Documentation

Parse any Int64 (big-endian).

Parse any Int64 (little-endian).

Parse any Int32 (big-endian).

Parse any Int32 (little-endian).

Parse any Int16 (big-endian).

Parse any Int16 (little-endian).

Parse any Word64 (big-endian).

Parse any Word64 (little-endian).

Parse any Word32 (big-endian).

Parse any Word32 (little-endian).

Parse any Word16 (big-endian).

Parse any Word16 (little-endian).

Parse any Int.

Parse any Int64.

Parse any Int32.

Parse any Int16.

Parse any Int8.

Skip any Word.

Parse any Word.

Skip any Word64.

Parse any Word64.

Skip any Word32.

Parse any Word32.

Skip any Word16.

Parse any Word16.

Skip any Word8 (byte).

Parse any Word8 (byte).

Template function, creates a Parser e () which unsafely scans a given sequence of bytes.

Decrease the current input position by the given number of bytes.

Unsafely read and return a byte from the input. It's not checked that the input is non-empty.

Unsafely read eight concrete bytes from the input. It's not checked that the input has enough bytes.

Unsafely read four concrete bytes from the input. It's not checked that the input has enough bytes.

Unsafely read two concrete bytes from the input. It's not checked that the input has enough bytes.

Unsafely read a concrete byte from the input. It's not checked that the input has enough bytes.

Check that the input has at least the given number of bytes.

Convert an UTF-8-coded ByteString to a String.

Get the rest of the input as a String, but restore the parsing state. Assumes UTF-8 encoding. This can be used for debugging.

Take the rest of the input as a String. Assumes UTF-8 encoding.

Parse the rest of the current line as a String, but restore the parsing state. Assumes UTF-8 encoding. This can be used for debugging.

Parse the rest of the current line as a String. Assumes UTF-8 encoding, throws an error if the encoding is invalid.

Create a Pos from a line and column number. Throws an error on out-of-bounds line and column numbers.

Create a ByteString from a Span. The result is invalid if the Span points outside the current buffer, or if the Span start is greater than the end position.

Compute corresponding line and column numbers for each Pos in a list. Throw an error on invalid positions. Note: computing lines and columns may traverse the ByteString, but it traverses it only once regardless of the length of the position list.

Check whether a Pos points into a ByteString.

Run a parser in a given input span. The input position and the Int state is restored after the parser is finished, so inSpan does not consume input and has no side effect. Warning: this operation may crash if the given span points outside the current parsing buffer. It's always safe to use inSpan if the span comes from a previous spanned or spanOf call on the current input.

CPS'd version of byteStringOf. Can be more efficient, because the result is more eagerly unboxed by GHC. It's more efficient to use spanOf or spanned instead.

Return the ByteString consumed by a parser. Note: it's more efficient to use spanOf and spanned instead.

Bind the result together with the span of the result. CPS'd version of spanOf for better unboxing.

Return the consumed span of a parser.

The end of the input.

Set the input position. Warning: this can result in crashes if the position points outside the current buffer. It is always safe to setPos values which came from getPos with the current input.

Get the current position in the input.

Succeed if the first parser succeeds and the second one fails.

Skip a parser one or more times.

Run a parser one or more times, collect the results in a list. Note: for optimal performance, try to avoid this. Often it is possible to get rid of the intermediate list by using a combinator or a custom parser.

Skip a parser zero or more times.

Run a parser zero or more times, collect the results in a list. Note: for optimal performance, try to avoid this. Often it is possible to get rid of the intermediate list by using a combinator or a custom parser.

An analogue of the list foldr function: parse zero or more a-s, terminated by a b, and combine the results in a right-nested way using the a -> b -> b function. Note: this is not the usual chainr function from the parsec libraries!

An analogue of the list foldl function: first parse a b, then parse zero or more a-s, and combine the results in a left-nested way by the b -> a -> b function. Note: this is not the usual chainl function from the parsec libraries!

Branch on a parser: if the first argument succeeds, continue with the second, else with the third. This can produce slightly more efficient code than (<|>). Moreover, ḃranch does not backtrack from the true/false cases.

Choose between two parsers. If the first parser fails, try the second one, but if the first one throws an error, propagate the error.

Read an Integer from the input, as a non-empty digit sequence.

Read an Int from the input, as a non-empty digit sequence. The Int may overflow in the result.

Skip any Char in the ASCII range. More efficient than anyChar_ if we're working only with ASCII.

Parse any Char in the ASCII range, fail if the next input character is not in the range. This is more efficient than anyChar if we are only working with ASCII.

Skip any UTF-8-encoded Char.

Parse any UTF-8-encoded Char.

Skipping variant of fusedSatisfy.

This is a variant of satisfy which allows more optimization. We can pick four testing functions for the four cases for the possible number of bytes in the UTF-8 character. So in fusedSatisfy f1 f2 f3 f4, if we read a one-byte character, the result is scrutinized with f1, for two-bytes, with f2, and so on. This can result in dramatic lexing speedups.

For example, if we want to accept any letter, the naive solution would be to use isLetter, but this accesses a large lookup table of Unicode character classes. We can do better with fusedSatisfy isLatinLetter isLetter isLetter isLetter, since here the isLatinLetter is inlined into the UTF-8 decoding, and it probably handles a great majority of all cases without accessing the character table.

Skip an ASCII Char for which a predicate holds. Assumption: the predicate must only return True for ASCII-range characters.

Parse an ASCII Char for which a predicate holds. Assumption: the predicate must only return True for ASCII-range characters. Otherwise this function might read a 128-255 range byte, thereby breaking UTF-8 decoding.

Skip a UTF-8 Char for which a predicate holds.

Parse a UTF-8 Char for which a predicate holds.

Version of switchWithPost without syntactic sugar. The second argument is the list of cases, the third is the default case.

Switch expression with an optional first argument for performing a post-processing action after every successful branch matching. For example, if we have ws :: Parser e () for a whitespace parser, we might want to consume whitespace after matching on any of the switch cases. For that case, we can define a "lexeme" version of switch as follows.

  switch' :: Q Exp -> Q Exp
  switch' = switchWithPost (Just [| ws |])

Note that this switch' function cannot be used in the same module it's defined in, because of the stage restriction of Template Haskell.

This is a template function which makes it possible to branch on a collection of string literals in an efficient way. By using switch, such branching is compiled to a trie of primitive parsing operations, which has optimized control flow, vectorized reads and grouped checking for needed input bytes.

The syntax is slightly magical, it overloads the usual case expression. An example:

    $(switch [| case _ of
        "foo" -> pure True
        "bar" -> pure False |])

The underscore is mandatory in case _ of. Each branch must be a string literal, but optionally we may have a default case, like in

    $(switch [| case _ of
        "foo" -> pure 10
        "bar" -> pure 20
        _     -> pure 30 |])

All case right hand sides must be parsers with the same type. That type is also the type of the whole switch expression.

A switch has longest match semantics, and the order of cases does not matter, except for the default case, which may only appear as the last case.

If a switch does not have a default case, and no case matches the input, then it returns with failure, without having consumed any input. A fallthrough to the default case also does not consume any input.

Parse a UTF-8 string literal. This is a template function, you can use it as $(string "foo"), for example, and the splice has type Parser e ().

Read a sequence of bytes. This is a template function, you can use it as $(bytes [3, 4, 5]), for example, and the splice has type Parser e ().

Read a Word8.

Parse a UTF-8 character literal. This is a template function, you can use it as $(char 'x'), for example, and the splice in this case has type Parser e ().

Succeed if the input is empty.

Run the parser, if we get a failure, throw the given error, but if we get an error, merge the inner and the newly given errors using the e -> e -> e function. This can be useful for implementing parsing errors which may propagate hints or accummulate contextual information.

CPS'd version of optional. This is usually more efficient, since it gets rid of the extra Maybe allocation.

Convert a parsing failure to a ().

Convert a parsing failure to a Maybe. If possible, use optioned instead.

Convert a parsing error into failure.

Convert a parsing failure to a success.

Save the parsing state, then run a parser, then restore the state.

Throw a parsing error. By default, parser choice (<|>) can't backtrack on parser error. Use try to convert an error to a recoverable failure.

The failing parser. By default, parser choice (<|>) arbitrarily backtracks on parser failure.

Run a parser on a String input. Reminder: OverloadedStrings for ByteString does not yield a valid UTF-8 encoding! For non-ASCII ByteString literal input, use runParserS or packUTF8 for testing.

Run a parser.

Contains return value and a pointer to the rest of the input buffer.

Constructor for errors which are by default non-recoverable.

Constructor for recoverable failure.

Primitive result of a parser. Possible results are given by OK#, Err# and Fail# pattern synonyms.

Parser e a has an error type e and a return type a.

Higher-level boxed data type for parsing results.

Convert a String to an UTF-8-coded ByteString.

Slice into a ByteString using a Span. The result is invalid if the Span is not a valid slice of the first argument.

isGreekLetter c = ('Α' <= c && c <= 'Ω') || ('α' <= c && c <= 'ω')

isLatinLetter c = ('A' <= c && c <= 'Z') || ('a' <= c && c <= 'z')

isDigit c = '0' <= c && c <= '9'

Byte offset counted backwards from the end of the buffer.

A pair of positions.