Closed sbrugman closed 2 years ago
After a lot of debugging, I found that there is a bug in histogrammar with Spark DataFrames. The histogrammar converts timestamps in nanoseconds for binning calculations. However, it fails to convert it back to timestamp for Spark DataFrames. It is a bit more explained in the issue here. I have fixed the issue in histogrammar repo and submitted a PR.
When the user provides a timestamp-typed time_axis in PySpark, the time axis is binned in (nano)seconds. This should be displayed in as datetimes in the plots.