-
-
Notifications
You must be signed in to change notification settings - Fork 17.9k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Refactor groupby helper from tempita to fused types #24954
Conversation
pandas/_libs/groupby.pyx
Outdated
@@ -382,5 +382,61 @@ def group_any_all(uint8_t[:] out, | |||
out[lab] = flag_val | |||
|
|||
|
|||
ctypedef fused floating: | |||
float32_t |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
we likely have this type elsewhere
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
cython has a builtin floating
fused type for float+double, better to re-use that as long as we're OK with declaring that float==float32_t and double==float64_t
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
As you said, floating is a fused type for float + double.
From what I saw in the cython code here and here , float is np.float32_t and double is np.float64_t.
So the fused type I created, floating, can be replaced with cython's floating.
It is possible that sometime in the future cython will use different definitions for np.float32_t and np.float64_t. If that is a concern, we can stick with defining our own floating fused type.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@jbrockmendel @WillAyd @jreback I made a change that the group_add function is not looked up by type (e.g. grou_add_float64) but rather directly by name.
It seems that after the change, when group_add is used with a ndarray[int64], the ndarray is first cast into ndarray[float64], then passed into group_add[float64], then cast back into ndarray[int64].
This is different then before the change (which passed the tests), because the cast back into ndarray[int64] did not occur so the result was an ndarray[float64].
See here.
What do you think is the best solution? keep the group_add_{type} = group_add_floating statements?.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
[...] ndarray is first cast into ndarray[float64] [...]
This is definitely not desirable. Any guesses why this might be happening?
What do you think is the best solution? keep the group_add_{type} = group_add_floating statements?
The best case would be to just have group_add
and when we call it from outside cython, cython will figure out what its supposed to do (correctly, without doing an unnecessary casting). In my experience, doing this lets us get rid of a bunch of ugly non-cython code.
Next best would be to keep group_add_dtype
functions. In these cases, I wouldn't suggest group_add_{type} = group_add_floating
but more specifically group_add_{type} = group_add_floating[{type}]
. And if feasible, I'd prefer group_add[{type}]
over group_add_floating[{type}]
.
Is this helpful?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@noamher generally we want to try and simplify the calls we are making from the groupby module back to the Cython layer, so I'd be a little hesitant to add more logic therein.
Might be missing the point but can you provide a specific example of the issue you are trying to solve? Certain types simply not resolving to the function correctly?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Just to be clear, the specialization works. I preferred to use the actual types rather than string thouh but couldn't make it work.
I guess the cyhton layer can do the cast into float and back into int.
I'll try to understand what happens to the nans when the output is cast back into floats.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I preferred to use the actual types rather than string thouh but couldn't make it work.
Yah I've found this tricky too. Don't sweat it.
I've looked into keeping the name group_add without the group_add_{type} members (i.e. changing things at the caller side).
Agreed with @WillAyd, definitely don't want to add more complexity on the python side. Removing the group_add_{type}
s would be nice, but if we have to do that in a follow-up that's OK too.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Used cython.floating instead of creating a fused type, is that okay?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@jreback green now.
Codecov Report
@@ Coverage Diff @@
## master #24954 +/- ##
==========================================
+ Coverage 92.37% 92.37% +<.01%
==========================================
Files 166 166
Lines 52408 52408
==========================================
+ Hits 48411 48412 +1
+ Misses 3997 3996 -1
Continue to review full report at Codecov.
|
Codecov Report
@@ Coverage Diff @@
## master #24954 +/- ##
==========================================
+ Coverage 92.38% 92.38% +<.01%
==========================================
Files 166 166
Lines 52398 52398
==========================================
+ Hits 48406 48407 +1
+ Misses 3992 3991 -1
Continue to review full report at Codecov.
|
@@ -380,7 +380,7 @@ def get_func(fname): | |||
# otherwise find dtype-specific version, falling back to object | |||
for dt in [dtype_str, 'object']: | |||
f = getattr(libgroupby, "{fname}_{dtype_str}".format( | |||
fname=fname, dtype_str=dtype_str), None) | |||
fname=fname, dtype_str=dt), None) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
So do we even need this object fallback loop?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
not for group_add and everything worked fine until now.
still it seemed like a bug.
pandas/_libs/groupby.pyx
Outdated
|
||
|
||
group_add_float32 = group_add_floating | ||
group_add_float64 = group_add_floating |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
group_add_float32 = group_add_floating[float32_t]
group_add_float64 = group_add_floating[float64_t]
or even better, track down the places where these are used and call the non-specialized one directly.
Probably also shorten the name to just group_add
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
After shortening the name to group_add, it's not required to define group_add_float32 and group_add_float64. The function is found by its name ("group_add").
fedab7f
to
c69da84
Compare
seems ok, ping when green. |
c69da84
to
a9b361b
Compare
a9b361b
to
937e44c
Compare
out[i, j] = sumx[i, j] | ||
|
||
|
||
group_add_float32 = _group_add['float'] |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Do we still need these? Thought from a previous comment that they weren't required
Side note - I see you've been force pushing changes. If you merge master you can push without forcing, which typically does a better job of maintaining comment history
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Yes, they are still needed (to avoid changing the code calling this function). atleast until some logic to handle int64_t differently is added to the _group_add function.
I force push to keep only 1 commit (avoid the bloat).
I can add commits and squash before merging, do you think that's better?
@@ -382,5 +383,55 @@ def group_any_all(uint8_t[:] out, | |||
out[lab] = flag_val | |||
|
|||
|
|||
@cython.wraparound(False) | |||
@cython.boundscheck(False) | |||
def _group_add(floating[:, :] out, |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Unrelated to this PR but I see there are some functions at the end of the template file which have been converted to fused types but still remain in the template file. Could make sense to move them here as well (separate PR)
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I'll take a look at that (for next PR). thanks.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@WillAyd which functions did you find that are unnecessary? I couldn't find any.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
small comment, ping on green.
@@ -382,5 +383,55 @@ def group_any_all(uint8_t[:] out, | |||
out[lab] = flag_val | |||
|
|||
|
|||
@cython.wraparound(False) | |||
@cython.boundscheck(False) | |||
def _group_add(floating[:, :] out, |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
just call this group_add
I can't. Because then it will be found and used instead of the typed version. |
Any name but that will work. |
not sure I understand we do this for all other functions. |
@jreback my plan is to do this over a few pr to all functions in groupy_helper.pxi.in. I can move some other function in this PR but I think it may be easier to do it in another pr. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I can't. Because then it will be found and used instead of the typed version.
Which will cause problems for a rather odd case where int64 is the type and the function is supposed to not be found at first, then found after the ndarray is cast into float64.not sure I understand we do this for all other functions.
@jreback I already have the next PR ready for the rest of the functions (group_prod, group_var, group_mean, group_ohlc).
I had to do the same thing (e.g. creating the group_prod_float64) for them as well for the same reason.
Description of the issue:
The issue is that the calling mechanism, first looks up by funcname, then by funcname_{{type}}.
The code the calls the calling mechanism has a fallback so that if an group_add is looked up with type int64, then the fallback is to cast into the input ndarray into float64, then lookup group_add with type float64.
So group_add is eventually called with an ndarray of type float64 and returns a result which is an ndarray of float64. Now, before returning the result, the fallback code casts the result back into int64. Then returns the result.
Implications:
Creating a group_add function that will be looked-up directly (not using type) means that when the lookup type is int64, the method will still be found (without the type in the name anyway). Calling it will succeed because of an implicit cast from int64 to float64.
Because the fallback code was never needed, there will be no cast back into int64.
This will fail tests and change the behavior of whatever's calling it in the first place.
To avoid this implications, group_add_{{type}} are defined the the function group_add is renamed into something that will fail the first lookup (e.g. _group_add).
I know that's not the prettiest solution but changing the calling mechanism is something I really tried to avoid because it seems a bit sensitive code to change without breaking anything.
@@ -382,5 +383,55 @@ def group_any_all(uint8_t[:] out, | |||
out[lab] = flag_val | |||
|
|||
|
|||
@cython.wraparound(False) | |||
@cython.boundscheck(False) | |||
def _group_add(floating[:, :] out, |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I'll take a look at that (for next PR). thanks.
@@ -382,5 +383,55 @@ def group_any_all(uint8_t[:] out, | |||
out[lab] = flag_val | |||
|
|||
|
|||
@cython.wraparound(False) | |||
@cython.boundscheck(False) | |||
def _group_add(floating[:, :] out, |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@WillAyd which functions did you find that are unnecessary? I couldn't find any.
out[i, j] = sumx[i, j] | ||
|
||
|
||
group_add_float32 = _group_add['float'] |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Yes, they are still needed (to avoid changing the code calling this function). atleast until some logic to handle int64_t differently is added to the _group_add function.
I force push to keep only 1 commit (avoid the bloat).
I can add commits and squash before merging, do you think that's better?
@noamher my concern here is the naming consistency. IIRC @jbrockmendel did things like this a while back. Is this pattern consistent with other fused types are callled in groupby & algos? |
@jreback Currently (as if this PR), only group_add exist, because the other functions (e.g. group_mean) only exist in their specialized form (e.g. group_mean_float64) which is created by tempita. What other functions should group_add be consistent with? |
@noamher ok thanks. if you have followups to keep consistentcy would be good. |
* upstream/master: BUG: Fix exceptions when Series.interpolate's `order` parameter is missing or invalid (pandas-dev#25246) API: Ensure DatetimeTZDtype standardizes pytz timezones (pandas-dev#25254) Split Excel IO Into Sub-Directory (pandas-dev#25153) PR04 errors fix (pandas-dev#25157) DEPR: remove assert_panel_equal (pandas-dev#25238) BUG: pandas Timestamp tz_localize and tz_convert do not preserve `freq` attribute (pandas-dev#25247) Revert "BLD: prevent asv from calling sys.stdin.close() by using different launch method (pandas-dev#25237)" (pandas-dev#25253) REF/TST: resample/test_base.py (pandas-dev#25262) BUG: Duplicated returns boolean dataframe (pandas-dev#25234) CLN: Remove ipython 2.x compat (pandas-dev#25150) Refactor groupby group_add from tempita to fused types (pandas-dev#24954) CLN: For loops, boolean conditions, misc. (pandas-dev#25206) (Closes pandas-dev#25029) Removed extra bracket from cheatsheet code example. (pandas-dev#25032) BLD: prevent asv from calling sys.stdin.close() by using different launch method (pandas-dev#25237) BUG: Fix read_json orient='table' without index (pandas-dev#25170) (pandas-dev#25171) BUG: Fix regression in DataFrame.apply causing RecursionError (pandas-dev#25230) BUG-25061 fix printing indices with NaNs (pandas-dev#25202) DEPR: Add Deprecated warning for timedelta with passed units M and Y (pandas-dev#23264) DEPR: Remove Panel-specific parts of io.pytables (pandas-dev#25233) DEPR: remove tm.makePanel and all usages (pandas-dev#25231)
* ERR/TST: Add pytest idiom to dtypes/test_cast.py (pandas-dev#24847) * fix MacPython pandas-wheels failue (pandas-dev#24851) * DEPS: Bump pyarrow min version to 0.9.0 (pandas-dev#24854) Closes pandas-devgh-24767 * DOC: Document AttributeError for accessor (pandas-dev#24855) Closes pandas-dev#20579 * Start whatsnew for 0.24.1 and 0.25.0 (pandas-dev#24848) * DEPR/API: Non-ns precision in Index constructors (pandas-dev#24806) * BUG: Format mismatch doesn't coerce to NaT (pandas-dev#24815) * BUG: Properly parse unicode usecols names in CSV (pandas-dev#24856) * CLN: fix typo in asv eval.Query suite (pandas-dev#24865) * BUG: DataFrame respects dtype with masked recarray (pandas-dev#24874) * REF/CLN: Move private method (pandas-dev#24875) * BUG : ValueError in case on NaN value in groupby columns (pandas-dev#24850) * BUG: fix floating precision formatting in presence of inf (pandas-dev#24863) * DOC: Creating top-level user guide section, and moving pages inside (pandas-dev#24677) * DOC: Creating top-level development section, and moving pages inside (pandas-dev#24691) * DOC: Creating top-level getting started section, and moving pages inside (pandas-dev#24678) * DOC: Implementing redirect system, and adding user_guide redirects (pandas-dev#24715) * DOC: Implementing redirect system, and adding user_guide redirects * Using relative urls for the redirect * Validating that no file is overwritten by a redirect * Adding redirects for getting started and development sections * DOC: fixups (pandas-dev#24888) * Fixed heading on whatnew * Remove empty scalars.rst * CLN: fix typo in ctors.SeriesDtypesConstructors setup (pandas-dev#24894) * DOC: No clean in sphinx_build (pandas-dev#24902) Closes pandas-dev#24727 * BUG (output formatting): use fixed with for truncation column instead of inferring from last column (pandas-dev#24905) * DOC: also redirect old whatsnew url (pandas-dev#24906) * Revert BUG-24212 fix usage of Index.take in pd.merge (pandas-dev#24904) * Revert BUG-24212 fix usage of Index.take in pd.merge xref pandas-dev#24733 xref pandas-dev#24897 * test 0.23.4 output * added note about buggy test * DOC: Add experimental note to DatetimeArray and TimedeltaArray (pandas-dev#24882) * DOC: Add experimental note to DatetimeArray and TimedeltaArray * Disable M8 in nanops (pandas-dev#24907) * Disable M8 in nanops Closes pandas-dev#24752 * CLN: fix typo in asv benchmark of non_unique_sorted, which was not sorted (pandas-dev#24917) * API/VIS: remove misc plotting methods from plot accessor (revert pandas-dev#23811) (pandas-dev#24912) * DOC: some 0.24.0 whatsnew clean-up (pandas-dev#24911) * DOC: Final reorganization of documentation pages (pandas-dev#24890) * DOC: Final reorganization of documentation pages * Move ecosystem to top level * DOC: Adding redirects to API moved pages (pandas-dev#24909) * DOC: Adding redirects to API moved pages * DOC: Making home page links more compact and clearer (pandas-dev#24928) * DOC: 0.24 release date (pandas-dev#24930) * DOC: Adding version to the whatsnew section in the home page (pandas-dev#24929) * API: Remove IntervalArray from top-level (pandas-dev#24926) * RLS: 0.24.0 * DEV: Start 0.25 cycle * DOC: State that we support scalars in to_numeric (pandas-dev#24944) We support it and test it already. xref pandas-devgh-24910. * DOC: Minor what's new fix (pandas-dev#24933) * TST: GH#23922 Add missing match params to pytest.raises (pandas-dev#24937) * Add tests for NaT when performing dt.to_period (pandas-dev#24921) * DOC: switch headline whatsnew to 0.25 (pandas-dev#24941) * BUG-24212 fix regression in pandas-dev#24897 (pandas-dev#24916) * CLN: reduce overhead in setup for categoricals benchmarks in asv (pandas-dev#24913) * Excel Reader Refactor - Base Class Introduction (pandas-dev#24829) * TST/REF: Add pytest idiom to test_numeric.py (pandas-dev#24946) * BLD: silence npy_no_deprecated warnings with numpy>=1.16.0 (pandas-dev#24864) * CLN: Refactor cython to use memory views (pandas-dev#24932) * DOC: Clean sort_values and sort_index docstrings (pandas-dev#24843) * STY: use pytest.raises context syntax (indexing) (pandas-dev#24960) * Fixed itertuples usage in to_dict (pandas-dev#24965) * Fixed itertuples usage in to_dict Closes pandas-dev#24940 Closes pandas-dev#24939 * STY: use pytest.raises context manager (resample) (pandas-dev#24977) * DOC: Document breaking change to read_csv (pandas-dev#24989) * DEPR: Fixed warning for implicit registration (pandas-dev#24964) * STY: use pytest.raises context manager (indexes/datetimes) (pandas-dev#24995) * DOC: move whatsnew note of pandas-dev#24916 (pandas-dev#24999) * BUG: Fix broken links (pandas-dev#25002) The previous location of contributing.rst file was /doc/source/contributing.rst but has been moved to /doc/source/development/contributing.rst * fix for BUG: grouping with tz-aware: Values falls after last bin (pandas-dev#24973) * REGR: Preserve order by default in Index.difference (pandas-dev#24967) Closes pandas-dev#24959 * CLN: do not use .repeat asv setting for storing benchmark data (pandas-dev#25015) * CLN: isort asv_bench/benchmark/algorithms.py (pandas-dev#24958) * fix+test to_timedelta('NaT', box=False) (pandas-dev#24961) * PERF: significant speedup in sparse init and ops by using numpy in check_integrity (pandas-dev#24985) * BUG: Fixed merging on tz-aware (pandas-dev#25033) * Test nested PandasArray (pandas-dev#24993) * DOC: fix error in documentation pandas-dev#24981 (pandas-dev#25038) * BUG: support dtypes in column_dtypes for to_records() (pandas-dev#24895) * Makes example from docstring work (pandas-dev#25035) * CLN: typo fixups (pandas-dev#25028) * BUG: to_datetime(strs, utc=True) used previous UTC offset (pandas-dev#25020) * BUG: Better handle larger numbers in to_numeric (pandas-dev#24956) * BUG: Better handle larger numbers in to_numeric * Warn about lossiness when passing really large numbers that exceed (u)int64 ranges. * Coerce negative numbers to float when requested instead of crashing and returning object. * Consistently parse numbers as integers / floats, even if we know that the resulting container has to be float. This is to ensure consistent error behavior when inputs numbers are too large. Closes pandas-devgh-24910. * MAINT: Address comments * BUG: avoid usage in_qtconsole for recent IPython versions (pandas-dev#25039) * Drop IPython<4.0 compat * Revert "Drop IPython<4.0 compat" This reverts commit 0cb0452. * update a * whatsnew * REGR: fix read_sql delegation for queries on MySQL/pymysql (pandas-dev#25024) * DOC: Start 0.24.2.rst (pandas-dev#25026) [ci skip] * REGR: rename_axis with None should remove axis name (pandas-dev#25069) * clarified the documentation for DF.drop_duplicates (pandas-dev#25056) * Clarification in docstring of Series.value_counts (pandas-dev#25062) * ENH: Support fold argument in Timestamp.replace (pandas-dev#25046) * CLN: to_pickle internals (pandas-dev#25044) * Implement+Test Tick.__rtruediv__ (pandas-dev#24832) * API: change Index set ops sort=True -> sort=None (pandas-dev#25063) * BUG: to_clipboard text truncated for Python 3 on Windows for UTF-16 text (pandas-dev#25040) * PERF: use new to_records() argument in to_stata() (pandas-dev#25045) * DOC: Cleanup 0.24.1 whatsnew (pandas-dev#25084) * Fix quotes position in pandas.core, typos and misspelled parameters. (pandas-dev#25093) * CLN: Remove sentinel_factory() in favor of object() (pandas-dev#25074) * TST: remove DST transition scenarios from tc pandas-dev#24689 (pandas-dev#24736) * BLD: remove spellcheck from Makefile (pandas-dev#25111) * DOC: small clean-up of 0.24.1 whatsnew (pandas-dev#25096) * DOC: small doc fix to Series.repeat (pandas-dev#25115) * TST: tests for categorical apply (pandas-dev#25095) * CLN: use dtype in constructor (pandas-dev#25098) * DOC: frame.py doctest fixing (pandas-dev#25097) * DOC: 0.24.1 release (pandas-dev#25125) [ci skip] * Revert set_index inspection/error handling for 0.24.1 (pandas-dev#25085) * DOC: Minor what's new fix (pandas-dev#24933) * Backport PR pandas-dev#24916: BUG-24212 fix regression in pandas-dev#24897 (pandas-dev#24951) * Revert "Backport PR pandas-dev#24916: BUG-24212 fix regression in pandas-dev#24897 (pandas-dev#24951)" This reverts commit 84056c5. * DOC/CLN: Timezone section in timeseries.rst (pandas-dev#24825) * DOC: Improve timezone documentation in timeseries.rst * edit some of the examples * Address review * DOC: Fix validation type error RT04 (pandas-dev#25107) (pandas-dev#25129) * Reading a HDF5 created in py2 (pandas-dev#25058) * BUG: Fixing regression in DataFrame.all and DataFrame.any with bool_only=True (pandas-dev#25102) * Removal of return variable names (pandas-dev#25123) * DOC: Improve docstring of Series.mul (pandas-dev#25136) * TST/REF: collect DataFrame reduction tests (pandas-dev#24914) * Fix validation error type `SS05` and check in CI (pandas-dev#25133) * Fixed tuple to List Conversion in Dataframe class (pandas-dev#25089) * STY: use pytest.raises context manager (indexes/multi) (pandas-dev#25175) * DOC: Updates to Timestamp document (pandas-dev#25163) * BLD: pin cython language level to '2' (pandas-dev#25145) Not explicitly pinning the language level has been producing future warnings from cython. The next release of cython is going to change the default level to '3str' under which the pandas cython extensions do not compile. The long term solution is to update the cython files to the next language level, but this is a stop-gap to keep pandas building. * CLN: Use ABCs in set_index (pandas-dev#25128) * DOC: update docstring for series.nunique (pandas-dev#25116) * DEPR: remove PanelGroupBy, disable DataFrame.to_panel (pandas-dev#25047) * BUG: DataFrame.merge(suffixes=) does not respect None (pandas-dev#24819) * fix MacPython pandas-wheels failure (pandas-dev#25186) * modernize compat imports (pandas-dev#25192) * TST: follow-up to Test nested pandas array pandas-dev#24993 (pandas-dev#25155) * revert changes to tests in pandas-devgh-24993 * Test nested PandasArray * isort test_numpy.py * change NP_VERSION_INFO * use LooseVersion * add _np_version_under1p16 * remove blank line from merge master * add doctstrings to fixtures * DOC/CLN: Fix errors in Series docstrings (pandas-dev#24945) * REF: Add more pytest idiom to test_holiday.py (pandas-dev#25204) * DOC: Fix validation type error SA05 (pandas-dev#25208) Create check for SA05 errors in CI * BUG: Fix Series.is_unique with single occurrence of NaN (pandas-dev#25182) * REF: Remove many Panel tests (pandas-dev#25191) * DOC: Fixes to docstrings and add PR10 (space before colon) to validation (pandas-dev#25109) * DOC: exclude autogenerated c/cpp/html files from 'trailing whitespace' checks (pandas-dev#24549) * STY: use pytest.raises context manager (indexes/period) (pandas-dev#25199) * fix ci failures (pandas-dev#25225) * DEPR: remove tm.makePanel and all usages (pandas-dev#25231) * DEPR: Remove Panel-specific parts of io.pytables (pandas-dev#25233) * DEPR: Add Deprecated warning for timedelta with passed units M and Y (pandas-dev#23264) * BUG-25061 fix printing indices with NaNs (pandas-dev#25202) * BUG: Fix regression in DataFrame.apply causing RecursionError (pandas-dev#25230) * BUG: Fix regression in DataFrame.apply causing RecursionError * Add feedback from PR * Add feedback after further code review * Add feedback after further code review 2 * BUG: Fix read_json orient='table' without index (pandas-dev#25170) (pandas-dev#25171) * BLD: prevent asv from calling sys.stdin.close() by using different launch method (pandas-dev#25237) * (Closes pandas-dev#25029) Removed extra bracket from cheatsheet code example. (pandas-dev#25032) * CLN: For loops, boolean conditions, misc. (pandas-dev#25206) * Refactor groupby group_add from tempita to fused types (pandas-dev#24954) * CLN: Remove ipython 2.x compat (pandas-dev#25150) * CLN: Remove ipython 2.x compat * trivial change to trigger asv * Update v0.25.0.rst * revert whatsnew * BUG: Duplicated returns boolean dataframe (pandas-dev#25234) * REF/TST: resample/test_base.py (pandas-dev#25262) * Revert "BLD: prevent asv from calling sys.stdin.close() by using different launch method (pandas-dev#25237)" (pandas-dev#25253) This reverts commit f67b7fd. * BUG: pandas Timestamp tz_localize and tz_convert do not preserve `freq` attribute (pandas-dev#25247) * DEPR: remove assert_panel_equal (pandas-dev#25238) * PR04 errors fix (pandas-dev#25157) * Split Excel IO Into Sub-Directory (pandas-dev#25153) * API: Ensure DatetimeTZDtype standardizes pytz timezones (pandas-dev#25254) * API: Ensure DatetimeTZDtype standardizes pytz timezones * Add whatsnew * BUG: Fix exceptions when Series.interpolate's `order` parameter is missing or invalid (pandas-dev#25246) * BUG: raise accurate exception from Series.interpolate (pandas-dev#24014) * Actually validate `order` before use in spline * Remove unnecessary check and dead code * Clean up comparison/tests based on feedback * Include invalid order value in exception * Check for NaN order in spline validation * Add whatsnew entry for bug fix * CLN: Make unit tests assert one error at a time * CLN: break test into distinct test case * PEP8 fix in test module * CLN: Test fixture for interpolate methods * BUG: DataFrame.join on tz-aware DatetimeIndex (pandas-dev#25260) * REF: use _constructor and ABCFoo to avoid runtime imports (pandas-dev#25272) * Refactor groupby group_prod, group_var, group_mean, group_ohlc (pandas-dev#25249) * Fix typo in Cheat sheet with regex (pandas-dev#25215) * Edit parameter type in pandas.core.frame.py DataFrame.count (pandas-dev#25198) * TST/CLN: remove test_slice_ints_with_floats_raises (pandas-dev#25277) * Removed Panel class from HDF ASVs (pandas-dev#25281) * DOC: Fix minor typo in docstring (pandas-dev#25285) * DOC/CLN: Fix errors in DataFrame docstrings (pandas-dev#24952) * Skipped broken Py2 / Windows test (pandas-dev#25323) * Rt05 documentation error fix issue 25108 (pandas-dev#25309) * Fix typos in docs (pandas-dev#25305) * Doc: corrects spelling in generic.py (pandas-dev#25333) * BUG: groupby.transform retains timezone information (pandas-dev#25264) * Fixes Formatting Exception (pandas-dev#25088) * Bug: OverflowError in resample.agg with tz data (pandas-dev#25297) * DOC/CLN: Fix various docstring errors (pandas-dev#25295) * COMPAT: alias .to_numpy() for timestamp and timedelta scalars (pandas-dev#25142) * ENH: Support times with timezones in at_time (pandas-dev#25280) * BUG: Fix passing of numeric_only argument for categorical reduce (pandas-dev#25304) * TST: use a fixed seed to have the same uniques across python versions (pandas-dev#25346) TST: add pytest-mock to handle mocker fixture * TST: xfail excel styler tests, xref GH25351 (pandas-dev#25352) * TST: xfail excel styler tests, xref GH25351 * CI: cleanup .c files for cpplint>1.4 * DOC: Correct doc mistake in combiner func (pandas-dev#25360) Closes pandas-devgh-25359. * DOC/BLD: fix --no-api option (pandas-dev#25209) * DOC: modify typos in Contributing section (pandas-dev#25365) * Remove spurious MultiIndex creation in `_set_axis_name` (pandas-dev#25371) * Resovles pandas-dev#25370 * Introduced by pandas-dev#22969 * pandas-dev#23049: test for Fatal Stack Overflow stemming From Misuse of astype('category') (pandas-dev#25366) * 9236: test for the DataFrame.groupby with MultiIndex having pd.NaT (pandas-dev#25310) * [BUG] exception handling of MultiIndex.__contains__ too narrow (pandas-dev#25268) * 14873: test for groupby.agg coercing booleans (pandas-dev#25327) * BUG/ENH: Timestamp.strptime (pandas-dev#25124) * BUG: constructor Timestamp.strptime() does not support %z. * Add doc string to NaT and Timestamp * updated the error message * Updated whatsnew entry. * Interval dtype fix (pandas-dev#25338) * [CLN] Excel Module Cleanups (pandas-dev#25275) Closes pandas-devgh-25153 Authored-By: tdamsma <tdamsma@gmail.com> * ENH: indexing and __getitem__ of dataframe and series accept zerodim integer np.array as int (pandas-dev#24924) * REGR: fix TimedeltaIndex sum and datetime subtraction with NaT (pandas-dev#25282, pandas-dev#25317) (pandas-dev#25329) * edited whatsnew typo (pandas-dev#25381) * fix typo of see also in DataFrame stat funcs (pandas-dev#25388) * API: more consistent error message for MultiIndex.from_arrays (pandas-dev#25189) * CLN: (re-)enable infer_dtype to catch complex (pandas-dev#25382) * DOC: Edited docstring of Interval (pandas-dev#25410) The docstring contained a repeated segment, which I removed. * Mark test_pct_max_many_rows as high memory (pandas-dev#25400) Fixes issue pandas-dev#25384 * Correct a typo of version number for interpolate() (pandas-dev#25418) * DEP: add pytest-mock to environment.yml (pandas-dev#25417) * BUG: Fix type coercion in read_json orient='table' (pandas-dev#21345) (pandas-dev#25219) * ERR: doc update for ParsingError (pandas-dev#25414) Closes pandas-devgh-22881 * ENH: Add in sort keyword to DatetimeIndex.union (pandas-dev#25110) * DOC: Rewriting of ParserError doc + minor spacing (pandas-dev#25421) Follow-up to pandas-devgh-25414. * API/ERR: allow iterators in df.set_index & improve errors (pandas-dev#24984) * BUG: Indexing with UTC offset string no longer ignored (pandas-dev#25263) * PERF/REF: improve performance of Series.searchsorted, PandasArray.searchsorted, collect functionality (pandas-dev#22034) * TST: remove never-used singleton fixtures (pandas-dev#24885) * BUG: fixed merging with empty frame containing an Int64 column (pandas-dev#25183) (pandas-dev#25289) * DOC: fixed geo accessor example in extending.rst (pandas-dev#25420) I realised "lon" and "lat" had just been switched with "longitude" and "latitude" in the following code block. So I used those names here as well. * TST: numpy RuntimeWarning with Series.round() (pandas-dev#25432) * CI: add __init__.py to isort skip list (pandas-dev#25455) * DOC: CategoricalIndex doc string (pandas-dev#24852) * DataFrame.drop Raises KeyError definition (pandas-dev#25474) * BUG: Keep column level name in resample nunique (pandas-dev#25469) Closes pandas-devgh-23222 xref pandas-devgh-23645 * ERR: Correct error message in to_datetime (pandas-dev#25467) * ERR: Correct error message in to_datetime Closes pandas-devgh-23830 xref pandas-devgh-23969 * Fix minor typo (pandas-dev#25458) Signed-off-by: Philippe Ombredanne <pombredanne@nexb.com> * CI: Set pytest minversion to 4.0.2 (pandas-dev#25402) * CI: Set pytest minversion to 4.0.2 * STY: use pytest.raises context manager (indexes) (pandas-dev#25447) * STY: use pytest.raises context manager (tests/test_*) (pandas-dev#25452) * STY: use pytest.raises context manager (tests/test_*) * fix ci failures * skip py2 ci failure * Fix minor error in dynamic load function (pandas-dev#25256) * Cythonized GroupBy Quantile (pandas-dev#20405) * BUG: Fix regression on DataFrame.replace for regex (pandas-dev#25266) * BUG: Fix regression on DataFrame.replace for regex The commit ensures that the replacement for regex is not confined to the beginning of the string but spans all the characters within. The behaviour is then consistent with versions prior to 0.24.0. One test has been added to account for character replacement when the character is not at the beginning of the string. * Correct contribution guide docbuild instruction (pandas-dev#25479) * TST/REF: Add pytest idiom to test_frequencies.py (pandas-dev#25430) * BUG: Fix index type casting in read_json with orient='table' and float index (pandas-dev#25433) (pandas-dev#25434) * BUG: Groupby.agg with reduction function with tz aware data (pandas-dev#25308) * BUG: Groupby.agg cannot reduce with tz aware data * Handle output always as UTC * Add whatsnew * isort and add another fixed groupby.first/last issue * bring condition at a higher level * Add try for _try_cast * Add comments * Don't pass the utc_dtype explicitly * Remove unused import * Use string dtype instead * DOC: Fix docstring for read_sql_table (pandas-dev#25465) * ENH: Add Series.str.casefold (pandas-dev#25419) * Fix PR10 error and Clean up docstrings from functions related to RT05 errors (pandas-dev#25132) * Fix unreliable test (pandas-dev#25496) * DOC: Clarifying doc/make.py --single parameter (pandas-dev#25482) * fix MacPython / pandas-wheels ci failures (pandas-dev#25505) * DOC: Reword Series.interpolate docstring for clarity (pandas-dev#25491) * Changed insertion order to sys.path (pandas-dev#25486) * TST: xfail non-writeable pytables tests with numpy 1.16x (pandas-dev#25517) * STY: use pytest.raises context manager (arithmetic, arrays, computati… (pandas-dev#25504) * BUG: Fix RecursionError during IntervalTree construction (pandas-dev#25498) * STY: use pytest.raises context manager (plotting, reductions, scalar...) (pandas-dev#25483) * STY: use pytest.raises context manager (plotting, reductions, scalar...) * revert removed testing in test_timedelta.py * remove TODO from test_frame.py * skip py2 ci failure * BUG: Fix potential segfault after pd.Categorical(pd.Series(...), categories=...) (pandas-dev#25368) * Make DataFrame.to_html output full content (pandas-dev#24841) * BUG-16807-1 SparseFrame fills with default_fill_value if data is None (pandas-dev#24842) Closes pandas-devgh-16807. * DOC: Add conda uninstall pandas to contributing guide (pandas-dev#25490) * fix pandas-dev#25487 add modify documentation * fix segfault when running with cython coverage enabled, xref cython#2879 (pandas-dev#25529) * TST: inline empty_frame = DataFrame({}) fixture (pandas-dev#24886) * DOC: Polishing typos out of doc/source/user_guide/indexing.rst (pandas-dev#25528) * STY: use pytest.raises context manager (frame) (pandas-dev#25516) * DOC: Fix pandas-dev#24268 by updating description for keep in Series.nlargest (pandas-dev#25358) * DOC: Fix pandas-dev#24268 by updating description for keep * fix MacPython / pandas-wheels ci failures (pandas-dev#25537) * TST/CLN: Remove more Panel tests (pandas-dev#25550) * BUG: caught typeError in series.at (pandas-dev#25506) (pandas-dev#25533) * ENH: Add errors parameter to DataFrame.rename (pandas-dev#25535) * ENH: GH13473 Add errors parameter to DataFrame.rename * TST: Skip IntervalTree construction overflow test on 32bit (pandas-dev#25558) * DOC: Small fixes to 0.24.2 whatsnew (pandas-dev#25559) * minor typo error (pandas-dev#25574) * BUG: in error message raised when invalid axis parameter (pandas-dev#25553) * BLD: Fixed pip install with no numpy (pandas-dev#25568) * Document the behavior of `axis=None` with `style.background_gradient` (pandas-dev#25551) * fix minor typos in dsintro.rst (pandas-dev#25579) * BUG: Handle readonly arrays in period_array (pandas-dev#25556) * BUG: Handle readonly arrays in period_array Closes pandas-dev#25403 * DOC: Fix typo in tz_localize (pandas-dev#25598) * BUG: secondary y axis could not be set to log scale (pandas-dev#25545) (pandas-dev#25586) * TST: add test for groupby on list of empty list (pandas-dev#25589) * TYPING: Small fixes to make stubgen happy (pandas-dev#25576) * CLN: Parmeterize test cases (pandas-dev#25355)
Refactoring groupby_helper to use fused types instead of tempita for functions such as group_add.
@jbrockmendel if this is what you meant in the previous PR I'll do the refactoring for the rest of the functions in groupby_helper.pxi.in.