replace function type with function isinstance

[frankfqchen]

What changes were proposed in this pull request?

replace function type with function isinstance

How was this patch tested?

(Please explain how this patch was tested. E.g. unit tests, integration tests, manual tests)

(If this patch involves UI changes, please attach a screenshot; otherwise, remove this)

[AmplabJenkins]

Can one of the admins verify this patch?

[HyukjinKwon]

I think we need a JIRA because type and isinstance are not exactly same. Also, maybe it'd better if the PR descriptions explains the bug and how this PR tries to resolve it.

BTW, it seems you intend to support sub-classes via isinstance consistently across the API, right?

If so, there are some instances similar with this. Maybe we should check those as well.

$ grep -r "type(.*) [=|\!]" . | grep .python | grep -v "tests.py"

./python/pyspark/ml/linalg/__init__.py:    elif type(v) == np.ndarray:
./python/pyspark/ml/linalg/__init__.py:        if type(other) == np.ndarray:
./python/pyspark/ml/linalg/__init__.py:            if type(pairs) == dict:
./python/pyspark/ml/param/__init__.py:        if type(value) == list:
./python/pyspark/ml/param/__init__.py:        elif type(value) == np.unicode_:
./python/pyspark/ml/param/__init__.py:        if type(value) == bool:
./python/pyspark/mllib/linalg/__init__.py:    elif type(v) == np.ndarray:
./python/pyspark/mllib/linalg/__init__.py:        if type(other) == np.ndarray:
./python/pyspark/mllib/linalg/__init__.py:            if type(pairs) == dict:
./python/pyspark/mllib/stat/_statistics.py:        if type(y) == str:
./python/pyspark/sql/column.py:        if type(startPos) != type(length):
./python/pyspark/sql/readwriter.py:            if type(path) != list:
./python/pyspark/sql/readwriter.py:        if type(path) == list:
./python/pyspark/sql/streaming.py:        if type(interval) != str or len(interval.strip()) == 0:
./python/pyspark/sql/streaming.py:            if type(path) != str or len(path.strip()) == 0:
./python/pyspark/sql/streaming.py:        if not outputMode or type(outputMode) != str or len(outputMode.strip()) == 0:
./python/pyspark/sql/streaming.py:        if not queryName or type(queryName) != str or len(queryName.strip()) == 0:
./python/pyspark/sql/streaming.py:            if type(processingTime) != str or len(processingTime.strip()) == 0:
./python/pyspark/sql/types.py:        return type(self) == type(other)

[HyukjinKwon]

If this change is legitimate, we should change this to isinstance(v, (array.array, list, tuple, xrange))

[holdenk]

Thanks for looking to get involved @frankfqchen :)

I think this PR definitely needs a JIRA and maybe some more description about what the intent of the change is rather than just the contents of the change. The Spark Contributing guide can help you out - https://cwiki.apache.org/confluence/display/SPARK/Contributing+to+Spark

[HyukjinKwon]

@frankfqchen Could you follow the comment above? If you are not able to proceed further, I think it might be better closed for now. Actually, IMHO, I am not too sure if it is worth sweeping them.