API: multi-line, not inplace eval by chris-b1 · Pull Request #11149 · pandas-dev/pandas (original) (raw)

Conversation

This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.Learn more about bidirectional Unicode characters

[ Show hidden characters]({{ revealButtonHref }})

closes #9297
closes #8664

This is just a proof of concept at this point - the idea is to allow multiple assignments within an eval block, and also include an inplace argument (currently default True to match old behavior, although maybe that should be changed) for chaining.

In [15]: df = pd.DataFrame({'a': np.linspace(0, 100), 'b': range(50)})

In [16]: df.eval("""
    ...: c = a * 2
    ...: d = c / 2.0
    ...: e = a + b + c""")

In [17]: df.head()
Out[17]: 
        a  b          c         d          e
0  0.000000  0   0.000000  0.000000   0.000000
1  2.040816  1   4.081633  2.040816   7.122449
2  4.081633  2   8.163265  4.081633  14.244898
3  6.122449  3  12.244898  6.122449  21.367347
4  8.163265  4  16.326531  8.163265  28.489796

In [18]: df.eval("""
    ...: f = e * 2
    ...: g = f * 1.5""", inplace=False).head()
Out[18]: 
        a  b          c         d          e          f          g
0  0.000000  0   0.000000  0.000000   0.000000   0.000000   0.000000
1  2.040816  1   4.081633  2.040816   7.122449  14.244898  21.367347
2  4.081633  2   8.163265  4.081633  14.244898  28.489796  42.734694
3  6.122449  3  12.244898  6.122449  21.367347  42.734694  64.102041
4  8.163265  4  16.326531  8.163265  28.489796  56.979592  85.469388

jreback added this to the Next Major Release milestone

Oct 8, 2015

jreback

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

add the issue as a comment

@chris-b1 this looks reasonable. I think that we should change the inplace to False by default. That obviously is a breaking change. So what I think we could do is this:

make inplace=None the default
detect if an assignment is being made. if inplace is None, then show a deprecation warning.
this implicity makes inplace=None -> inplace=False for non-assignments (which is what it currently does, and is quite reasonable). To get inplace then assignment you would have to be explict about passing inplace=True, which is what we would want.

yes?

Yes, this looks great!

I agree that we should change the default behavior to inplace=True in a version or two. I think @jreback has the right idea on the migration path.

@chris-b1 going to mark this for 0.18.0 as putting the deprecation warning in-place (pun intended)

@chris-b1 want to update / add in deprecation warning as discussed

@jreback - this is updated - added a whatsnew and deprecation warnings.

chris-b1 changed the title~~Experimental/WIP: multi-line, not inplace eval~~ API: multi-line, not inplace eval

Dec 6, 2015

jreback

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

let's add .assign in the See also (as .assign is basically what df.eval('c=a+b',inplace=False) IS)

pls review the eval docs for any updates (in enhancingperf.rst)

@jreback
Updated the enhancingperf docs, and added a fix/test for #8664

jreback

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

add a versionadded tag for inplace

jreback

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

add a versionadded

lgtm. couple of doc changes. squash & ping when green.

Updated for your doc comment and PEP8 - nothing major but did have to add a couple # noqa where a name was only used in an eval block - is that the right approach?

merged via 99c2d93

thanks for the nice changes!

yeh the # noqa is fine (I took one out, but otherwise ok).

This was referenced

Jan 4, 2016