SEPURI-SAI-KRISHNA commented on PR #45083:
URL: https://github.com/apache/superset/pull/45083#issuecomment-6542385369
Resolved on the shape you gave, keyed by target since a rename means the two
names differ:
```python
for source, target in overwritten.items():
try:
_base_df.loc[:, [target]] = append_df[[source]].rename(
columns={source: target}
)
except TypeError:
_base_df[target] = append_df[source]
```
The added `rename` is free. Against master's loop on the five same-name
shapes
(lossless float into int64, NaN into int64, float into float, int into int,
fractional float into int64) both forms take the same branch and return equal
frames. The swap, the plain append and three MultiIndex shapes are right too.
You were right that the bulk set breaks wider than NaN alone. Six of eight
cases raise under it, plain `{"y": "y"}` included.
Added `test_append_columns_widens_an_int_target_the_new_values_do_not_fit`
for
the fallback: an int `z`, a NaN in the source, reached through a rename,
which
is the path only this PR creates. 215 pass in the directory, pre-commit
clean.
Re-censused against the new base: 13 conditions, 12 fail a named test when
reverted on their own. The thirteenth is the list-wrap on the write target,
which no reachable input separates. Only a tuple-keyed mapping would, and
`validate_column_args` matches keys against level 0 on a MultiIndex frame, so
one cannot arrive. On a real pivot cross product the two forms are identical,
and `rolling` with `{"sum_m": "sum_n"}` over one comes out correct.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
