Text
Text.Parsec
Text.Parsec.ByteString
Text.Parsec.ByteString.Lazy
Text.Parsec.Char
Text.Parsec.Combinator
Text.Parsec.Error
Text.Parsec.Expr
Text.Parsec.Language
Text.Parsec.Perm
Text.Parsec.Pos
Text.Parsec.Prim
Text.Parsec.String
Text.Parsec.Token
ParserCombinators
Text.ParserCombinators.Parsec
Text.ParserCombinators.Parsec.Char
Text.ParserCombinators.Parsec.Combinator
Text.ParserCombinators.Parsec.Error
Text.ParserCombinators.Parsec.Expr
Text.ParserCombinators.Parsec.Language
Text.ParserCombinators.Parsec.Perm
Text.ParserCombinators.Parsec.Pos
Text.ParserCombinators.Parsec.Prim
Text.ParserCombinators.Parsec.Token
Are they the same?
At the moment there are two widely used major versions of Parsec: Parsec 2 and Parsec 3.
My advice is simply to use the latest release of Parsec 3. But if you want to make a conscientious choice, read on.
Parsec 3 introduces a monad transformer, ParsecT
, which can be used to combine
parsing with other monadic effects.
Although Parsec 2 lets you to choose the token type (which is useful when you want to separate lexical analysis from the parsing), the tokens are always arranged into lists. List may be not the most efficient data structure in which to store large texts.
Parsec 3 can work with arbitrary streams -- data structures with a list-like
interface. You can define your own streams, but Parsec 3 also includes a popular
and efficient Stream implementation based on ByteString (for Char
-based
parsing), exposed through the modules Text.Parsec.ByteString
and
Text.Parsec.ByteString.Lazy
.
Advanced features provided by Parsec 3 do not come for free; to implement them several language extensions are required.
Neither of the two versions is Haskell-2010 (i.e. both use extensions), but Parsec 2 uses fewer extensions than Parsec 3, so chances that any given compiler can compile Parsec 2 are higher than those for Parsec 3.
By this time both versions work with GHC, while Parsec 2 is also reported to build with JHC and is included as one of the JHC's standard libraries.
Originally (i.e. as of 3.0 version) Parsec 3 was considerably slower than Parsec 2. However, work on improving Parsec 3 performance has been done, and as of version 3.1 Parsec 3 is only slightly slower than Parsec 2 (benchmarks: 1, 2).
It has been possible to "reimplement" all of the Parsec 2 API in Parsec 3. This
compatibility layer is provided by the Parsec 3 package under the module hierarchy
Text.ParserCombinators.Parsec
(the same hierarchy which is used by Parsec 2),
while the new Parsec 3 API is available under the Text.Parsec
hierarchy.
This means that you can use Parsec 3 as a drop-in replacement for Parsec 2.