uninitialized array leads to failure in tz_localize ("UserWarning: Inferring time-zone from CET in column __null_dask_index__ failed, using time-zone-agnostic")

What happened:

Cannot get proper tz while reading a dataframe written with tz=“CET” (get naive datetimeindex instead).

Issue probably due to use of np.empty instead of np.zeros in https://github.com/dask/fastparquet/blob/master/fastparquet/dataframe.py#L100 which leads to an exception in https://github.com/dask/fastparquet/blob/master/fastparquet/dataframe.py#L103 as pandas cannot infer the right timezone as the input is garbage (not initalised to zero).

Minimal Complete Verifiable Example:

Difficult to come with as depends on the content of np.empty . issue fixed when replacing np.empty by np.zeros

[martindurant]

cc @jsignell - I forget the dask issue, but it seems we do do timezones, but probably this isn’t applied to the max/min statistics values.

[martindurant]

I don’t think it will fix that - the timezone needs to be applied to the “statistics” values, not to the data values as here.