CLN: ASV frame_methods benchmark #18536
Conversation
Codecov Report
@@ Coverage Diff @@
## master #18536 +/- ##
==========================================
- Coverage 91.35% 91.33% -0.02%
==========================================
Files 164 164
Lines 49801 49801
==========================================
- Hits 45494 45485 -9
- Misses 4307 4316 +9
Continue to review full report at Codecov.
|
Codecov Report
@@ Coverage Diff @@
## master #18536 +/- ##
==========================================
- Coverage 91.35% 91.33% -0.02%
==========================================
Files 164 164
Lines 49802 49802
==========================================
- Hits 45496 45487 -9
- Misses 4306 4315 +9
Continue to review full report at Codecov.
|
| goal_time = 0.2 | ||
|
|
||
| def setup(self): | ||
| np.random.seed(1234) |
There was a problem hiding this comment.
You mentioned a module level setup function. Can we use that for the random seed to avoid repeating it for each benchmark?
There was a problem hiding this comment.
In my comment I was actually referring to the setup function of each benchmark class instead of a module level setup function since the asv docs imply that the setup function of each class is run before each time_* function in that class.
Eventually I think we should transition to using 'standardized' data that could be initialized once at the module level.
There was a problem hiding this comment.
But in your comment you included this quote from the asv docs:
You can also include a module-level setup function, which will be run for every benchmark within the module, prior to any setup assigned specifically to each function.
which I understood as a module level def setup() that is called before each benchmark. So I think that should work, but not sure.
Eventually I think we should transition to using 'standardized' data that could be initialized once at the module level.
Yes, I agree we should try to do that more (at least for those cases where the data is not modified)
| for col in df: | ||
| df[col] | ||
| for col in self.df3: | ||
| self.df3[col] | ||
|
|
||
| def time_itertuples(self): | ||
| for row in self.df2.itertuples(): | ||
| pass |
There was a problem hiding this comment.
Now you are at it, can you add a iterrows one as well?
|
|
||
|
|
||
| #----------------------------------------------------------------------------- | ||
| # from_records issue-6700 | ||
| class FromRecords(object): |
There was a problem hiding this comment.
maybe move to the frame constructor benchmarks?
| K = 10 | ||
| self.df = DataFrame({'key1': tm.makeStringIndex(N).values.repeat(K), | ||
| 'key2': tm.makeStringIndex(N).values.repeat(K), | ||
| 'value': np.random.randn(N * K)}) | ||
|
|
||
| def time_frame_sort_index_by_columns(self): | ||
| self.df.sort_index(by=['key1', 'key2']) |
There was a problem hiding this comment.
This is actually deprecated, I think we should do sort_values if that exists, otherwise, sort_index
| self.df.info() | ||
|
|
||
|
|
||
| class Nlargest(object): |
There was a problem hiding this comment.
Maybe combine this one with Quantile? (same setup, and related methods)
Move relevant benchmark to other files Add more cleaning
b0956ae to
12e4686
Compare
|
@jorisvandenbossche Nice catch; you were correct about the setup. I was able to define a Otherwise, moved |
| import pandas.util.testing as tm | ||
| from pandas import (DataFrame, Series, MultiIndex, date_range, period_range, | ||
| isnull, NaT) | ||
| from .pandas_vb_common import setup |
There was a problem hiding this comment.
you can put a # noqa on this line to avoid linter warnings
|
Apart from the small comment, looks good to me! |
|
Since this is not linted anyhow, not that urgent (although nice for local use that it doesn't give a lint error). So you can always fix it if you do another PR on this. Thanks! |
Added
np.random.seed(1234)in setup classes where random data is created xref BENCH: put in np.random.seed on vbenches #8144Ran flake8 and replaced star imports
Moved
GetItemSingleColumn,AssignTimeseriesIndex, andInsertColumnstoindexing.py, andStringSlicetostrings.pyRefactored to use
paramswhere relevant