Closed dmcclean closed 9 years ago
Wait, don't merge this. While I was driving home from Akamai I had a thought that it might be wrong "with probability 0" (on the reals) when the interval starts exactly at a local extreme and extends beyond the other local extreme.
Will test and patch in a bit.
OK, fixed that problem. I'm not a big fan of it because it is a bit redundant (lines 107-110 are almost exactly like 101-104) but I can't think of a way to clean it up without either pattern synonyms, view patterns, or complicated pattern guards.
Closing this, as it is better done as part of a larger pull request I have prepared.
Fair nuff
Fixes issue #20 and related issues.
I have doctests that test this, but they also test everything else (#4), and I didn't want to send up one pull request that was so huge.
See if you think that this approach can be modified to work for empty/kaucher intervals or whether you want to go a different way to address #20.