Replace Koalas with pandas API on Spark #1949
Conversation
| def replace_nan_with_flag(pdf, flag=-1): | ||
| def replace_nan_with_flag(pdf, flag=-1.): |
There was a problem hiding this comment.
Spark series doesn't support an array that contains floats and integers
ps.from_pandas(pd.Series([[0.0, 0.0], [7.0, 3.0], [14.0, 6.0], [-1, -1], [-1, -1]]))TypeError: element in array field 0: DoubleType can not accept object -1 in type <class 'int'>
featuretools/entityset/entityset.py
Outdated
| if isinstance(df, dd.DataFrame): | ||
| if isinstance(df, (dd.DataFrame, ps.DataFrame)): | ||
| df[index] = 1 | ||
| df[index] = df[index].cumsum() - 1 |
There was a problem hiding this comment.
Spark doesn't support range type for column assignment
setup.cfg
Outdated
| woodwork >= 0.8.1 | ||
| woodwork @ git+https://github.com/alteryx/woodwork.git@migrate-to-pyspark-api#egg=woodwork |
There was a problem hiding this comment.
We'll need to change this to the corresponding woodwork version before merging
|
we'll need to switch the required unit tests from koalas to spark |
Codecov Report
@@ Coverage Diff @@
## main #1949 +/- ##
==========================================
- Coverage 98.99% 98.99% -0.01%
==========================================
Files 146 146
Lines 16478 16437 -41
==========================================
- Hits 16313 16271 -42
- Misses 165 166 +1
Continue to review full report at Codecov.
|
|
I think we'll need to merge and release the changes in Woodwork first before doing the same in featuretools. |
|
@jeff-hernandez Just a quick heads-up. The |
Closes #1864