Bugs item #940294, was opened at 2004-04-22 13:16 Message generated for change (Comment added) made by nobody You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=108032&aid=940294&group_id=8032
Category: libraries (other) Group: 6.2.1 Status: Open Resolution: None Priority: 5 Submitted By: Jon Cast (jcast) Assigned to: Nobody/Anonymous (nobody) Summary: Language.Haskell.Pretty misformats Initial Comment: ghci gives the following results: $ ghci -fglasgow-exts ___ ___ _ / _ \ /\ /\/ __(_) / /_\// /_/ / / | | GHC Interactive, version 6.2.1, for Haskell 98. / /_\/ __ / /___| | http://www.haskell.org/ghc/ \____/\/ /_/\____/|_| Type :? for help. Loading package base ... linking ... done. Prelude> :m + Language.Haskell.Syntax Prelude Language.Haskell.Syntax> :m + Language.Haskell.Pretty Prelude Language.Haskell.Syntax Language.Haskell.Pretty> putStrLn $ prettyPrint $ HsInfixApp (HsDo [HsQualifier (HsVar (UnQual (HsIdent "undefined"))), HsQualifier (HsVar (UnQual (HsIdent "undefined")))]) (HsQVarOp (UnQual (HsSymbol "<|>"))) (HsDo [HsQualifier (HsVar (UnQual (HsIdent "undefined")))]) Loading package haskell98 ... linking ... done. Loading package haskell-src ... linking ... done. do undefined undefined <|> do undefined Prelude Language.Haskell.Pretty Language.Haskell.Syntax> putStrLn $ prettyPrint $ HsApp (HsVar (UnQual (HsIdent "foo"))) $ HsApp (HsVar (UnQual (HsIdent "bar"))) $ HsVar (UnQual (HsIdent "baz")) foo bar baz In other words, it mis-indents infix applications and does /not/ parenthesize nested applications. There may be other mis-formatting issues, but these are the ones that have me stopped atm. ---------------------------------------------------------------------- Comment By: Nobody/Anonymous (nobody) Date: 2004-04-23 16:55 Message: Logged In: NO I'm not clear on what behaviour you want for the original example. It must indent the <|>, because the do might itself be a statement of another do expression. In the bigger one, you've explicitly asked for an onsideIndent that's the same as the doIndent, when it should always be smaller. I'll add this and similar preconditions to the documentation. The documentation is also lacking in relation to HsParen. I'm reluctant to improve pretty by adding parentheses, partly because it's work, and partly because it's only a partial solution and it's not clear how infix operators should be handled. -- Ross ---------------------------------------------------------------------- Comment By: Jon Cast (jcast) Date: 2004-04-23 15:14 Message: Logged In: YES user_id=762054 Patterson: Well, the first one is more of a QOI issue (and possibly debatable, although I can't imagine who'd want that behavior). As for the second issue: are you saying that (HsApp <foo> (HsApp <bar> <baz>)) is simply illegal or in some way wrong? That doesn't seem to be documented, and seems like a bad idea to me anyway. IMHO, it's the most natural use from the POV of someone reading the Haddock documentation HsParen is very hard to find, and it's not at all obvious that it is /required/ whenever the Haskell syntax requires parentheses. But, I was inspired to report the first issue by the following (more complex) test case, which I'm going to report in full here as it's not too long and definitely shows incorrect behavior: putStrLn $ prettyPrintStyleMode style (PPHsMode 2 2 2 2 2 2 True PPOffsideRule False False) $ (HsInfixApp (HsDo [HsQualifier (HsApp (HsVar (UnQual (HsIdent "reserved"))) (HsLit (HsString "LineTo"))), HsGenerator (SrcLoc {srcFilename = "stdin", srcLine = 1, srcColumn = 1}) (HsPVar (HsIdent "curveSegmentEndPoint")) (HsVar (UnQual (HsIdent"parse"))), HsQualifier (HsApp (HsVar (UnQual (HsIdent "return"))) (HsApp (HsCon (UnQual (HsIdent "CurveLineTo"))) (HsVar (UnQual (HsIdent "curveSegmentEndPoint")))))]) (HsQVarOp (UnQual (HsSymbol "<|>"))) (HsDo [HsQualifier (HsApp (HsVar (UnQual (HsIdent "reserved"))) (HsLit (HsString "LineAlong"))),HsGenerator (SrcLoc {srcFilename = "stdin", srcLine = 1, srcColumn = 1}) (HsPVar (HsIdent "curveSegmentAlong")) (HsVar (UnQual (HsIdent "parse"))),HsQualifier (HsApp (HsVar (UnQual (HsIdent "return"))) (HsApp (HsCon (UnQual (HsIdent "CurveAlong"))) (HsVar (UnQual (HsIdent "curveSegmentAlong")))))])) do reserved "LineTo" curveSegmentEndPoint <- parse return CurveLineTo curveSegmentEndPoint <|> do reserved "LineAlong" curveSegmentAlong <- parse return CurveAlong curveSegmentAlong Note that the operator here is indented by /two/ spaces, same as the statements in the first do expression. This behavior can, oddly enough, be fixed by using defaultMode instead of an explicit PPHsMode constructor, but I can't find any other work-around. ---------------------------------------------------------------------- Comment By: Nobody/Anonymous (nobody) Date: 2004-04-22 16:10 Message: Logged In: NO I don't see the problem with the first one: <|> must be indented less than the do body to force a }, but not so much that it forces a ; in an outer layout context. Regarding the second one, prettyPrint never inserts parentheses in expressions. You have to use HsParen, and the parser leaves those in the parse tree. If you're generating source, you'll need to add it. Maybe it should be smarter than that, but it would still be limited, since there's no information about operator fixity. -- Ross Paterson ---------------------------------------------------------------------- You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=108032&aid=940294&group_id=8032 _______________________________________________ Glasgow-haskell-bugs mailing list [EMAIL PROTECTED] http://www.haskell.org/mailman/listinfo/glasgow-haskell-bugs
