Rice distribution: extend to b=0, add an explicit rvs method.

[ev-br]

closes gh-2164 and gh-2742.

As a bonus, rvs get a bit faster. With current master:

In [10]: scipy.__version__
Out[10]: '0.14.0.dev-efbde60'

In [11]: %timeit stats.rice.rvs(b=3.)
100 loops, best of 3: 4.18 ms per loop

In [12]: %timeit stats.rice.rvs(b=3., size=1000)
1 loops, best of 3: 4.13 s per loop

with this PR:

In [6]: scipy.__version__
Out[6]: '0.14.0.dev-bf7d8d9'

In [7]: %timeit stats.rice.rvs(b=3.)
10000 loops, best of 3: 45.5 µs per loop

In [8]: %timeit stats.rice.rvs(b=3., size=1000)
10000 loops, best of 3: 192 µs per loop

[josef-pkt]

nice, ticket count is going down

looks good to me

[rgommers]

I think you meant to use assert_allclose here?

[rgommers]

Looks good. One thing to add would be to remove 'rice' from the list of slow distributions in test_continuous_basic.py

[ev-br]

made changed suggested by @rgommers, rebased on master, pushed the rebased version.

[rgommers]

Slight git mishap here....

[ev-br]

ouch.
I screwed it up.
any suggestion of how to fix this mess? Or is it easier to close this, apply the changes to the master, and send a fresh PR?

[rgommers]

You can create a new branch from current master, cherry-pick the relevant commits and then force push to this PR ($ git push origin newbranch:rice_enh).

[ev-br]

@rgommers done just that, thanks a whole lot!
Just for me to understand (and avoid in the future): what was the error?
What I did was

• fetch upstream master and merge it to the local master
• rebased the feature branch on local master
• tried pushing to the origin, which failed
• pulled the feature branch from the origin, resolved merge conflicts,
• pushed the feature branch back to origin
  As there was no force-pushing involved, I thought it'd to the right thing.
  Apparently not...

[josef-pkt]

after rebase you need to force push to your github branch, to overwrite it

[rgommers]

Yep. Step 1, 2 are correct, 3 needed -f.

[rgommers]

Merged, thanks!