briochh
commented
9 months ago e.g.
- There seem to be issues setting a new column as a string using the
.loc[idx, col]slice accessor: https://github.com/pandas-dev/pandas/issues/55025. Presumably the internals set the column to np.nan which is then falling short of the impending strict upcasting ban https://pandas.pydata.org/pdeps/0006-ban-upcasting.html when it try to putstrvalues in there. Results in > "FutureWarning: Setting an item of incompatible dtype is deprecated and will raise in a future error of pandas." - There is a bug in 2.1 that is not dealing with columns containing commas when reading whitespace delimited https://github.com/pandas-dev/pandas/issues/54918
- Concatenating with an empty dataframe now also raises a warning. Either we need to ensure not empty before or (eventually) dropna(all) after concatenation (I think); I guess the latter will continue to raise a warning for now...
- Accessing series entries using numpy-like index
df.col[i]is deprecated. The expectation now is to usedf.col.iloc[i]or presumably,df.col.values[i]or maybedf.iloc[i].col(maybe?!)
A few issues with behaviour of pandas 2.1.0 that causes failed tests and a whole string of Future and Deprecation warnings
Need to assess what actually reflects an impending breakage in how we are using pandas, going forwards and what might be temporary pandas "features" as they move towards pyarrow.