id	summary	reporter	owner	description	type	status	priority	milestone	version	resolution	keywords	cc
59	Template Haskell generated declarations problematic	nomeata		"Hi,

with haddock > 2.0, it is possible to generate documentation for sources with TH generated top level declartaions. This is very handy, especially as it is often not clear to the user of a library, what declarations a TH function generates, and what types they have.

With haddock-2.2.2, I observed that haddock includes ''some'' TH generated declarations. Aside the fact that TH-generated functions need to have TH-generated type signatures for haddock to notice them, just like manual functoins, it skipped some of them without apparent reason.

With some code reading, I found out that haddock throws aways all but one declaration per source location – but with Template Haskell, there can be more than one per source location! The following patch fixes it for me:

{{{
diff -rN -u old-haddock/src/Haddock/Interface/Create.hs new-haddock/src/Haddock/Interface/Create.hs
--- old-haddock/src/Haddock/Interface/Create.hs	2008-10-05 01:40:22.785419074 +0200
+++ new-haddock/src/Haddock/Interface/Create.hs	2008-10-05 01:40:22.813422082 +0200
@@ -327,9 +327,7 @@
 
     _ -> case d of
       Nothing -> collect (Just e) doc_so_far es
-      Just d0
-        | sameDecl d0 e -> collect d doc_so_far es  
-        | otherwise -> finishedDoc d0 doc_so_far (collect (Just e) DocEmpty es)
+      Just d0 -> finishedDoc d0 doc_so_far (collect (Just e) DocEmpty es)
 
 
 finishedDoc :: LHsDecl Name -> HsDoc Name -> [DeclWithDoc] -> [DeclWithDoc]
@@ -340,10 +338,6 @@
     notDocDecl _              = True
 finishedDoc _ _ rest = rest
 
-
-sameDecl d1 d2 = getLoc d1 == getLoc d2
-
-
 mkSubMap :: HsGroup Name -> Map Name [Name]
 mkSubMap group = Map.fromList [ (name, subs) | L _ tycld <- hs_tyclds group,
  let name:subs = map unLoc (tyClDeclNames tycld) ]
}}}

I’m not entirely sure what the side effects are. In what cases are more than one declarations per source location where one of them should be dropped? I don’t think that that happens often, because (if I read the code correctly), it is not specified ''which'' declaration should be dropped in such a case.

I have attached a darcs patch with this change. I’d greatly welcome better TH support in haddock."	defect	closed	major			fixed	TH