Portability | unknown |
---|---|
Stability | experimental |
Maintainer | bos@serpentine.com |
Safe Haskell | Safe-Infered |
Simple, efficient combinator parsing for Text
strings,
loosely based on the Parsec library.
- type Parser = Parser Text
- type Result = IResult Text
- data IResult t r
- parse :: Parser a -> Text -> Result a
- feed :: Result r -> Text -> Result r
- parseOnly :: Parser a -> Text -> Either String a
- parseWith :: Monad m => m Text -> Parser a -> Text -> m (Result a)
- parseTest :: Show a => Parser a -> Text -> IO ()
- maybeResult :: Result r -> Maybe r
- eitherResult :: Result r -> Either String r
- (<?>) :: Parser a -> String -> Parser a
- try :: Parser a -> Parser a
- module Data.Attoparsec.Combinator
- char :: Char -> Parser Char
- anyChar :: Parser Char
- notChar :: Char -> Parser Char
- satisfy :: (Char -> Bool) -> Parser Char
- satisfyWith :: (Char -> a) -> (a -> Bool) -> Parser a
- skip :: (Char -> Bool) -> Parser ()
- peekChar :: Parser (Maybe Char)
- digit :: Parser Char
- letter :: Parser Char
- space :: Parser Char
- inClass :: String -> Char -> Bool
- notInClass :: String -> Char -> Bool
- string :: Text -> Parser Text
- stringCI :: Text -> Parser Text
- skipSpace :: Parser ()
- skipWhile :: (Char -> Bool) -> Parser ()
- scan :: s -> (s -> Char -> Maybe s) -> Parser Text
- take :: Int -> Parser Text
- takeWhile :: (Char -> Bool) -> Parser Text
- takeWhile1 :: (Char -> Bool) -> Parser Text
- takeTill :: (Char -> Bool) -> Parser Text
- (.*>) :: Text -> Parser a -> Parser a
- (<*.) :: Parser a -> Text -> Parser a
- takeText :: Parser Text
- takeLazyText :: Parser Text
- endOfLine :: Parser ()
- isEndOfLine :: Char -> Bool
- isHorizontalSpace :: Char -> Bool
- decimal :: Integral a => Parser a
- hexadecimal :: (Integral a, Bits a) => Parser a
- signed :: Num a => Parser a -> Parser a
- double :: Parser Double
- data Number
- number :: Parser Number
- rational :: Fractional a => Parser a
- endOfInput :: Parser ()
- atEnd :: Parser Bool
Differences from Parsec
Compared to Parsec 3, Attoparsec makes several tradeoffs. It is not intended for, or ideal for, all possible uses.
- While Attoparsec can consume input incrementally, Parsec cannot. Incremental input is a huge deal for efficient and secure network and system programming, since it gives much more control to users of the library over matters such as resource usage and the I/O model to use.
- Much of the performance advantage of Attoparsec is gained via
high-performance parsers such as
takeWhile
andstring
. If you use complicated combinators that return lists of characters, there is less performance difference between the two libraries. - Unlike Parsec 3, Attoparsec does not support being used as a monad transformer.
- Attoparsec is specialised to deal only with strict
Text
input. Efficiency concerns rule out both lists and lazy text. The usual use for lazy text would be to allow consumption of very large input without a large footprint. For this need, Attoparsec's incremental input provides an excellent substitute, with much more control over when input takes place. If you must use lazy text, see theLazy
module, which feeds lazy chunks to a regular parser. - Parsec parsers can produce more helpful error messages than Attoparsec parsers. This is a matter of focus: Attoparsec avoids the extra book-keeping in favour of higher performance.
Incremental input
Attoparsec supports incremental input, meaning that you can feed it
a Text
that represents only part of the expected total amount
of data to parse. If your parser reaches the end of a fragment of
input and could consume more input, it will suspend parsing and
return a Partial
continuation.
Supplying the Partial
continuation with another string will
resume parsing at the point where it was suspended. You must be
prepared for the result of the resumed parse to be another
Partial
continuation.
To indicate that you have no more input, supply the Partial
continuation with an empty
Text
.
Remember that some parsing combinators will not return a result
until they reach the end of input. They may thus cause Partial
results to be returned.
If you do not need support for incremental input, consider using
the parseOnly
function to run your parser. It will never
prompt for more input.
Performance considerations
If you write an Attoparsec-based parser carefully, it can be realistic to expect it to perform within a factor of 2 of a hand-rolled C parser (measuring megabytes parsed per second).
To actually achieve high performance, there are a few guidelines that it is useful to follow.
Use the Text
-oriented parsers whenever possible,
e.g. takeWhile1
instead of many1
anyChar
. There is
about a factor of 100 difference in performance between the two
kinds of parser.
For very simple character-testing predicates, write them by hand
instead of using inClass
or notInClass
. For instance, both
of these predicates test for an end-of-line character, but the
first is much faster than the second:
endOfLine_fast c = c == '\r' || c == '\n' endOfLine_slow = inClass "\r\n"
Make active use of benchmarking and profiling tools to measure, find the problems with, and improve the performance of your parser.
Parser types
The result of a parse. This is parameterised over the type t
of string that was processed.
This type is an instance of Functor
, where fmap
transforms the
value in a Done
result.
Fail t [String] String | The parse failed. The |
Partial (t -> IResult t r) | Supply this continuation with more input so that the parser can resume. To indicate that no more input is available, use an empty string. |
Done t r | The parse succeeded. The |
Running parsers
feed :: Result r -> Text -> Result rSource
If a parser has returned a Partial
result, supply it with more
input.
parseOnly :: Parser a -> Text -> Either String aSource
Run a parser that cannot be resupplied via a Partial
result.
:: Monad m | |
=> m Text | An action that will be executed to provide the parser
with more input, if necessary. The action must return an
|
-> Parser a | |
-> Text | Initial input for the parser. |
-> m (Result a) |
Run a parser with an initial input string, and a monadic action that can supply more input if needed.
parseTest :: Show a => Parser a -> Text -> IO ()Source
Run a parser and print its result to standard output.
Result conversion
maybeResult :: Result r -> Maybe rSource
eitherResult :: Result r -> Either String rSource
Combinators
Name the parser, in case failure occurs.
try :: Parser a -> Parser aSource
Attempt a parse, and if it fails, rewind the input so that no input appears to have been consumed.
This combinator is provided for compatibility with Parsec. Attoparsec parsers always backtrack on failure.
module Data.Attoparsec.Combinator
Parsing individual characters
satisfy :: (Char -> Bool) -> Parser CharSource
The parser satisfy p
succeeds for any character for which the
predicate p
returns True
. Returns the character that is
actually parsed.
digit = satisfy isDigit where isDigit c = c >= '0' && c <= '9'
satisfyWith :: (Char -> a) -> (a -> Bool) -> Parser aSource
The parser satisfyWith f p
transforms a character, and succeeds
if the predicate p
returns True
on the transformed value. The
parser returns the transformed character that was parsed.
skip :: (Char -> Bool) -> Parser ()Source
The parser skip p
succeeds for any character for which the
predicate p
returns True
.
skipDigit = skip isDigit where isDigit c = c >= '0' && c <= '9'
peekChar :: Parser (Maybe Char)Source
Match any character. Returns Nothing
if end of input has been
reached. Does not consume any input.
Note: Because this parser does not fail, do not use it with
combinators such as many
, because such parsers loop until a
failure occurs. Careless use will thus result in an infinite loop.
Special character parsers
Character classes
inClass :: String -> Char -> BoolSource
Match any character in a set.
vowel = inClass "aeiou"
Range notation is supported.
halfAlphabet = inClass "a-nA-N"
To add a literal '-'
to a set, place it at the beginning or end
of the string.
notInClass :: String -> Char -> BoolSource
Match any character not in a set.
Efficient string handling
string :: Text -> Parser TextSource
string s
parses a sequence of characters that identically match
s
. Returns the parsed string (i.e. s
). This parser consumes no
input if it fails (even if a partial match).
Note: The behaviour of this parser is different to that of the
similarly-named parser in Parsec, as this one is all-or-nothing.
To illustrate the difference, the following parser will fail under
Parsec given an input of for
:
string "foo" <|> string "for"
The reason for its failure is that that the first branch is a
partial match, and will consume the letters 'f'
and 'o'
before failing. In Attoparsec, the above parser will succeed on
that input, because the failed first branch will consume nothing.
stringCI :: Text -> Parser TextSource
Satisfy a literal string, ignoring case.
Note: this function is currently quite inefficient. Unicode case folding can change the length of a string ("ß" becomes ss), which makes a simple, efficient implementation tricky. We have (for now) chosen simplicity over efficiency.
skipWhile :: (Char -> Bool) -> Parser ()Source
Skip past input for as long as the predicate returns True
.
scan :: s -> (s -> Char -> Maybe s) -> Parser TextSource
A stateful scanner. The predicate consumes and transforms a
state argument, and each transformed state is passed to successive
invocations of the predicate on each character of the input until one
returns Nothing
or the input ends.
This parser does not fail. It will return an empty string if the
predicate returns Nothing
on the first character of input.
Note: Because this parser does not fail, do not use it with
combinators such as many
, because such parsers loop until a
failure occurs. Careless use will thus result in an infinite loop.
takeWhile :: (Char -> Bool) -> Parser TextSource
Consume input as long as the predicate returns True
, and return
the consumed input.
This parser does not fail. It will return an empty string if the
predicate returns False
on the first character of input.
Note: Because this parser does not fail, do not use it with
combinators such as many
, because such parsers loop until a
failure occurs. Careless use will thus result in an infinite loop.
takeTill :: (Char -> Bool) -> Parser TextSource
Consume input as long as the predicate returns False
(i.e. until it returns True
), and return the consumed input.
This parser does not fail. It will return an empty string if the
predicate returns True
on the first character of input.
Note: Because this parser does not fail, do not use it with
combinators such as many
, because such parsers loop until a
failure occurs. Careless use will thus result in an infinite loop.
String combinators
The .*>
and <*.
combinators are intended for use with the
OverloadedStrings
language extension. They simplify the common
task of matching a statically known string, then immediately
parsing something else.
An example makes this easier to understand:
{-# LANGUAGE OverloadedStrings #-} shoeSize = "Shoe size: ".*>
decimal
If we were to try to use *>
above instead, the type checker would
not be able to tell which IsString
instance to use for the text
in quotes. We would have to be explicit, using either a type
signature or the string
parser.
Consume all remaining input
takeLazyText :: Parser TextSource
Consume all remaining input and return it as a single string.
Text parsing
Match either a single newline character '\n'
, or a carriage
return followed by a newline character "\r\n"
.
isEndOfLine :: Char -> BoolSource
A predicate that matches either a carriage return '\r'
or
newline '\n'
character.
isHorizontalSpace :: Char -> BoolSource
A predicate that matches either a space ' '
or horizontal tab
'\t'
character.
Numeric parsers
hexadecimal :: (Integral a, Bits a) => Parser aSource
Parse and decode an unsigned hexadecimal number. The hex digits
'a'
through 'f'
may be upper or lower case.
This parser does not accept a leading "0x"
string.
signed :: Num a => Parser a -> Parser aSource
Parse a number with an optional leading '+'
or '-'
sign
character.
Parse a rational number.
The syntax accepted by this parser is the same as for rational
.
Note: This function is almost ten times faster than rational
,
but is slightly less accurate.
The Double
type supports about 16 decimal places of accuracy.
For 94.2% of numbers, this function and rational
give identical
results, but for the remaining 5.8%, this function loses precision
around the 15th decimal place. For 0.001% of numbers, this
function will lose precision at the 13th or 14th decimal place.
This function does not accept string representations of "NaN" or "Infinity".
A numeric type that can represent integers accurately, and
floating point numbers to the precision of a Double
.
Parse a number, attempting to preserve both speed and precision.
The syntax accepted by this parser is the same as for rational
.
Note: This function is almost ten times faster than rational
.
On integral inputs, it gives perfectly accurate answers, and on
floating point inputs, it is slightly less accurate than
rational
.
This function does not accept string representations of "NaN" or "Infinity".
rational :: Fractional a => Parser aSource
Parse a rational number.
This parser accepts an optional leading sign character, followed by
at least one decimal digit. The syntax similar to that accepted by
the read
function, with the exception that a trailing '.'
or
'e'
not followed by a number is not consumed.
Examples with behaviour identical to read
, if you feed an empty
continuation to the first result:
rational "3" == Done 3.0 "" rational "3.1" == Done 3.1 "" rational "3e4" == Done 30000.0 "" rational "3.1e4" == Done 31000.0, ""
Examples with behaviour identical to read
:
rational ".3" == Fail "input does not start with a digit" rational "e3" == Fail "input does not start with a digit"
Examples of differences from read
:
rational "3.foo" == Done 3.0 ".foo" rational "3e" == Done 3.0 "e"
This function does not accept string representations of "NaN" or "Infinity".
State observation and manipulation functions
Match only if all input has been consumed.