Pandas updates and deprecations

[doctor-phil]

Pandas is constantly updating and making big changes and deprecating methods; some of which are core to certain lectures. It's getting harder to keep that pinned version <2.0, since the student's environments are more up to date.

I am worried that eventually upgrading pandas will have huge impacts on some of these notes. Especially the lecture on reshaping.

Collecting relevant substantial changes here so I can go back and try to update it for next term.

• stack and unstack changed behavior somehow
• issue with downcasting on pd.DataFrame.fillna?
• possible changes to bfill and ffill

add more here as they arise

[doctor-phil]

• applymap -> map

[doctor-phil]

• value_counts won't apply twice in the chapter on merging, goodreads example. Something to do with count column

[doctor-phil]

• in groupby.md, in the section on pd.Grouper, the line start=pd.datetime.today()... should be start = pd.Timestamp.today()...

[doctor-phil]

• quandl python API is deprecated in favor of nasdaq-data-link. Different syntax and data (columns) also different in pandas/timeseries.md
• after updating to use nasdaq-data-link, the "delorean" exercises are no longer possible (no open or close price in the new table)