Define column-wise na/default value for merge operations.

[gerigk]

I find myself often using fillna on a lot of columns after a merge operation and having to cast columns back to their original dtype.
Pseudo code now

data_types = df.dtypes
df_new = merge(df, df_other)
for col in former_int64_columns:
df_new[col] = df_new[col].fillna(0).astype(np.int64)

in some cases I may want a different default value, also for other column types than int64.

An idea would be something like a keyword argument "column_default_values = { 'a': 0, 'b': -1, ...}".

[petehuang]

This issue's last interaction was 4 years ago. In the most recent version of pandas, you just need to call df.fillna(0) instead of iterating through each of the columns. You can also use a dict of values for column-specific fill values.

However, passing a default value into merge is still not supported.

[jreback]

merge columns are pretty good dtype-wise. if there is a more specific example, please open a new issue with a repro.