Module hierarchy: GHC.Hs (#13009)

Next step into introducing module hierarchy (cf #13009 (closed)):

 hsSyn/HsSyn.hs       GHC/Hs.hs
 hsSyn/HsBinds.hs     GHC/Hs/Binds.hs
 hsSyn/HsDecls.hs     GHC/Hs/Decls.hs
 hsSyn/HsDoc.hs       GHC/Hs/Doc.hs
 hsSyn/HsDumpAst.hs   GHC/Hs/Dump.hs
 hsSyn/HsExpr.hs      GHC/Hs/Expr.hs
 hsSyn/HsExtension.hs GHC/Hs/Extension.hs
 hsSyn/HsImpExp.hs    GHC/Hs/ImpExp.hs
 hsSyn/HsInstances.hs GHC/Hs/Instances.hs
 hsSyn/HsLit.hs       GHC/Hs/Lit.hs
 hsSyn/HsPat.hs       GHC/Hs/Pat.hs
 hsSyn/PlaceHolder.hs GHC/Hs/PlaceHolder.hs
 hsSyn/Convert.hs     GHC/ThToHs.hs
 hsSyn/HsTypes.hs     GHC/Hs/Types.hs
 hsSyn/HsUtils.hs     GHC/Hs/Utils.hs

IMPORTANT: it requires some small changes in haddock and nofib. I have redirected the .gitmodules to my fork so don't merge as is.

The Hs part is not useful (it doesn't mean anything), and the Syn part is gone. Why not GHC/Syn/Binds instead of GHC/Hs/Binds? Syn part is useful I think, it says that these are about front-end syntax.

I read Hs as "Haskell", and I find it useful. It's not Core, it's not STG, it's not C--, etc. Maybe HsSyn? I think I prefer just Hs for brevity, but I could be swayed.

Convert is an uneasy fit here, but I don't see another place to put it. Actually, it's really a conversion pass, converting from the TH syn to the Haskell syn. Maybe it should be named like other conversion passes? Except that it would be all alone, because one module is enough for that whole pass.

Do we want Util or Utils? I guess !1571 (merged) used Utils, so let's just go with that (as you have here).

Maybe DumpAst should just be Dump. If I see the prefix GHC.Hs., I know it's dealing with an AST.

Thanks!

added 1 commit

• 91cdf978 - Module hierarchy: Hs (#13009 (closed))

Compare with previous version

I read Hs as "Haskell", and I find it useful. It's not Core, it's not STG, it's not C--, etc. Maybe HsSyn? I think I prefer just Hs for brevity, but I could be swayed.

I totally agree. Moreover all the datatypes are prefixed with Hs.

Convert is an uneasy fit here, but I don't see another place to put it. Actually, it's really a conversion pass, converting from the TH syn to the Haskell syn. Maybe it should be named like other conversion passes? Except that it would be all alone, because one module is enough for that whole pass.

Maybe GHC.ThToHs to make it similar to other conversion/compilation modules?

Maybe DumpAst should just be Dump. If I see the prefix GHC.Hs., I know it's dealing with an AST.

Done.

changed the description

added 1 commit

• e40d489f - Module hierarchy: Hs (#13009 (closed))

Compare with previous version

added 1 commit

• 2c7ac08f - Module hierarchy: Hs (#13009 (closed))

Compare with previous version

changed the description

added 1 commit

• 452b3411 - Module hierarchy: Hs (#13009 (closed))

Compare with previous version

Maybe GHC.ThToHs to make it similar to other conversion/compilation modules?

Yes, if that's how the others will be named.

added 1 commit

• 84c7011e - Module hierarchy: Hs (#13009 (closed))

Compare with previous version

changed the description

The MR is ready to be merged. nofib and haddock have to be updated first with the commits in my forks. The MR currently modifies .gitmodules but this mustn't be committed.

assigned to @bgamari

added 23 commits

• 84c7011e...7915afc6 - 22 commits from branch ghc:master
• 1b1bf5d1 - Module hierarchy: Hs (#13009 (closed))

Compare with previous version

approved this merge request

added 1 commit

• 41ad3c5b - Module hierarchy: Hs (#13009 (closed))

Compare with previous version

@bgamari Can you merge this? It requires updating haddock and nofib submodules.

added 6 commits

• 41ad3c5b...c77fc3b2 - 5 commits from branch ghc:master
• 572ad909 - Module hierarchy: Hs (#13009 (closed))

Compare with previous version

assigned to @marge-bot and unassigned @bgamari