added mising numeric_only option for DataFrame.std/var/sem #9209
Conversation
jreback
added
Bug
Numeric Operations
| @@ -11458,6 +11458,21 @@ def test_var_std(self): | |||
| self.assertFalse((result < 0).any()) | |||
| nanops._USE_BOTTLENECK = True | |||
|
|
|||
| def test_numeric_only_flag(self): | |||
| methods = ['sem', 'var', 'std'] | |||
There was a problem hiding this comment.
pls add the issue number in a comment here
|
pls add a release note in the Bug Fix section of 0.16 |
|
@jreback updated, please take a look, thanks |
|
|
||
| df2 = DataFrame(np.random.randn(5, 3), columns=['foo', 'bar', 'baz']) | ||
| # set one entry in a non-existing column to a str | ||
| df2.ix[0, 'a'] = '150' |
There was a problem hiding this comment.
I mean set the entry to 'a', e.g. make it an object series that has a completely invalid string (and not a string-like number)
mortada
force-pushed
the
numeric_only
branch
2 times, most recently
from
c5adc3e to
c74b932
Compare
|
@jreback PR updated, please take a look |
|
I think that if u have a string 100 in a column and the rest are actual float/int values (thus dtype is object) thrn this should always raise a ValueError and not do the coercion it hides errors to evaluate the string as a number |
|
that's a good point. I'll dig into this. |
|
@mortada pls rebase and update according to the last. thanks |
|
pls rebase |
|
@jreback just rebased, sorry I've been pretty swamped and haven't got around to this. But I should have more time going forward. |
mortada
force-pushed
the
numeric_only
branch
2 times, most recently
from
2e51776 to
414b31a
Compare
|
@jreback rebased and addressed the issue you mentioned about type coercion. Please take a look. With this PR This is now consistent with the behavior of other methods such as |
added mising numeric_only option for DataFrame.std/var/sem
|
thanks! |
closes #9201, the
numeric_onlyoption is missing forDataFrame.std()(and alsoDataFrame.var()andDataFrame.sem()), this is a fix for it