sushu0 opened a new issue, #45466:
URL: https://github.com/apache/superset/issues/45466
### Bug description
The `compare` pandas post-processing operation raises
`decimal.DivisionByZero` (or `decimal.InvalidOperation` for 0/0) when
percentage or ratio comparisons receive Decimal columns with a zero baseline.
The same operation with float columns completes and returns the existing
infinity/NaN result. Both branches convert their result to float, but that
conversion happens **after** Decimal division has already raised.
This affects raw-column post-processing requests: `columns: ["current",
"previous"]`, `metrics: []`, followed by `compare`. `ExploreMixin.normalize_df`
only coerces metric columns to numbers, so raw SQL DECIMAL columns remain
Decimal objects. I am not claiming that the usual UI time-comparison metric
path is affected; those metric columns are normally converted before
post-processing.
Reproduction against master `99ecfe5801c4cbb79eedd475003b134b6ba7ff28`,
using the real imported operation:
```python
from decimal import Decimal
import pandas as pd
from superset.constants import PandasPostprocessingCompare as PPC
from superset.utils import pandas_postprocessing as pp
df = pd.DataFrame({"current": [Decimal("10")], "previous": [Decimal("0")]})
pp.compare(
df,
source_columns=["current"],
compare_columns=["previous"],
compare_type=PPC.RAT, # PPC.PCT also fails
)
```
Traceback ends at `superset/utils/pandas_postprocessing/compare.py:73`,
before `.astype(float)`:
```text
diff_df = (s_df / c_df).astype(float).round(precision)
decimal.DivisionByZero: [<class 'decimal.DivisionByZero'>]
```
Expected: percentage/ratio comparison of Decimal input follows the
operation's existing float zero-baseline behavior, without changing
absolute-difference precision or deciding new infinity/null rendering semantics.
### Screenshots/recordings
Not applicable; backend post-processing reproduction.
### Superset version
master / latest-dev
### Python version
3.12.14; pandas 2.3.3 from requirements/base.txt; Windows
### Node version
Not applicable
### Browser
Not applicable
### Additional context
Reproduced through the actual operation and project pytest tests, including
a raw-column QueryObject passed through SqlaTable.normalize_df and
exec_post_processing. The regression also covers plain and MultiIndex columns.
No live database/browser/HTTP endpoint reproduction is claimed.
I can prepare the focused fix and regression tests. Codex assisted with
investigation, implementation preparation and local validation; this report
does not represent prior human review of the code. Related Decimal contribution
fixes concern a different operation; I found no matching Apache compare fix in
current issue/PR searches.
### Checklist
- [ ] I have searched Superset docs and Slack and didn't find a solution to
my problem. (Public docs and GitHub searched; Slack not searched.)
- [x] I have searched the GitHub issue tracker and didn't find a similar bug
report.
- [ ] I have checked Superset's logs for errors. (Local Python/pytest
traceback is included above; no running deployment was inspected.)
--
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]
