Re: Proposal: Add these two handy functions to Data.List

On Fri, 2010-07-02 at 14:53 -0700, Bryan O'Sullivan wrote:

On Fri, Jul 2, 2010 at 1:59 AM, Conor McBride wrote: The question for library designers is at what level to engage with this structure. In doing so, we should of course take care to protect Joe Programmer from the Screaming Heebie-Jeebies. I am not qualified to judge how best this is to be done, but I thought I might at least offer some of the raw data for that calculation.

I like Cale's proposal, and in fact it reminds me of something Duncan nudged me on a few months ago:

http://www.serpentine.com/blog/2010/07/02/whats-in-a-find-function/

Yes, we discussed revising find yet again on the #ghc channel a month or so ago. Despite having pushed you towards the one with type Text -> Text -> (Text, [(Text, Text)]), I realised I still was not happy with it either. The one we discussed more recently is similar to what you are proposing now. I think the difference was that in what I suggested, the middle component of the tuple went to the end of the text, i.e. it was a prefix of the last component. I think the rationale for that was that you already know what you searched for and we can save time and space by providing the full tail rather than making new texts by recomposing the search term with the trailing string. Duncan