WRN: let + * and - pass thru on boolean with a warning

[cpcloud]

closes #7210

[jreback]

maybe make it a UserWarning as its not deprecated per se, just not supported really ?

[cpcloud]

ok

[jorisvandenbossche]

not supported also sounds a bit strange (as this is not clear that your code will actually work). In what regard is it not supported? Or is it just not recommended as it can have performance consequences?

[cpcloud]

ah right ... that reminds me i actually need to test this with numexpr which was the whole point! numexpr doesn't support boolean arithmetic, in this case that's what I mean when I say not supported

[jreback]

I thought u were going to do a warning, then just translate the symbols and do it anyhow

(e.g. '*' -> '&') etc

[jorisvandenbossche]

ah, I thought you were going to detect this and then just use python evaluation and not numexpr (in that sense my comment above).

What I mean is: not supported <-> but my code will still run is not clear from the message now.

[cpcloud]

i see @jreback i misunderstood what you meant by translation...that's actually probably the simplest method

[jreback]

yes I agree with @jorisvandenbossche (that's even better), but I would still do the warning; you can just bail out of the evaluation and 'fallback' to python eval

[cpcloud]

ok so translate or no? very simple and would keep the performance for those ops (not that they are really used that much)

[jreback]

@cpcloud I think just fallback as @jorisvandenbossche suggests (with a warning)

[cpcloud]

ok

[cpcloud]

how about 'operator "BLAH" not supported by numexpr for bool dtype, use "BLARG" instead'?

[cpcloud]

still not totally clear that it works tho

[cpcloud]

@jorisvandenbossche ok with the latest message?

[jorisvandenbossche]

ok!

[jreback]

Looks good to me

need to adjust the release notes / v0.14.0.txt to reflect this (and add this issue number too).
go ahead do that and squash and can merge when ready.

In [2]: df = DataFrame(dict(A=np.random.randn(50000))).isnull()

In [3]: df*df
pandas/computation/expressions.py:184: UserWarning: evaluating in Python space because the '*' operator is not supported by numexpr for the bool dtype, use '&' instead
  unsupported[op_str]))
Out[3]: 
           A
0      False
1      False
2      False
3      False
4      False
5      False
6      False
7      False
8      False
9      False
10     False
11     False
12     False
13     False
14     False
15     False
16     False
17     False
18     False
19     False
20     False
21     False
22     False
23     False
24     False

[cpcloud]

@jreback good on the rls notes?

[jreback]

yep