Improve caching experience on InteractiveRunner with dataframes

Reusing the default label for to_pcollection when using the interactive runner results in caching errors when used with multiple pipelines:

 

 

{{Traceback (most recent call last):
File "/home/srohde/Workdir/beam/sdks/python/apache_beam/runners/interactive/interactive_runner_test.py", line 389, in test_dataframes_with_multi_index_get_result
pd.testing.assert_series_equal(df_expected, ib.collect(deferred_df, n=10))
File "/home/srohde/Workdir/beam/sdks/python/apache_beam/runners/interactive/utils.py", line 247, in run_within_progress_indicator
return func(*args, **kwargs)
File "/home/srohde/Workdir/beam/sdks/python/apache_beam/runners/interactive/interactive_beam.py", line 579, in collect
recording = recording_manager.record([pcoll], max_n=n, max_duration=duration)
File "/home/srohde/Workdir/beam/sdks/python/apache_beam/runners/interactive/recording_manager.py", line 433, in record
self._watch(pcolls)
File "/home/srohde/Workdir/beam/sdks/python/apache_beam/runners/interactive/recording_manager.py", line 306, in _watch
for pcoll in to_pcollection(*watched_dataframes, always_return_tuple=True):
File "/home/srohde/Workdir/beam/sdks/python/apache_beam/dataframe/convert.py", line 196, in to_pcollection
new_results =

{p: extract_input(p) File "/home/srohde/Workdir/beam/sdks/python/apache_beam/transforms/ptransform.py", line 1086, in __ror__ return self.transform.__ror__(pvalueish, self.label) File "/home/srohde/Workdir/beam/sdks/python/apache_beam/transforms/ptransform.py", line 587, in __ror__ raise ValueError( ValueError: Mixing value from different pipelines not allowed.}

}