xpeg
xpeg copied to clipboard
Experimental PEG library for Elixir
More documentation will be added, for now please refer to the documentation of Xpeg, a Nim implementation of a similar PEG parser.
Introduction
Xpeg is a pure Elixir pattern matching library. It provides macros to compile patterns and grammars (PEGs) to Elixir function which will parse a string and collect selected parts of the input. PEGs are not unlike regular expressions, but offer more power and flexibility, and have less ambiguities. (More about PEGs on Wikipedia)
╭───────────»──────────╮
Object o──'{'─»─fn()─»─┬─[Obj_pair]─»─┴─┬─","─»─[Obj_pair]─┬─┴─┬─»─"}"──o
│ ╰─────────«────────╯ │
╰─[S]───────────────────────────────────╯
Some use cases where Xpeg is useful are configuration or data file parsers, robust protocol implementations, input validation, lexing of programming languages or domain specific languages.
Some Xpeg highlights:
-
Grammar definitions and Elixir code acting on or transforming the parsed fragments can be freely mixed.
-
Xpeg-generated parsers can be used both at run and at compile time.
-
Xpeg offers various methods for tracing, optimizing and debugging your parsers.
-
Xpeg can draw cool diagrams.
Installation
def deps do
[
{:xpeg, "~> 0.8.1"}
]
end
Quickstart
Here is a simple example showing the power of Xpeg: The macro peg
compiles a
grammar definition into a parser
function, which is used to match a string and
place the key-value pairs into a list of tuples:
p = Xpeg.peg Dict do
Dict <- Pair * star("," * Pair) * !1
Pair <- Word * "=" * Number * fn [a,b|cs] -> [{b,a}|cs] end
Word <- str(+{'a'..'z'})
Number <- int(+{'0'..'9'})
end
Xpeg.match(p, "grass=4,horse=1,star=2")
Output:
[{"star", 2}, {"horse", 1}, {"grass", 4}]
Usage
The basic operation consists of the provided grammar, which consists of a set
of named rules. A name is an elixir atom, in the form :name
or Name
,
whichever you prefer. A rule is made up of a number of atoms (not to be
confused with Elixirs atoms. I should probably find another name for this) and
operators, which are executed to match the input string. Rules can also call
into other rules, allowing for recursive grammars.
For example, the grammar below matches a comma-separated list of words
p = peg List do
List <- Word * star( "," * Word )
Word <- +{'a'..'z'}
end
- The
List
rule matches oneWord
, followed by zero or more (star(P)
) times a,
followed by aWord
- The
Word
rule matches one-or-more (+P
) times the set of characters ({}
) consisting of all letters from'a'
to'z'
During the execution of the grammar, matching parts of the subject strings can
be captured with the str()
operator. All captures are stored on the
captures
list inside the parser state. This list is returned by the match()
function, but can also be used by in-grammar functions to perform conversions
or transformations.
Below is the same grammar as above, but in this case it captures all
the individual Word
s:
p = peg List do
List <- Word * star( "," * Word )
Word <- str(+{'a'..'z'})
end
match(p, "one,two,three")
The above will return these following list of captures:
["three", "two", "one"]
A powerful feature allows mixing of Elixir functions with the grammar, which can be used to perform transformations of the captures or build abstract syntax trees (ASTs) on-the-fly.
For example, the grammar above is changed to match numbers instead of words,
and a conversion function is called after every matching number that
converts the last captured value on the captures
list to an integer:
p = peg List do
List <- Word * star( "," * Word )
Word <- str(+{'0'..'9'}) *
fn [v|cs] ->
[String.to_integer(v)|cs]
end
end
match(p, "123,42,31415")
which results in the following captures:
[31415, 42, 123]
More elaborate examples can be found in examples_test.exs, including a parser for arithmatic expressions and a full JSON parser.
Grammars
The peg
macro provides a method to define (recursive) grammars. The first
argument is the name of initial patterns, followed by a list of named patterns.
Patterns can now refer to other patterns by name, allowing for recursion.
The order in which the grammar patterns are defined affects the generated parser. Although Xpeg could always reorder, this is a design choice to give the user more control over the generated parser:
-
when a pattern P1 refers to pattern P2 which is defined before P1, P2 will be inlined in P1. This increases the generated code size, but generally improves performance.
-
when a pattern P1 refers to pattern P2 which is defined after P1, P2 will be generated as a subroutine which gets called from P1. This will reduce code size, but might also result in a slower parser.
Syntax
The Xpeg syntax is similar to normal PEG notation, but some changes were made to allow the grammar to be properly parsed by the Elixir compiler:
- Xpeg uses prefix operators instead of suffix operators for
+
,-
- Elixir does not support the
*
and?
prefix operators, so insteadstar(P)
andopt(P)
are used - The explicit
*
infix operator is used for concatenation
Xpeg patterns and grammars can be composed of the following parts:
Atoms:
0 # matches always and consumes nothing
1 # matches any character
n # matches exactly n characters
'x' # matches literal character 'x'
"xyz" # matches literal string "xyz"
{'x'..'y'} # matches any character in the range from 'x'..'y'
{'x','y','z'} # matches any character from the set
Operators:
P1 * P2 # concatenation
P1 | P2 # ordered choice
P1 - P2 # matches P1 if P2 does not match
(P) # grouping
!P # matches everything but P
&P # matches P without consuming input
opt(P) # matches P zero or one times
star(P) # matches P zero or more times
+P # matches P one or more times
P[n] # matches P n times
P[m..n] # matches P m to n times
@P # searches for P
Captures:
str(P) # Adds the matched string to the capture list
int(P) # Adds the matched integer to the capture list
float(P) # Adds the matched float to the capture list
Elixir function:
fn(captures) # Elixir function for transformations
Performance
Generated parsers will typically never reach the spead of a hand-crafted and fine tuned parser for a specific grammar. Having said that, Xpeg parsers can still be pretty fast; for example, the JSON parser from the examples runs at approximately 2/3 of the speed of the Poison JSON parser, which is said to be "wicked-fast"
Tracing and debugging
Syntax diagrams
When passing the option :dump_graph
to Xpeg.peg()
, Xpeg will dump syntax
diagrams (also known as railroad diagrams) for all parsed rules.
Syntax diagrams are sometimes helpful to understand or debug a grammar, or to get more insight in a grammars' complexity.
╭───────────»──────────╮
Object o──'{'─»─fn()─»─┬─[Obj_pair]─»─┴─┬─","─»─[Obj_pair]─┬─┴─┬─»─"}"──o
│ ╰─────────«────────╯ │
╰─[S]───────────────────────────────────╯
- Optionals (?) are indicated by a forward arrow overhead.
- Repeats ('+') are indicated by a backwards arrow underneath.
- Non-terminals are printed in square brackets.
Tracing
When passing the flag :dump_ir
to Xpeg.peg, it will print the IR representation of the
parsed grammar at compile time. The option :trace
will print the IR instructions and the matched subject
string during parsing - this will dramatically slow down the parsing, however.
For example, the following program:
Xpeg.peg Line, trace: true, dump_ir: true do
Space <- ' '
Line <- Word * star(Space * Word)
Word <- +{'a'..'z'}
end
will output the following intermediate representation at compile time. From the
IR it can be seen that the space rule has been inlined in the line rule, but
that the Word
rule has been emitted as a subroutine which gets called from
Line
:
Line:
0 :call 6
1 :choice 5 1
2 :chr 32
3 :call 6
4 :commit
5 :return
Word:
6 :set 'abcdefghijklmnopqrstuvwxyz'
7 :span 'abcdefghijklmnopqrstuvwxyz'
8 :return
fail :fail
At runtime, the following trace is generated. The trace consists of a number of columns:
- The current instruction pointer, which maps to the compile time dump.
- The substring of the subject.
- The instruction being executed.
0 | 'one two' | {:call, 6}
6 | 'one two' | {:set, 'abcdefghijklmnopqrstuvwxyz'}
7 | 'ne two' | {:span, 'abcdefghijklmnopqrstuvwxyz'}
7 | 'e two' | {:span, 'abcdefghijklmnopqrstuvwxyz'}
7 | ' two' | {:span, 'abcdefghijklmnopqrstuvwxyz'}
8 | ' two' | {:return}
1 | ' two' | {:choice, 5, 1}
2 | ' two' | {:chr, 32}
3 | 'two' | {:call, 6}
6 | 'two' | {:set, 'abcdefghijklmnopqrstuvwxyz'}
7 | 'wo' | {:span, 'abcdefghijklmnopqrstuvwxyz'}
7 | 'o' | {:span, 'abcdefghijklmnopqrstuvwxyz'}
7 | [] | {:span, 'abcdefghijklmnopqrstuvwxyz'}
8 | [] | {:return}
4 | [] | {:commit}
1 | [] | {:choice, 5, 1}
2 | [] | {:chr, 32}
fail | [] | {:fail}
5 | [] | {:return}
The exact meaning of the IR instructions is not discussed here
TODO
- I do not like the
star()
andopt()
syntax of the AST, but given the limited support for prefix operators in Elixir I'm not yet sure how to make this better