Re: [trac@galois.com: Re: [GHC] #1218: Add sortNub and sortNubBy to Data.List]

On Thu, 2007-03-22 at 12:10 +0100, Nils Anders Danielsson wrote:

On Thu, 22 Mar 2007, Duncan Coutts wrote:

...

It's interesting to note that the implementations in base for some functions in Data.List have been inconsistent with the report for years and nobody has noticed (or at least if they noticed they didn't complain very loudly).

I have certainly found some discrepancies in the past. If you never reason about partial or infinite values you're not likely to stumble over these inconsistencies, though.

Certainly. I've come across some some very subtle examples that I'd not have spotted normally. I came across your ChasingBottoms lib yesterday. It's very nice. I'd started to implement something similar to test the strictness properties of our list reimplementation but what you've got is a good deal more complete. I started by looking at Olaf Chitil's StrictCheck module which uses your lib however the SYB technique it uses to generate partial values of various types doesn't help when it comes to function types. So instead I modified SmallCheck to generate values with bottoms in them. That copes with function types very nicely. It can generate all the partial monotone functions up to a given depth ( I think :-) ). My one gripe with ChasingBottoms is that I can't customise the ApproxShow instances. I want to show functions in the SmallCheck style but deal with _|_'s properly. SmallCheck defines a Show instance for functions that shows the mappings up to a certain depth. It's very cute. I can't define my own ApproxShow instance for functions since it overlaps with the Data a => ApproxShow a instance. Duncan