Implement Akima1DInterpolator #12833

gliptak · 2016-04-09T00:34:14Z

closes Add Akima Interpolation to Pandas #7588
tests added / passed
passes git diff upstream/master | flake8 --diff
whatsnew entry

gliptak · 2016-04-09T14:56:06Z

This build failed https://travis-ci.org/pydata/pandas/jobs/121843938 likely because of the Scipy version.

axis was added to Akima1DInterpolator at 0.17.0 as per scipy/scipy@b0395b4

Is there a minimum version of Scipy for pandas? What would be a good way to handle lower versions?

Thanks

jreback · 2016-04-09T14:56:19Z

pandas/tests/test_generic.py

+                                           50.75]))
+        interp_s = ser.reindex(new_index).interpolate(method='akima')
+        # does not blow up, GH5977
+        interp_s[49:51]


pls have an expected case and use assert_series_equal

Added assert

jreback · 2016-04-09T14:58:22Z

you can you call this in a try: except to handle the scipy signature differences. (then if it doesn't work in the excpet, raise again). we don't have a min version.

jreback · 2016-04-09T20:40:38Z

pandas/core/missing.py

+    try:
+        P = interpolate.Akima1DInterpolator(xi, yi, axis=axis)
+    except TypeError:
+        # Scipy earlier than 0.17.0 missing axis


so need a test for this. on the 2.7_LOCALE build we install a really old scipy. We may need to install a somewhat newer one on another build to explicity test for this. When did this Akima1DInterpolate appear in scipy? and seems the axis parm was added in 0.17.0?

0.14.0 (as per scipy/scipy@a0a3b38) I implemented a different check.

ok the 3.4 build has scipy==0.14.0 installed so pls verify that the test works corretctly (IOW it is NOT skipped, but passes).

remove scipy from ci/requirements-3.5_OSX. Make sure that you DO see some skips. I don't think we were testing having NO scipy.

ok , just remove scipy from ci/requirements-3.5_OSX and I think will be good to go.

from ci/requirements-3.5_OSX.run ?

yep (we are including a version of scipy on EVERY build, so need to take it from one)

gliptak · 2016-04-09T21:44:20Z

The two test failures seem to be unrelated.

gliptak · 2016-04-10T12:45:27Z

The two test failures seem to be unrelated.

jreback · 2016-04-10T15:48:09Z

pandas/core/missing.py

                              "PCHIP interpolation.")

+    try:


can you change this (and the pchip one to something like):

if method == 'akima': alt_methods['akima'] = _akima_interpolate

Move to _akima_interpolate

from scipy.interpolate import Akima1DInterpolator # noqa

jreback · 2016-04-10T15:48:58Z

pls add a whatsnew as well

gliptak · 2016-04-10T16:07:22Z

Added whatsnew

jreback · 2016-04-10T16:16:53Z

pandas/core/missing.py

-    except ImportError:
-        if method == 'akima':
+    if method == 'akima':
+        try:


yes this is fine

jreback · 2016-04-10T16:22:23Z

pls squash

jreback · 2016-04-10T16:23:58Z

add a small note when to use this in : http://pandas-docs.github.io/pandas-docs-travis/missing_data.html#interpolation (you might want to break up that section a bit and make it into a table)

jreback · 2016-04-10T16:24:19Z

add to the doc-string for .interpolate with a versionadded tag

gliptak · 2016-04-10T17:01:00Z

Added notes and versionadded tag

jreback · 2016-04-10T17:03:52Z

pandas/core/missing.py

@@ -101,6 +101,7 @@ def interpolate_1d(xvalues, yvalues, method='linear', limit=None,

    Bounds_error is currently hardcoded to False since non-scipy ones don't
    take it as an argumnet.
+    .. versionadded:: 0.18.1 support for 'akima'.


no this needs to be updated in pandas.core.generic.NDFrame.interpolate

Moved section.

gliptak · 2016-04-10T18:07:25Z

Updated per comments.

jreback · 2016-04-10T18:15:10Z

doc/source/missing_data.rst

+
+* If you are dealing with a time series that is growing at an increasing rate,
+``method='quadratic'`` may be appropriate.
+* If you have values approximating a cumulative distribution function,


jreback · 2016-04-10T19:11:10Z

ok, looks good. ping when green (and pls verify that tests are skipped as appropriate)

jorisvandenbossche · 2016-04-10T19:26:42Z

doc/source/missing_data.rst

-distribution function, then ``method='pchip'`` should work well.
+
+* If you are dealing with a time series that is growing at an increasing rate,
+``method='quadratic'`` may be appropriate.


You need two spaces here at the beginning of this line, so the start of the text is aligned with the start of the text in the previous line ('If ...')

jorisvandenbossche · 2016-04-10T19:28:48Z

pandas/core/generic.py

              wrappers around the scipy interpolation methods of similar
              names. These use the actual numerical values of the index. See
              the scipy documentation for more on their behavior
              `here <http://docs.scipy.org/doc/scipy/reference/interpolate.html#univariate-interpolation>`__  # noqa
              `and here <http://docs.scipy.org/doc/scipy/reference/tutorial/interpolate.html>`__  # noqa

+            .. versionadded:: 0.18.1 support for 'akima'.


The 'support for 'akima'' should go on the next line I think. See the sphinx docs: http://www.sphinx-doc.org/en/stable/markup/para.html#directive-versionadded for an example. And then maybe change the text to something like "Added support for the 'akima' method"

jorisvandenbossche · 2016-04-10T19:30:19Z

@gliptak I just added some more doc comments, looks very good for the rest!

gliptak · 2016-04-11T02:19:53Z

@jreback Build is green. Tests skipped https://s3.amazonaws.com/archive.travis-ci.org/jobs/122101334/log.txt

#268 pandas.tests.test_generic.TestSeries.test_interpolate_akima: no scipy.stats module
#269 pandas.tests.test_generic.TestSeries.test_interpolate_corners: no scipy.stats module
#270 pandas.tests.test_generic.TestSeries.test_interpolate_pchip: no scipy.stats module

I also opened #12856 (consistently fails the same way)

jreback · 2016-04-11T15:05:07Z

@gliptak thanks!

pls review the built docs for accuracy (and good looking ness) and issue a follow up if necessary. (prob will take a few hours before they are built)

gliptak · 2016-04-11T15:12:38Z

@jreback @jorisvandenbossche Thank you for guiding this through

gliptak · 2016-04-11T22:32:44Z

Docs updated good.

jreback reviewed Apr 9, 2016
View reviewed changes

jreback added Enhancement Missing-data np.nan, pd.NaT, pd.NA, dropna, isnull, interpolate labels Apr 9, 2016

jreback reviewed Apr 9, 2016
View reviewed changes

jreback reviewed Apr 10, 2016
View reviewed changes

pandas/core/missing.py

except ImportError:

if method == 'akima':

if method == 'akima':

try:

Copy link

Contributor

jreback Apr 10, 2016

Choose a reason for hiding this comment

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

yes this is fine

jreback reviewed Apr 10, 2016
View reviewed changes

jreback added this to the 0.18.1 milestone Apr 10, 2016

jorisvandenbossche reviewed Apr 10, 2016
View reviewed changes

Implement Akima1DInterpolator

a63dc31

jreback closed this in 1c8816f Apr 11, 2016

gliptak mentioned this pull request Apr 11, 2016

BLD: Travis dedup via webhooks #12438

Closed

gliptak deleted the akima branch April 12, 2016 19:52

Implement Akima1DInterpolator #12833

Implement Akima1DInterpolator #12833

Conversation

gliptak commented Apr 9, 2016

gliptak commented Apr 9, 2016

Choose a reason for hiding this comment

Choose a reason for hiding this comment

jreback commented Apr 9, 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

gliptak commented Apr 9, 2016

gliptak commented Apr 10, 2016

Choose a reason for hiding this comment

Choose a reason for hiding this comment

jreback commented Apr 10, 2016

gliptak commented Apr 10, 2016

Choose a reason for hiding this comment

jreback commented Apr 10, 2016

jreback commented Apr 10, 2016

jreback commented Apr 10, 2016

gliptak commented Apr 10, 2016

Choose a reason for hiding this comment

Choose a reason for hiding this comment

gliptak commented Apr 10, 2016

Choose a reason for hiding this comment

jreback commented Apr 10, 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

jorisvandenbossche commented Apr 10, 2016

gliptak commented Apr 11, 2016

jreback commented Apr 11, 2016

gliptak commented Apr 11, 2016

gliptak commented Apr 11, 2016