Fix Pandas 0.24 DateOffset bug pt. 2

[villebro]

CATEGORY

Choose one

• Bug Fix

SUMMARY

This should fix the DateOffset bug that was introduced when moving from Pandas 0.23 to 0.24. Credits for this go to @jbrockmendel, see pandas-dev/pandas#27728

TEST PLAN

Tested locally

ADDITIONAL INFORMATION

• Has associated issue: Fixes Time-series Period Pivot - DateOffset objects are immutable #7966

REVIEWERS

@betodealmeida @AndLLA

[AndLLA]

The patch works for my use case (time grain=day, frequency 1week)
Thanks !!!!

I tried an alternative configuration (freq = 4weeks), but the chart is still the same as for 1week.

I also tried "frequency = day", but I get an error (but maybe I'm miss-using the chart): Tick offset with normalize=True are not allowed.

Traceback (most recent call last):
  File "/home/superset/venvSS/lib/python3.6/site-packages/superset/views/base.py", line 122, in wraps
    return f(self, *args, **kwargs)
  File "/home/superset/venvSS/lib/python3.6/site-packages/superset/utils/decorators.py", line 70, in wrapper
    return f(*args, **kwargs)
  File "/home/superset/venvSS/lib/python3.6/site-packages/superset/views/core.py", line 1097, in explore_json
    viz_obj, csv=csv, query=query, results=results, samples=samples
  File "/home/superset/venvSS/lib/python3.6/site-packages/superset/views/core.py", line 1018, in generate_json
    payload = viz_obj.get_payload()
  File "/home/superset/venvSS/lib/python3.6/site-packages/superset/viz.py", line 379, in get_payload
    payload["data"] = self.get_data(df)
  File "/home/superset/venvSS/lib/python3.6/site-packages/superset/viz.py", line 1401, in get_data
    freq = type(freq)(freq.n, normalize=True, **freq.kwds)
  File "/home/superset/venvSS/lib/python3.6/site-packages/pandas/tseries/offsets.py", line 2247, in __init__
    raise ValueError("Tick offset with `normalize=True` are not "
ValueError: Tick offset with `normalize=True` are not allowed.

[villebro]

Thanks for testing @AndLLA ! Let's get that last error sorted, I think we're pretty close.

[villebro]

Can you do another round of testing @AndLLA ? Seems to work for me.

[jbrockmendel]

FYI the Tick offsets (Day, Hour, Minute, ...) had normalize disallowed because they caused unexpected arithmetic behavior.

More generally, the immutability was put in place because it led to major performance improvements when working with PeriodIndex/Array.

[AndLLA]

It works also for me :)
Thanks !!!!!

[codecov-io]

Codecov Report

Merging #7981 into master will decrease coverage by 0.01%.
The diff coverage is 0%.

@@            Coverage Diff             @@
##           master    #7981      +/-   ##
==========================================
- Coverage   65.58%   65.57%   -0.02%     
==========================================
  Files         469      469              
  Lines       22407    22409       +2     
  Branches     2432     2432              
==========================================
- Hits        14695    14694       -1     
- Misses       7591     7594       +3     
  Partials      121      121

Impacted Files	Coverage Δ
superset/viz.py	71.59% <0%> (-0.16%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update b856666...0b42137. Read the comment docs.

[betodealmeida]

LGTM, but we should address the underlying problem at some point.