PERF: indexing #43274

debnathshoham · 2021-08-28T18:07:09Z

closes #xxxx
tests added / passed
Ensure all linting tests pass, see here for how to run them
whatsnew entry

Added a check as suggested in #39280 (comment)

In [1]: from asv_bench.benchmarks.indexing import InsertColumns

In [2]: self = InsertColumns()

In [3]: self.setup()

In [4]: %timeit self.time_assign_with_setitem()
27.6 ms ± 68.1 µs per loop (mean ± std. dev. of 7 runs, 10 loops each) #master

In [4]: %timeit self.time_assign_with_setitem()
8.6 ms ± 6.43 µs per loop (mean ± std. dev. of 7 runs, 100 loops each) #PR

Edit:

       before           after         ratio
     [2b3a9b16]       [fa9c7140]
     <master>         <indexing.InsertColumns.time_assign_with_setitem>
-         179±4μs        163±0.6μs     0.91  indexing.NumericSeriesIndexing.time_iloc_array(<class 'pandas.core.indexes.numeric.UInt64Index'>, 'unique_monotonic_inc')
-        838±50μs          759±7μs     0.90  indexing.NonNumericSeriesIndexing.time_getitem_list_like('datetime', 'nonunique_monotonic_inc')
-        37.7±4μs       34.0±0.5μs     0.90  indexing.NumericSeriesIndexing.time_getitem_slice(<class 'pandas.core.indexes.numeric.UInt64Index'>, 'nonunique_monotonic_inc')
-     2.91±0.09ms       2.61±0.1ms     0.90  indexing.NumericSeriesIndexing.time_loc_list_like(<class 'pandas.core.indexes.numeric.UInt64Index'>, 'unique_monotonic_inc')
-        60.1±2μs       54.0±0.5μs     0.90  indexing.NumericSeriesIndexing.time_iloc_list_like(<class 'pandas.core.indexes.numeric.Int64Index'>, 'unique_monotonic_inc')
-     1.56±0.03ms      1.40±0.04ms     0.90  indexing.NumericSeriesIndexing.time_loc_list_like(<class 'pandas.core.indexes.numeric.Int64Index'>, 'nonunique_monotonic_inc')
-        183±20μs          164±1μs     0.89  indexing.NumericSeriesIndexing.time_iloc_array(<class 'pandas.core.indexes.numeric.Int64Index'>, 'unique_monotonic_inc')
-     2.39±0.05ms         2.14±0ms     0.89  indexing.NumericSeriesIndexing.time_getitem_list_like(<class 'pandas.core.indexes.numeric.UInt64Index'>, 'nonunique_monotonic_inc')
-        879±30μs         780±50μs     0.89  indexing.CategoricalIndexIndexing.time_get_indexer_list('monotonic_incr')
-        184±20μs          163±1μs     0.88  indexing.NumericSeriesIndexing.time_iloc_array(<class 'pandas.core.indexes.numeric.Int64Index'>, 'nonunique_monotonic_inc')
-         131±7μs          115±4μs     0.87  indexing.NumericSeriesIndexing.time_loc_scalar(<class 'pandas.core.indexes.numeric.Float64Index'>, 'nonunique_monotonic_inc')
-     3.02±0.03ms      2.63±0.09ms     0.87  indexing.NumericSeriesIndexing.time_loc_list_like(<class 'pandas.core.indexes.numeric.Float64Index'>, 'nonunique_monotonic_inc')
-     3.19±0.04ms      2.78±0.04ms     0.87  indexing.NumericSeriesIndexing.time_getitem_list_like(<class 'pandas.core.indexes.numeric.Float64Index'>, 'nonunique_monotonic_inc')
-      11.7±0.2ms       10.2±0.2ms     0.87  indexing.NumericSeriesIndexing.time_loc_array(<class 'pandas.core.indexes.numeric.Float64Index'>, 'unique_monotonic_inc')
-     1.89±0.04ms         1.65±0ms     0.87  indexing.NumericSeriesIndexing.time_getitem_list_like(<class 'pandas.core.indexes.numeric.Int64Index'>, 'unique_monotonic_inc')
-      4.63±0.1ms      3.99±0.02ms     0.86  indexing.NumericSeriesIndexing.time_getitem_array(<class 'pandas.core.indexes.numeric.Int64Index'>, 'unique_monotonic_inc')
-     1.82±0.04ms      1.56±0.01ms     0.86  indexing.NumericSeriesIndexing.time_getitem_list_like(<class 'pandas.core.indexes.numeric.Int64Index'>, 'nonunique_monotonic_inc')
-      3.69±0.1ms       3.14±0.1ms     0.85  indexing.NumericSeriesIndexing.time_loc_list_like(<class 'pandas.core.indexes.numeric.Float64Index'>, 'unique_monotonic_inc')
-     4.08±0.06ms      3.43±0.02ms     0.84  indexing.NumericSeriesIndexing.time_getitem_list_like(<class 'pandas.core.indexes.numeric.Float64Index'>, 'unique_monotonic_inc')
-        150±30μs          120±3μs     0.80  indexing.NumericSeriesIndexing.time_loc_slice(<class 'pandas.core.indexes.numeric.UInt64Index'>, 'unique_monotonic_inc')
-        961±30μs         762±20μs     0.79  indexing.CategoricalIndexIndexing.time_get_indexer_list('monotonic_decr')
-        295±30μs          231±3μs     0.78  indexing.NumericSeriesIndexing.time_loc_slice(<class 'pandas.core.indexes.numeric.Int64Index'>, 'nonunique_monotonic_inc')
-      1.11±0.3ms         825±20μs     0.74  indexing_engines.NumericEngineIndexing.time_get_loc((<class 'pandas._libs.index.UInt8Engine'>, <class 'numpy.uint8'>), 'monotonic_incr')
-        741±80μs          288±9μs     0.39  indexing.AssignTimeseriesIndex.time_frame_assign_timeseries_index
-        49.3±2ms       18.2±0.2ms     0.37  indexing.InsertColumns.time_assign_with_setitem
-       54.3±20ms         18.5±1ms     0.34  indexing.InsertColumns.time_assign_list_like_with_setitem

jbrockmendel · 2021-08-28T20:12:51Z

nice cc @phofl

jreback · 2021-08-28T20:14:30Z

can u run the indexing asvs and report here

phofl · 2021-08-28T20:15:37Z

Really nice, good catch.

jreback · 2021-08-28T20:18:07Z

also pls add a whatsnew note for 1.4

mzeitlin11 · 2021-08-28T20:26:35Z

Could also go in 1.3.3, looks like this helps top regression here: https://simonjayhawkins.github.io/fantastic-dollop/#regressions?sort=3&dir=desc&branch=1.3.x

phofl · 2021-08-28T20:28:02Z

Yeah good point, this check was added in 1.3

debnathshoham · 2021-08-28T20:32:55Z

should I put the whatsnew in 1.3.3 fixed regression or under others?
@jreback - I have added the indexing asvs report in the top comment

jreback · 2021-08-28T20:36:32Z

great yeah 1.3.3 is fine (add a perf section)

phofl · 2021-08-28T21:58:21Z

doc/source/whatsnew/v1.3.3.rst

+
+Performance improvements
+~~~~~~~~~~~~~~~~~~~~~~~~
+- Performance improvement for :meth:`DataFrame.__setitem__`


Could you clarify in which cases?

Hi @phofl - updated. Let me know if this is fine.

simonjayhawkins · 2021-08-29T10:02:33Z

Could also go in 1.3.3, looks like this helps top regression here: https://simonjayhawkins.github.io/fantastic-dollop/#regressions?sort=3&dir=desc&branch=1.3.x

side note: class InsertColumns was updated in #42998. Changing the code or setup code invalidates previous results for that benchmark. So that regression is not currently reported.

simonjayhawkins · 2021-08-30T08:41:48Z

side note: class InsertColumns was updated in #42998. Changing the code or setup code invalidates previous results for that benchmark. So that regression is not currently reported.

benchmarks have been re-run and gh-pages updated.

doc/source/whatsnew/v1.3.3.rst

phofl · 2021-08-30T10:29:27Z

doc/source/whatsnew/v1.3.3.rst

+
+Performance improvements
+~~~~~~~~~~~~~~~~~~~~~~~~
+- Performance improvement for :meth:`DataFrame.__setitem__` with dataframes containing unique columns (:issue:`43274`)


This is not specific enough. If key is any list-like or frame we don't hit this at all. Key has to be a scalar I think and value is not a df

Updated. let me know how this looks

phofl · 2021-08-30T12:29:25Z

doc/source/whatsnew/v1.3.3.rst

+
+Performance improvements
+~~~~~~~~~~~~~~~~~~~~~~~~
+- Performance improvement for :meth:`DataFrame.__setitem__` when the key or value is not a dataframe, or key is not list-like (:issue:`43274`)


:class:DataFrame, otherwise lgtm

jreback

lgtm. @phofl merge when ready.

phofl · 2021-08-30T14:36:43Z

doc/source/whatsnew/v1.3.3.rst

+~~~~~~~~~~~~~~~~~~~~~~~~
+- Performance improvement for :meth:`DataFrame.__setitem__` when the key or value is not a :class:`DataFrame`, or key is not list-like (:issue:`43274`)
+-
+-


We need a blank line, docs aren't building like that

ah! I see. I was confused why the docs were consistently failing.

phofl · 2021-08-30T15:48:58Z

thx @debnathshoham

Co-authored-by: Shoham Debnath <debnathshoham@gmail.com>

* PERF: indexing.InsertColumns.time_assign_with_setitem * added whatsnew * more descriptive whatsnew * rectified whatsnew * added PR ref in whatsnew * clear whatsnew * suggested change * added blank line

PERF: indexing.InsertColumns.time_assign_with_setitem

fa9c714

debnathshoham changed the title ~~PERF: indexing.InsertColumns.time_assign_with_setitem~~ PERF: indexing Aug 28, 2021

added whatsnew

9568c73

phofl reviewed Aug 28, 2021

View reviewed changes

debnathshoham added 2 commits August 29, 2021 12:23

more descriptive whatsnew

9cad0b1

rectified whatsnew

1567d3c

simonjayhawkins added this to the 1.3.3 milestone Aug 29, 2021

simonjayhawkins added Indexing Related to indexing on series/frames, not to indexes themselves Performance Memory or execution speed performance labels Aug 29, 2021

simonjayhawkins reviewed Aug 30, 2021

View reviewed changes

doc/source/whatsnew/v1.3.3.rst Outdated Show resolved Hide resolved

added PR ref in whatsnew

256f80f

phofl reviewed Aug 30, 2021

View reviewed changes

clear whatsnew

ed10166

phofl reviewed Aug 30, 2021

View reviewed changes

suggested change

aaa5fdf

jreback approved these changes Aug 30, 2021

View reviewed changes

phofl reviewed Aug 30, 2021

View reviewed changes

added blank line

549b39e

phofl approved these changes Aug 30, 2021

View reviewed changes

phofl merged commit 1009a7d into pandas-dev:master Aug 30, 2021

meeseeksmachine mentioned this pull request Aug 30, 2021

Backport PR #43274 on branch 1.3.x (PERF: indexing) #43303

Merged

meeseeksmachine pushed a commit to meeseeksmachine/pandas that referenced this pull request Aug 30, 2021

Backport PR pandas-dev#43274: PERF: indexing

e91c562

debnathshoham deleted the indexing.InsertColumns.time_assign_with_setitem branch August 30, 2021 15:49

phofl pushed a commit that referenced this pull request Aug 30, 2021

Backport PR #43274: PERF: indexing (#43303)

9fbb61f

Co-authored-by: Shoham Debnath <debnathshoham@gmail.com>

Uh oh!

PERF: indexing #43274

PERF: indexing #43274

Uh oh!

Conversation

debnathshoham commented Aug 28, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

jbrockmendel commented Aug 28, 2021

Uh oh!

jreback commented Aug 28, 2021

Uh oh!

phofl commented Aug 28, 2021

Uh oh!

jreback commented Aug 28, 2021

Uh oh!

mzeitlin11 commented Aug 28, 2021

Uh oh!

phofl commented Aug 28, 2021

Uh oh!

debnathshoham commented Aug 28, 2021

Uh oh!

jreback commented Aug 28, 2021

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

simonjayhawkins commented Aug 29, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

simonjayhawkins commented Aug 30, 2021

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

jreback left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

phofl commented Aug 30, 2021

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants

debnathshoham commented Aug 28, 2021 •

edited

Loading

simonjayhawkins commented Aug 29, 2021 •

edited

Loading