A parser is a fold that can fail and is represented as Parser step initial extract. Before we drive a parser we call the initial action to retrieve the initial state of the fold. The parser driver invokes step with the state returned by the previous step and the next input element. It results into a new state and a command to the driver represented by Step type. The driver keeps invoking the step function until it stops or fails. At any point of time the driver can call extract to inspect the result of the fold. If the parser hits the end of input extract is called. It may result in an error or an output value.

Pre-release

This exception is used when a parser ultimately fails, the user of the parser is intimated via this exception.

Like ParseError but reports the stream position where the error ocurred. The Int is the position in the stream where the error ocurred. This exception is used by position reporting parser drivers.

Make a Parser from a Fold. This parser sends all of its input to the fold.

A parser that always yields a pure value without consuming any input.

A parser that always yields the result of an effectful action without consuming any input.

A parser that always fails with an error message without consuming any input.

Peek the head element of a stream, without consuming it. Fails if it encounters end of input.

>>> Stream.parse ((,) <$> Parser.peek <*> Parser.satisfy (> 0)) $ Stream.fromList [1]
Right (1,1)

peek = lookAhead (satisfy True)

Succeeds if we are at the end of input, fails otherwise.

>>> Stream.parse ((,) <$> Parser.satisfy (> 0) <*> Parser.eof) $ Stream.fromList [1]
Right (1,())

Consume one element from the head of the stream. Fails if it encounters end of input.

>>> one = Parser.satisfy $ const True

Match any one of the elements in the supplied list.

>>> oneOf xs = Parser.satisfy (`Foldable.elem` xs)

When performance matters a pattern matching predicate could be more efficient than a Foldable datatype:

let p x =
   case x of
      a -> True
      e -> True
       _  -> False
in satisfy p

GHC may use a binary search instead of linear search in the list. Alternatively, you can also use an array instead of list for storage and search.

See performance notes in oneOf.

>>> noneOf xs = Parser.satisfy (`Foldable.notElem` xs)

Returns the next element if it passes the predicate, fails otherwise.

>>> Stream.parse (Parser.satisfy (== 1)) $ Stream.fromList [1,0,1]
Right 1

>>> toMaybe f x = if f x then Just x else Nothing
>>> satisfy f = Parser.maybe (toMaybe f)

Like listEqBy but uses a stream instead of a list and does not return the stream.

Match the given sequence of elements using the given comparison function. Returns the original sequence if successful.

Definition:

>>> listEqBy cmp xs = Parser.streamEqBy cmp (Stream.fromList xs) *> Parser.fromPure xs

Examples:

>>> Stream.parse (Parser.listEqBy (==) "string") $ Stream.fromList "string"
Right "string"

>>> Stream.parsePos (Parser.listEqBy (==) "mismatch") $ Stream.fromList "match"
Left (ParseErrorPos 2 "streamEqBy: mismtach occurred")

Match the input sequence with the supplied list and return it if successful.

>>> listEq = Parser.listEqBy (==)

lmap f parser maps the function f on the input of the parser.

>>> Stream.parse (Parser.lmap (\x -> x * x) (Parser.fromFold Fold.sum)) (Stream.enumerateFromTo 1 100)
Right 338350

lmap = Parser.lmapM return

lmapM f parser maps the monadic function f on the input of the parser.

rmapM f parser maps the monadic function f on the output of the parser.

>>> rmap = fmap

Include only those elements that pass a predicate.

>>> Stream.parse (Parser.filter (> 5) (Parser.fromFold Fold.sum)) $ Stream.fromList [1..10]
Right 40

Run a parser without consuming the input.

Stops after taking exactly n input elements.

• Stops - after consuming n elements.
• Fails - if the stream or the collecting fold ends before it can collect exactly n elements.

>>> Stream.parsePos (Parser.takeEQ 2 Fold.toList) $ Stream.fromList [1,0,1]
Right [1,0]

>>> Stream.parsePos (Parser.takeEQ 4 Fold.toList) $ Stream.fromList [1,0,1]
Left (ParseErrorPos 3 "takeEQ: Expecting exactly 4 elements, input terminated on 3")

Collect stream elements until an element fails the predicate. The element on which the predicate fails is returned back to the input stream.

• Stops - when the predicate fails or the collecting fold stops.
• Fails - never.

>>> Stream.parse (Parser.takeWhile (== 0) Fold.toList) $ Stream.fromList [0,0,1,0,1]
Right [0,0]

>>> takeWhile cond f = Parser.takeWhileP cond (Parser.fromFold f)

We can implement a breakOn using takeWhile:

breakOn p = takeWhile (not p)

Like takeWhile but takes at least one element otherwise fails.

>>> takeWhile1 cond p = Parser.takeWhileP cond (Parser.takeBetween 1 maxBound p)

Drain the input as long as the predicate succeeds, running the effects and discarding the results.

This is also called skipWhile in some parsing libraries.

>>> dropWhile p = Parser.takeWhile p Fold.drain

Like splitOn but strips leading, trailing, and repeated separators. Therefore, ".a..b." having . as the separator would be parsed as ["a","b"]. In other words, its like parsing words from whitespace separated text.

• Stops - when it finds a word separator after a non-word element
• Fails - never.

>>> wordBy = Parser.wordFramedBy (const False) (const False) (const False)

S.wordsBy pred f = S.parseMany (PR.wordBy pred f)

Given an input stream [a,b,c,...] and a comparison function cmp, the parser assigns the element a to the first group, then if a `cmp` b is True b is also assigned to the same group. If a `cmp` c is True then c is also assigned to the same group and so on. When the comparison fails the parser is terminated. Each group is folded using the Fold f and the result of the fold is the result of the parser.

• Stops - when the comparison fails.
• Fails - never.

>>> :{
 runGroupsBy eq =
     Stream.fold Fold.toList
         . Stream.parseMany (Parser.groupBy eq Fold.toList)
         . Stream.fromList
:}

>>> runGroupsBy (<) []
[]

>>> runGroupsBy (<) [1]
[Right [1]]

>>> runGroupsBy (<) [3, 5, 4, 1, 2, 0]
[Right [3,5,4],Right [1,2],Right [0]]

Unlike groupBy this combinator performs a rolling comparison of two successive elements in the input stream. Assuming the input stream is [a,b,c,...] and the comparison function is cmp, the parser first assigns the element a to the first group, then if a `cmp` b is True b is also assigned to the same group. If b `cmp` c is True then c is also assigned to the same group and so on. When the comparison fails the parser is terminated. Each group is folded using the Fold f and the result of the fold is the result of the parser.

• Stops - when the comparison fails.
• Fails - never.

>>> :{
 runGroupsByRolling eq =
     Stream.fold Fold.toList
         . Stream.parseMany (Parser.groupByRolling eq Fold.toList)
         . Stream.fromList
:}

>>> runGroupsByRolling (<) []
[]

>>> runGroupsByRolling (<) [1]
[Right [1]]

>>> runGroupsByRolling (<) [3, 5, 4, 1, 2, 0]
[Right [3,5],Right [4],Right [1,2],Right [0]]

Pre-release

Like groupByRolling, but if the predicate is True then collects using the first fold as long as the predicate holds True, if the predicate is False collects using the second fold as long as it remains False. Returns Left for the first case and Right for the second case.

For example, if we want to detect sorted sequences in a stream, both ascending and descending cases we can use 'groupByRollingEither (<=) Fold.toList Fold.toList'.

Pre-release

Quote and bracket aware word splitting with escaping. Like wordBy but word separators within specified quotes or brackets are ignored. Quotes and escape characters can be processed. If the end quote is different from the start quote it is called a bracket. The following quoting rules apply:

• In an unquoted string a character may be preceded by an escape character. The escape character is removed and the character following it is treated literally with no special meaning e.g. e.g. h e l l o is a single word, n is same as n.
• Any part of the word can be placed within quotes. Inside quotes all characters are treated literally with no special meaning. Quoting character itself cannot be used within quotes unless escape processing within quotes is applied to allow it.
• Optionally escape processing for quoted part can be specified. Escape character has no special meaning inside quotes unless it is followed by a character that has a escape translation specified, in that case the escape character is removed, and the specified translation is applied to the character following it. This can be used to escape the quoting character itself within quotes.
• There can be multiple quoting characters, when a quote starts, all other quoting characters within that quote lose any special meaning until the quote is closed.
• A starting quote char without an ending char generates a parse error. An ending bracket char without a corresponding bracket begin is ignored.
• Brackets can be nested.

We should note that unquoted and quoted escape processing are different. In unquoted part escape character is always removed. In quoted part it is removed only if followed by a special meaning character. This is consistent with how shell performs escape processing.

Collect zero or more parses. Apply the supplied parser repeatedly on the input stream and push the parse results to a downstream fold.

Stops: when the downstream fold stops or the parser fails. Fails: never, produces zero or more results.

>>> many = Parser.countBetween 0 maxBound

Compare with many.

Collect one or more parses. Apply the supplied parser repeatedly on the input stream and push the parse results to a downstream fold.

Stops: when the downstream fold stops or the parser fails. Fails: if it stops without producing a single result.

>>> some p f = Parser.manyP p (Parser.takeGE 1 f)
>>> some = Parser.countBetween 1 maxBound

Compare with some.

manyTill p test f tries the parser test on the input, if test fails it backtracks and tries p, after p succeeds test is tried again and so on. The parser stops when test succeeds. The output of test is discarded and the output of p is accumulated by the supplied fold. The parser fails if p fails.

Stops when the fold f stops.

Apply two parsers alternately to an input stream. The input stream is considered an interleaving of two patterns. The two parsers represent the two patterns. Parsing starts at the first parser and stops at the first parser. It can be used to parse a infix style pattern e.g. p1 p2 p1 . Empty input or single parse of the first parser is accepted.

>>> p1 = Parser.takeWhile1 (not . (== '+')) Fold.toList
>>> p2 = Parser.satisfy (== '+')
>>> p = Parser.deintercalate p1 p2 Fold.toList
>>> Stream.parse p $ Stream.fromList ""
Right []
>>> Stream.parse p $ Stream.fromList "1"
Right [Left "1"]
>>> Stream.parse p $ Stream.fromList "1+"
Right [Left "1"]
>>> Stream.parse p $ Stream.fromList "1+2+3"
Right [Left "1",Right '+',Left "2",Right '+',Left "3"]

See also chainl1.