BUG: Matched SearchedSorted Signature with NumPy's #12413

gfyoung · 2016-02-22T06:50:35Z

Title is self-explanatory. Closes #12238.

gfyoung · 2016-02-22T06:51:33Z

Where should I put tests for this compatibility issue? It would be great if I could just write one test for all of the classes that have a searchsorted method.

jreback · 2016-02-22T07:58:35Z

test_base will exercise this

gfyoung · 2016-02-22T10:03:59Z

Added test, and the test passes on Travis. Should be good to merge if there is nothing else.

gfyoung · 2016-02-22T10:04:37Z

I need to get a new mouse. In any case, I did let the tests run with my new test, and Travis gave the green light.

jreback · 2016-02-22T13:50:49Z

pandas/tests/test_base.py

+    # See gh-12238
+    def testSearchSorted(self):
+        mixin = IndexOpsMixin()
+        mixin.values = np.array([1, 2, 3])


no, see how we do for example .value_counts where an object is constructed and looped thru. pls follow the same pattern

gfyoung · 2016-02-23T11:14:31Z

Tests are finally passing! Should be good to merge if there is nothing else.

jreback · 2016-02-23T15:22:30Z

doc/source/whatsnew/v0.18.0.txt

@@ -1117,3 +1117,4 @@ Bug Fixes
 - Bug in ``DataFrame.apply`` in which reduction was not being prevented for cases in which ``dtype`` was not a numpy dtype (:issue:`12244`)
 - Bug when initializing categorical series with a scalar value. (:issue:`12336`)
 - Bug when specifying a UTC ``DatetimeIndex`` by setting ``utc=True`` in ``.to_datetime`` (:issue:11934)
+- Bug in ``searchsorted`` argument signatures for ``IndexOpsMixin``, ``DatetimeIndex``, ``PeriodIndex``, and ```TimedeltaIndex`` that broke compatibility with ``searchsorted`` in ``numpy`` (:issue:`12238`)


you don't need to say all this, just say searchsorted now accepts sorter arg for compat with numpy.

jreback · 2016-02-23T15:25:12Z

you can put in 0.18.0

gfyoung · 2016-02-23T17:26:11Z

You accidentally cancelled all of my builds for this PR.

jreback · 2016-02-23T17:26:57Z

oh sorry, push again :)

gfyoung · 2016-02-23T17:38:51Z

FYI: I believe you can restart a build, but no worries here since I just rebased onto master as my new "commit".

gfyoung · 2016-03-16T16:30:18Z

@jreback : Also, regarding that side typo I made with Categorical initially, I don't think that was ever an issue because the line before it made sure that the side parameter was enforced.

jreback · 2016-03-16T16:34:06Z

@gfyoung ok then

jreback · 2016-03-16T16:36:09Z

pandas/core/base.py

+
+        Parameters
+        ----------
+        v : array_like


I know this is annoying, but how about we rename v -> value. More readable args.

In a previous comment we changed it to v so that it matches the signature of numpy, or maybe I misunderstood you then?

no I think it was always like this. But we generally name things as value in pandas (single character is too like q/kdb .... )

Let's maybe just keep it at v (as long as we do it better in our own methods!), you never know that someone used it as a keyword argument

@jorisvandenbossche : v is not used as a keyword in numpy, and from a semantic point of view, I see no reason why it would be.

Yikes, this is news to me. I guess I never encountered that before. Well, at least I know now and am slightly less of a Python noob because of that. Thanks for clarifying @kawochen!

ok, just about to merge, then I saw you changed some of the v -> value. (you missed the datetimelikes).

So originally we had datetimelikes used key (and I had you change it to v). While Series used v.

So we have to break something.

Let's just keep the rename to value (and fix the datetimelikes). Move the whatsnew note to API Changes (and note that the argument name changed). I seriously doubt anyone is actually passing as a kw arg (the first arg).

Okay, so rename to value and make a note in the API changes? Also, what do you mean by "datetimelikes"?

datetimes, timedelta, period

Got it. Done.

codecov-io · 2016-03-16T21:26:04Z

Powered by Codecov. Updated on successful CI builds.

gfyoung · 2016-03-17T11:48:47Z

@jreback : Changes have been made, and Travis approves. Should be good to merge then.

jorisvandenbossche · 2016-03-17T12:56:13Z

Sorry, but I want to say it again. Changing a keyword's name is an API change that can break people's code, and since we want to change it because "we don't like the short name", it is certainly not an unavoidable API change, so IMO should not go into a minor release.

gfyoung · 2016-03-17T12:58:59Z

@jorisvandenbossche : I find myself caught in a tug-of-war between you and @jreback on this issue, not to mention the push to land this is very high. It would be great if some consensus could be found on this.

jreback · 2016-03-17T13:40:36Z

doc/source/whatsnew/v0.18.1.txt

-
-
+- ``searchsorted`` for ``Index`` and ``TimedeltaIndex`` now accept a ``sorter`` argument to maintain compatibility with numpy's ``searchsorted`` function (:issue:`12238`)
+- The parameter for specifying values to be inserted into a ``pandas`` object when calling its ``searchsorted`` method is now uniformly denoted as ``value``


pls just make the whatsnew simpler. just say the signature changed the names of the arguments and put in a pointer to the signature.

Done. I put [ci skip] in my commit because this is only a doc change. So it should be good to merge if there is nothing else.

jreback · 2016-03-17T16:32:59Z

@jorisvandenbossche well, we already have that issue as key is the name for the datetimelikes and v for Series. I guess its ok to move this part of the change to 0.19.0 (the actual name changes). Though really don't think this is a big deal to change this now as its the first argument.

jorisvandenbossche · 2016-03-17T16:46:24Z

But I think it is also not a big deal to just leave it for 0.19.0, just in case to be safe. But I agree the discrepancy with Index key is indeed also not nice.

@gfyoung would you like to split the PR in two? one for the actual fix to searchsorted and one for the name change?

jreback · 2016-03-17T16:49:51Z

ok, let's move the name change to 0.19.0. (you can just create an issue I think and I'll mark it). otherwise ok then. (back out both names changes, IOW reverse the value -> key for datetimelikes)

Closes pandas-devgh-12238.

gfyoung · 2016-03-17T23:14:36Z

@jreback : Travis has finally given the green light on this one as well. Should be good to merge if there is nothing else to add.

jreback · 2016-03-17T23:47:25Z

whoosh! thanks for the patience

gfyoung · 2016-03-17T23:54:43Z

@jreback: No worries. Onto fixing the rest of fromnumeric.py!

gfyoung force-pushed the searchsorted_compat branch from d22a921 to 6c70e12 Compare February 22, 2016 09:25

gfyoung closed this Feb 22, 2016

gfyoung reopened this Feb 22, 2016

jreback reviewed Feb 22, 2016
View reviewed changes

jreback added Indexing Related to indexing on series/frames, not to indexes themselves Compat pandas objects compatability with Numpy or Python functions labels Feb 22, 2016

gfyoung force-pushed the searchsorted_compat branch 4 times, most recently from d3792d2 to 6cb665f Compare February 23, 2016 08:33

jreback reviewed Feb 23, 2016
View reviewed changes

jreback added this to the 0.18.0 milestone Feb 23, 2016

gfyoung force-pushed the searchsorted_compat branch 4 times, most recently from a28afb0 to 62ca892 Compare February 23, 2016 16:57

gfyoung force-pushed the searchsorted_compat branch from 62ca892 to e147977 Compare February 23, 2016 17:29

gfyoung force-pushed the searchsorted_compat branch from 887d4dc to f350442 Compare March 16, 2016 16:30

jreback reviewed Mar 16, 2016
View reviewed changes

gfyoung force-pushed the searchsorted_compat branch 2 times, most recently from aa140f7 to 7e37570 Compare March 16, 2016 17:28

gfyoung force-pushed the searchsorted_compat branch 2 times, most recently from 5b33a6c to 3a0de8b Compare March 17, 2016 04:08

jreback reviewed Mar 17, 2016
View reviewed changes

gfyoung force-pushed the searchsorted_compat branch 2 times, most recently from a30e10f to d7690c2 Compare March 17, 2016 15:48

gfyoung mentioned this pull request Mar 17, 2016

COMPAT: Standardize method signatures #12662

Closed

BUG: Matched searchsorted signature with numpy's

1e6df5e

Closes pandas-devgh-12238.

gfyoung force-pushed the searchsorted_compat branch from d7690c2 to 1e6df5e Compare March 17, 2016 20:30

jreback closed this in 28327ce Mar 17, 2016

gfyoung deleted the searchsorted_compat branch March 17, 2016 23:55



		- ``searchsorted`` for ``Index`` and ``TimedeltaIndex`` now accept a ``sorter`` argument to maintain compatibility with numpy's ``searchsorted`` function (:issue:`12238`)
		- The parameter for specifying values to be inserted into a ``pandas`` object when calling its ``searchsorted`` method is now uniformly denoted as ``value``

BUG: Matched SearchedSorted Signature with NumPy's #12413

BUG: Matched SearchedSorted Signature with NumPy's #12413

Conversation

gfyoung commented Feb 22, 2016

gfyoung commented Feb 22, 2016

jreback commented Feb 22, 2016

gfyoung commented Feb 22, 2016

gfyoung commented Feb 22, 2016

Choose a reason for hiding this comment

gfyoung commented Feb 23, 2016

Choose a reason for hiding this comment

Choose a reason for hiding this comment

jreback commented Feb 23, 2016

gfyoung commented Feb 23, 2016

jreback commented Feb 23, 2016

gfyoung commented Feb 23, 2016

gfyoung commented Mar 16, 2016

jreback commented Mar 16, 2016

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

codecov-io commented Mar 16, 2016

gfyoung commented Mar 17, 2016

jorisvandenbossche commented Mar 17, 2016

gfyoung commented Mar 17, 2016

Choose a reason for hiding this comment

Choose a reason for hiding this comment

jreback commented Mar 17, 2016

jorisvandenbossche commented Mar 17, 2016

jreback commented Mar 17, 2016

gfyoung commented Mar 17, 2016

jreback commented Mar 17, 2016

gfyoung commented Mar 17, 2016