Fix decoding of string in python 3 when rerunning time series

[xylar]

There is a bug when rerunning analysis in python 3 with time series output. The xtime_startMonthly variable was not being converted correctly to a string. This has been fixed.

[xylar]

@milenaveneziani, the problem can be reproduced by running a case of your choice on edison using develop with:

export PATH=/global/project/projectdirs/acme/software/anaconda_envs/edison_mpas_analysis_py3/bin:$PATH
./run_mpas_analysis --purge config.<case> --generate=timeSeriesSST
./run_mpas_analysis config.<case> --generate=timeSeriesSST

If you run the same with this branch, you should no longer see the error.

[milenaveneziani]

Did the test you mentioned and all worked.

[milenaveneziani]

@xylar, leaving a comment here just in case it is not necessary to open an issue: I noticed that, even if I request a time series starting from say year=10, I still get a message on screen that says:

Computing MPAS time series from files:
    mpaso.hist.am.timeSeriesStatsMonthly.0001-01-01.nc through
    mpaso.hist.am.timeSeriesStatsMonthly.0010-12-01.nc

ie, starting from year 1. I assume just the message should be adjusted?

[xylar]

@milenaveneziani, please make an issue for the time series range. I think the issue might be that the first year is included in every time series, regardless of what is requested.

[xylar]

Thanks for the quick review!

[milenaveneziani]

I think the issue might be that the first year is included in every time series, regardless of what is requested.

yes, that is what I think is happening too. I looked into it a bit, and I think it all originates here:

self.inputFiles = self.historyStreams.readpath(
                streamName, startDate=startDate, endDate=endDate,
                calendar=self.calendar)

on line 144 of mpas_time_series_task.py.
So, I think we just have to leave it be.