9]+"">
                        <regex "~"[0-9]+"">
                            <Node "~">
                            <literal ""[0-9]+"">
                            <Node "">
            <Node "">
            <eol "
    <Node "">

Note the parse tree tacked onto the exception. The node whose visitor method raised the error is pointed out.

Future Directions

Language Changes

  • Do we need a LookAhead? It might be faster, but A Lookahead(B) is equivalent to AB & A.

  • Maybe support left-recursive rules like PyMeta, if anybody cares.

  • The ability to mark certain nodes as undesired, so we don't bother constructing them and cluttering the tree with them. For example, we might only care to see the OneOf node in the final tree, not the boring Literals inside it:

    greeting = "hi" / "hello" / "bonjour"

    Perhaps we could express it like this:

    greeting = -"hi" / -"hello" / -"bonjour"

    On the other hand, parentheses for anonymous subexpressions could largely solve this problem--and in a more familiar way--if we implicitly omitted their nodes. (The exception would be subexpressions that you end up having to repeat several times in the grammar.)

  • Pijnu has a raft of tree manipulators. I don't think I want all of them, but a judicious subset might be nice. Don't get into mixing formatting with tree manipulation.

  • Think about having the ability, like PyParsing, to get irrevocably into a pattern so that we don't backtrack out of it. Then, if things don't end up matching, we complain with an informative error message rather than backtracking to nonsense.


  • Make RAM use almost constant by automatically inserting "cuts", as described in
  • Think about having the user (optionally) provide some representative input along with a grammar. We can then profile against it, see which expressions are worth caching, and annotate the grammar. Perhaps there will even be positions at which a given expression is more worth caching. Or we could keep a count of how many times each cache entry has been used and evict the most useless ones as RAM use grows.
  • We could possibly compile the grammar into VM instructions, like in "A parsing machine for PEGs" by Medeiros.
  • If the recursion gets too deep in practice, use trampolining to dodge it.