ENH: DataFrame.corr(method='spearman') is cythonized.

[jniznan]

It should be more than 10 times faster than the old version.

[jreback]

ok....can you put a release notes mention? (enhancement section),

and if you are gung ho, pls add...a vbench entry (vb_suite/stat_ops.py)....it could have a start date of say 2013/1/1 (or earlier if you can figure out about when spearman was added....was sometime last year, but I don't remember)

[cpcloud]

@jreback is there any consensus about fused types/memoryviews? if there's a consensus to use should we "require" that new cython use it?

[jreback]

should use them
but in this case doesn't matter as float64 is all that is needed

[cpcloud]

fused types sure but what about memoryviews here?

[jreback]

if u can get them to work
should do it

[cpcloud]

why not just make this default to 1 and then u can give it a type of Py_ssize_t...

[jreback]

@jniznan does @cpcloud change make sense?

[jniznan]

Yes, I will push the changes in the next day or two.

[jreback]

@jniznan can you see if you can finish this up ? thanks

[jniznan]

@jreback sorry about the delay,
it seems that the spearman functionality was added on 2011/12/3 (GH #428) , so i will put 2011/12/4 in the benchmark.

[jniznan]

@jreback i put in the release notes mention and the vbench entry, i have also implemented the change @cpcloud proposed

[cpcloud]

@jniznan can u rebase?

[jniznan]

@cpcloud it should be rebased now.

[jreback]

Merged...thanks.....nice improvements

-------------------------------------------------------------------------------
Test name                                    | head[ms] | base[ms] |  ratio   |
-------------------------------------------------------------------------------
stats_corr_spearman                          |  83.9807 | 3735.8600 |   0.0225 |

Target [a70b48a] : PERF: stats_corr a smaller vbsuite
Base   [c4a218c] : Merge pull request #3929 from y-p/PR_PTF_no_more