We've noticed two major issues with z.show() after upgrading Zeppelin 1) z.show(df) used to work directly on spark dataframe object, now it produces TypeError: object of type 'DataFrame' has no len() Full exception stack in [1].
We tried disabling ipython and it seems to be a workaround. I there is a way to have compatibility with previous Zeppelin release on z.show() without disabling ipython altogether? https://issues.apache.org/jira/browse/ZEPPELIN-3234 2) The new UI grid displays just an empty box when output is cut with a message like > Output is truncated to 102400 bytes. Learn more about > ZEPPELIN_INTERPRETER_OUTPUT_LIMIT It doesn't happen every time, I think it depends on how interpreter has cut the table? https://issues.apache.org/jira/browse/ZEPPELIN-3235 More minor - the new data grid visualization seems much slower on wider datasets. Not sure if there is a way to fallback to older table data grid UI option? [1] TypeErrorTraceback (most recent call last) <ipython-input-4-47beac9208ab> > in <module>() > ----> 1 z.show(spark.sql('select * from disc_mrt.unified_fact')) > <ipython-input-1-b24458af29e1> in show(self, p, **kwargs) > 73 # `isinstance(p, DataFrame)` would req `import > pandas.core.frame.DataFrame` > 74 # and so a dependency on pandas > ---> 75 self.show_dataframe(p, **kwargs) > 76 elif hasattr(p, '__call__'): > 77 p() #error reporting <ipython-input-1-b24458af29e1> in show_dataframe(self, > df, show_index, **kwargs) > 80 """Pretty prints DF using Table Display System 81 """ ---> 82 limit = > len(df) > self.max_result 83 header_buf = StringIO("") > 84 if show_index: TypeError: object of type 'DataFrame' has no len() > > -- Ruslan Dautkhanov
