PERF: pd.to_datetime, unit='s' much slower for float64 than for int64 #35027
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
|
As a quick measure of the efficacy of this PR, I ran the checks from #20445 on current branch and master. The set-up is: For ints I get and for floats I get |
WillAyd
reviewed
Jun 29, 2020
WillAyd
added
Performance
|
@WillAyd @jbrockmendel I have something that works on some Linux environments but not on others and not on Mac/Windows. Not sure what the problem(s) are and the best way to go about debugging |
|
ping @WillAyd @jbrockmendel |
WillAyd
reviewed
Jul 24, 2020
pandas/_libs/tslib.pyx
Outdated
| # fill by comparing to NPY_NAT constant | ||
| mask = fresult == NPY_NAT | ||
| fresult[mask] = 0 | ||
| fvalues = fresult * (<float64_t>m) |
There was a problem hiding this comment.
This cast doesn't seem right and is probably the cause of your portability issue. What is the warning that this generates without?
There was a problem hiding this comment.
The traceback if I remove this line:
File "/workspaces/pandas-arw2019/pandas/core/tools/datetimes.py", line 802, in to_datetime
values = convert_listlike(arg._values, format)
File "/workspaces/pandas-arw2019/pandas/core/tools/datetimes.py", line 338, in _convert_listlike_datetimes
result, tz_parsed = tslib.array_with_unit_to_datetime(
File "pandas/_libs/tslib.pyx", line 262, in pandas._libs.tslib.array_with_unit_to_datetime
if ((fvalues < Timestamp.min.value).any()
SystemError: /home/conda/feedstock_root/build_artifacts/python_1591030388223/work/Objects/object.c:769: bad argument to internal function
jreback
requested changes
Jul 24, 2020
|
Benchmark results are below. Would appreciate any comments on how to triage the areas which take a hit |
asv_bench/benchmarks/timeseries.py
Outdated
| ) | ||
| self.timestamp_seconds_float = self.timestamp_seconds_int.astype("float64") | ||
|
|
||
| def to_datetime_int(self): |
There was a problem hiding this comment.
i think these need to be def time_foo
jreback
requested changes
Sep 18, 2020
pandas/_libs/tslib.pyx
Outdated
| result = values.astype('M8[ns]') | ||
| if unit == "ns": | ||
| if issubclass(values.dtype.type, (np.integer, np.float_)): | ||
| result = values.astype("M8[ns]") |
There was a problem hiding this comment.
does astype of floats directly to M8 work?
There was a problem hiding this comment.
i guess it does (as you use it below), but do we have a test specifically for float with unit='ns'?
also can try .astype(..., copy=False)
There was a problem hiding this comment.
does astype of floats directly to M8 work?
would it be better to do, here and below:
ivalues = values.view("i8")
result = ivalues.astype("M8[ns]")There was a problem hiding this comment.
i guess it does (as you use it below), but do we have a test specifically for float with unit='ns'?
I'll look some more but I think we don't. Will add unless I find one
|
@jreback with your suggestion (bb8c35b) benchmarks are now (largely) unaffected: |
jreback
approved these changes
Sep 19, 2020
|
thanks @arw2019 very nice! |
|
thanks @jreback @WillAyd @jbrockmendel for reviewing!!! |
kesmit13
pushed a commit
to kesmit13/pandas
that referenced
this pull request
Nov 2, 2020
3 tasks
3 tasks
3 tasks
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
black pandasgit diff upstream/master -u -- "*.py" | flake8 --diffAs per discussion in #20445 this PR addresses performance of
to_datetimeon a uniform type array of floats. The aim is to get a speed-up by implementingastype-ing, and avoid looping, for floats.