Migrate cabal.project parsing to parsec

[gbaz]

Cabal.project parsing goes through legacy parsers and has a legacy -> current step. Its all rather awkward and also makes figuring out the actual grammar pretty difficult. It would be good to clean this up, capture the current grammar properly, and go through modern parsec parsers.

[ptkato]

I think this could be tied to the exact printing work. A project file doesn't differ much, in terms of how it's written, from a cabal file.

[gbaz]

Well, maybe. As it is, we don't have any sane parsing at all. It would be nice to reuse as much of the cabal parsing machinery as possible, treating projects as a special grammar, at which point we could leverage the exact-print stuff.

[phadej]

The challenge is not using some specific new framework, as parsing the project file would be something like

grammar :: ParsecFieldGrammar Project Project
grammar = Project
    <$> C.monoidalFieldAla "packages"          (C.alaList' C.FSep PackageLocation) prjPackagesL
    <*> C.monoidalFieldAla "optional-packages" (C.alaList' C.FSep PackageLocation) prjOptPackagesL
    <*> C.monoidalFieldAla "constraints"       (C.alaList' C.CommaVCat NoCommas)   prjConstraintsL
    <*> C.monoidalFieldAla "allow-newer"       (C.alaList' C.CommaVCat NoCommas)   prjAllowNewerL
    <*> C.booleanFieldDef  "reorder-goals"                                         prjReorderGoalsL False
    <*> C.optionalFieldAla "max-backjumps"     Int'                                prjMaxBackjumpsL
    <*> C.optionalFieldDef "optimization"                                          prjOptimizationL OptimizationOn

... but actually figuring out what should or shouldn't be parsed.

The cabal.project parser reuses parts ~/.cabal/config parsing, there is some magic tying command line parsing in as well.

It's very messy, and not specified, there are only examples and some stuff which works, maybe intentionally, maybe accidentally.

I wouldn't tie exact print work with this. They are different challenges.

(EDIT: Just replicating what the current code does may be beneficial, already but i'm sure there are plenty of low-mid hanging cleanup&improvement opportunities).

[mmhat]

I think this can be closed since it is a duplicate of #6101.