Deprecation warnings with pandas 3.1 (inplace, read_csv(float_precision=...)) #570

pandas 3.1.0 deprecates the inplace keyword of drop, rename, reset_index, set_index, sort_index, sort_values and others (PDEP-8), and the float_precision argument of read_csv. Both emit Pandas4Warning (a DeprecationWarning subclass) and will stop working in pandas 4.

petab uses both when loading any PEtab table, so every Problem.from_yaml warns. Downstream test suites that turn warnings into errors fail (e.g., AMICI, see AMICI-dev/AMICI#3440).

Affected places on main (68be6d1):

• read_csv(..., float_precision="round_trip"): v1/conditions.py, v1/core.py (2×), v1/mapping.py, v1/measurements.py, v1/observables.py, v1/parameters.py, v2/conditions.py, v2/experiments.py
• reset_index(inplace=True) / set_index(..., inplace=True): v1/conditions.py, v1/mapping.py, v1/observables.py, v1/parameters.py (incl. l. 243), v1/simplify.py
• sort_index(inplace=True): v1/visualize/plotting.py (2×)
• rename(..., inplace=True) / drop(..., inplace=True): v2/petab1to2.py (5×)

(fillna(..., inplace=True) in v1/visualize/lint.py and v2/petab1to2.py is not affected.)

The inplace uses can be replaced by re-assignments. For float_precision, pandas 3.1 uses the same (improved) converter for all precision modes (pandas#64395), but older pandas versions don't parse round-trip by default. So the argument should probably only be passed for pandas < 3.1.

🤖 Generated with Claude Code