itholic commented on a change in pull request #33757:
URL: https://github.com/apache/spark/pull/33757#discussion_r690854299
##########
File path: python/pyspark/pandas/tests/test_categorical.py
##########
@@ -243,18 +243,16 @@ def test_astype(self):
self.assert_eq(kcser.astype("category"), pcser.astype("category"))
+ # CategoricalDtype is not updated if the dtype is same from pandas 1.3.
if LooseVersion(pd.__version__) >= LooseVersion("1.3"):
- # TODO(SPARK-36367): Fix the behavior to follow pandas >= 1.3
- pass
- elif LooseVersion(pd.__version__) >= LooseVersion("1.2"):
self.assert_eq(
kcser.astype(CategoricalDtype(["b", "c", "a"])),
pcser.astype(CategoricalDtype(["b", "c", "a"])),
)
else:
self.assert_eq(
kcser.astype(CategoricalDtype(["b", "c", "a"])),
Review comment:
Sure!
##########
File path: python/pyspark/pandas/tests/test_categorical.py
##########
@@ -243,18 +243,16 @@ def test_astype(self):
self.assert_eq(kcser.astype("category"), pcser.astype("category"))
+ # CategoricalDtype is not updated if the dtype is same from pandas 1.3.
if LooseVersion(pd.__version__) >= LooseVersion("1.3"):
- # TODO(SPARK-36367): Fix the behavior to follow pandas >= 1.3
- pass
- elif LooseVersion(pd.__version__) >= LooseVersion("1.2"):
self.assert_eq(
kcser.astype(CategoricalDtype(["b", "c", "a"])),
pcser.astype(CategoricalDtype(["b", "c", "a"])),
)
else:
self.assert_eq(
kcser.astype(CategoricalDtype(["b", "c", "a"])),
- pser.astype(CategoricalDtype(["b", "c", "a"])),
+ kcser,
Review comment:
Yes, because `kcser` is pandas-on-Spark Series, which always follows
latest pandas(1.3).
##########
File path: python/pyspark/pandas/tests/test_categorical.py
##########
@@ -243,18 +243,16 @@ def test_astype(self):
self.assert_eq(kcser.astype("category"), pcser.astype("category"))
+ # CategoricalDtype is not updated if the dtype is same from pandas 1.3.
if LooseVersion(pd.__version__) >= LooseVersion("1.3"):
- # TODO(SPARK-36367): Fix the behavior to follow pandas >= 1.3
- pass
- elif LooseVersion(pd.__version__) >= LooseVersion("1.2"):
self.assert_eq(
kcser.astype(CategoricalDtype(["b", "c", "a"])),
pcser.astype(CategoricalDtype(["b", "c", "a"])),
)
else:
self.assert_eq(
kcser.astype(CategoricalDtype(["b", "c", "a"])),
- pser.astype(CategoricalDtype(["b", "c", "a"])),
+ kcser,
Review comment:
Yes, because `kcser` is pandas-on-Spark Series, which always follows
latest pandas(1.3).
Latest pandas doesn't update the `dtype` after `astype`, if the unique
values of target `dtype` are the same as original unique values.
##########
File path: python/pyspark/pandas/tests/test_categorical.py
##########
@@ -243,18 +243,16 @@ def test_astype(self):
self.assert_eq(kcser.astype("category"), pcser.astype("category"))
+ # CategoricalDtype is not updated if the dtype is same from pandas 1.3.
if LooseVersion(pd.__version__) >= LooseVersion("1.3"):
- # TODO(SPARK-36367): Fix the behavior to follow pandas >= 1.3
- pass
- elif LooseVersion(pd.__version__) >= LooseVersion("1.2"):
self.assert_eq(
kcser.astype(CategoricalDtype(["b", "c", "a"])),
pcser.astype(CategoricalDtype(["b", "c", "a"])),
)
else:
self.assert_eq(
kcser.astype(CategoricalDtype(["b", "c", "a"])),
- pser.astype(CategoricalDtype(["b", "c", "a"])),
+ kcser,
Review comment:
Yes, because `kcser` is pandas-on-Spark Series, which always follows
latest pandas(1.3).
Latest pandas doesn't update the `dtype` after `astype`, if the unique
values of target `dtype` are the same as original unique values as described in
the PR description.
```python
>>> pcat1 = pser.astype(CategoricalDtype(["b", "c", "a"]))
>>> pcat2 = pser.astype(CategoricalDtype(["a", "b", "c"]))
>>> pcat1.dtype == pcat2.dtype
True
```
##########
File path: python/pyspark/pandas/tests/data_type_ops/test_categorical_ops.py
##########
@@ -183,22 +183,31 @@ def test_astype(self):
psser = ps.from_pandas(pser)
self.assert_eq(pser.astype(int), psser.astype(int))
self.assert_eq(pser.astype(float), psser.astype(float))
- self.assert_eq(pser.astype(np.float32), psser.astype(np.float32))
- self.assert_eq(pser.astype(np.int32), psser.astype(np.int32))
- self.assert_eq(pser.astype(np.int16), psser.astype(np.int16))
- self.assert_eq(pser.astype(np.int8), psser.astype(np.int8))
self.assert_eq(pser.astype(str), psser.astype(str))
- self.assert_eq(pser.astype(bool), psser.astype(bool))
self.assert_eq(pser.astype("category"), psser.astype("category"))
+ # Bug in pandas 1.2
+ if (LooseVersion(pd.__version__) >= LooseVersion("1.2")) and (
+ LooseVersion(pd.__version__) < LooseVersion("1.3")
+ ):
+ self.assert_eq(psser.astype(np.float32), ps.Series(data,
dtype=np.float32))
+ self.assert_eq(psser.astype(np.int32), ps.Series(data,
dtype=np.int32))
+ self.assert_eq(psser.astype(np.int16), ps.Series(data,
dtype=np.int16))
+ self.assert_eq(psser.astype(np.int8), ps.Series(data,
dtype=np.int8))
+ self.assert_eq(psser.astype(bool), ps.Series(data, dtype=bool))
+ else:
+ self.assert_eq(pser.astype(np.float32), psser.astype(np.float32))
+ self.assert_eq(pser.astype(np.int32), psser.astype(np.int32))
+ self.assert_eq(pser.astype(np.int16), psser.astype(np.int16))
+ self.assert_eq(pser.astype(np.int8), psser.astype(np.int8))
+ self.assert_eq(pser.astype(bool), psser.astype(bool))
Review comment:
Oh... then seems like we're missing tests for pandas 1.2 ??
It's failed on my local with pandas 1.2.
```
Attribute "dtype" are different
[left]: float64
[right]: float32
Left:
0 1.0
1 2.0
2 3.0
dtype: float64
float64
Right:
0 1.0
1 2.0
2 3.0
dtype: float32
float32
```
The result is actually different.
```python
>>> pd.__version__
'1.2.0'
>>> pser.astype(np.float32)
0 1.0
1 2.0
2 3.0
dtype: float64
>>> psser.astype(np.float32)
0 1.0
1 2.0
2 3.0
dtype: float32
```
##########
File path: python/pyspark/pandas/tests/data_type_ops/test_categorical_ops.py
##########
@@ -183,22 +183,31 @@ def test_astype(self):
psser = ps.from_pandas(pser)
self.assert_eq(pser.astype(int), psser.astype(int))
self.assert_eq(pser.astype(float), psser.astype(float))
- self.assert_eq(pser.astype(np.float32), psser.astype(np.float32))
- self.assert_eq(pser.astype(np.int32), psser.astype(np.int32))
- self.assert_eq(pser.astype(np.int16), psser.astype(np.int16))
- self.assert_eq(pser.astype(np.int8), psser.astype(np.int8))
self.assert_eq(pser.astype(str), psser.astype(str))
- self.assert_eq(pser.astype(bool), psser.astype(bool))
self.assert_eq(pser.astype("category"), psser.astype("category"))
+ # Bug in pandas 1.2
+ if (LooseVersion(pd.__version__) >= LooseVersion("1.2")) and (
+ LooseVersion(pd.__version__) < LooseVersion("1.3")
+ ):
+ self.assert_eq(psser.astype(np.float32), ps.Series(data,
dtype=np.float32))
+ self.assert_eq(psser.astype(np.int32), ps.Series(data,
dtype=np.int32))
+ self.assert_eq(psser.astype(np.int16), ps.Series(data,
dtype=np.int16))
+ self.assert_eq(psser.astype(np.int8), ps.Series(data,
dtype=np.int8))
+ self.assert_eq(psser.astype(bool), ps.Series(data, dtype=bool))
+ else:
+ self.assert_eq(pser.astype(np.float32), psser.astype(np.float32))
+ self.assert_eq(pser.astype(np.int32), psser.astype(np.int32))
+ self.assert_eq(pser.astype(np.int16), psser.astype(np.int16))
+ self.assert_eq(pser.astype(np.int8), psser.astype(np.int8))
+ self.assert_eq(pser.astype(bool), psser.astype(bool))
Review comment:
@ueshin Oh, yes I'll narrow down the version range after checking the
all 1.2.x versions.
Thanks!
##########
File path: python/pyspark/pandas/tests/indexes/test_category.py
##########
@@ -176,18 +176,16 @@ def test_astype(self):
self.assert_eq(kcidx.astype("category"), pcidx.astype("category"))
+ # CategoricalDtype is not updated if the dtype is same from pandas 1.3.
if LooseVersion(pd.__version__) >= LooseVersion("1.3"):
- # TODO(SPARK-36367): Fix the behavior to follow pandas >= 1.3
- pass
- elif LooseVersion(pd.__version__) >= LooseVersion("1.2"):
self.assert_eq(
kcidx.astype(CategoricalDtype(["b", "c", "a"])),
Review comment:
yeah, just fixed it. Thanks!
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
