Thank you for doing this Benedict. I've added your more detailed comments to ticket #2393 so that they are preserved. Ian: would you like to apply? I'm not sure how to integrate the QuickCheck tests, but I bet you know. Benedict: while you are in the area, would you like to take a swing at http://hackage.haskell.org/trac/ghc/ticket/1337, and 1176? Simon | -----Original Message----- | From: libraries-bounces@haskell.org [mailto:libraries-bounces@haskell.org] On Behalf Of Benedikt Huber | Sent: 24 June 2008 13:12 | To: libraries@haskell.org | Subject: [Ticket #2393] Text.PrettyPrint.HughesPJ: Bug fixes, performance improvement | | Hello, | | I'd like to propose bugfixes, documentation fixes and a performance | improvement for Text.PrettyPrint.HughesPJ. The changes shouldn't | effect the expected behaviour of the PP library. | | I've written a QuickCheck test suite for the pretty printer (to test | the improvement), and found two bugs and some misconceptions/ | ambiguities in the documentation. Additionally, there is a | microbenchmark for the suggested improvement. | Both are available at http://code.haskell.org/~bhuber/Text/ | PrettyPrint/. Note that the QuickCheck tests need access to all top- | level names in HughesPJ (i.e. ignore the export list). | | In summary, I propose to | * fix a bug in fillNB and one in fillNB/sepNB | * correct documentation on laws and invariants. | * add more efficient implementations of vcat,hsep,hcat | | More specifically: | | (1) Bugfix fillNB: Additional case for fillNB Empty (Empty : ys) | | (2) Bugfix [f](cat|sep): do not allow overlapping ($$) in vertical | composition | | (3) Lazy implementations of vcat,hcat and hsep | | (4) Law <t2> isn't always true | | (5) Invariant 5 should be made stronger | | (6) Change the comment about negative indentation |