Skip to content

scipy/rng-keyword

Rule rng-keyword
Pack SciPy
Severity info
Fix unsafe, with xpd check --fix --unsafe-fixes
Library versions SciPy 1.15 and later

Since SciPy 1.15, these functions take rng=, and SciPy's docs say new code should use it in place of random_state= or seed=. Pass rng=, and expect other random numbers for the same seed.

engine = qmc.Sobol(d=2, seed=0)
U, s, Vt = svds(A, k=3, random_state=0)
engine = qmc.Sobol(d=2, rng=0)
U, s, Vt = svds(A, k=3, rng=0)

The rule reports random_state= and seed= passed by keyword to the functions and classes that SciPy 1.15 moved to rng=. SciPy's docs say new code should pass rng. Examples are sparse.random, svds, bootstrap, permutation_test, kmeans2, differential_evolution, basinhopping, dual_annealing, directed_hausdorff, Rotation.random and the qmc engines. The old keywords run without a warning on SciPy 1.18.1.

The rule is skipped on SciPy older than 1.15. It leaves alone a call under a SciPy version check, because rng= raises on an older SciPy. xpd reads only the installed SciPy and can't see which SciPy versions a library supports. The message therefore says to keep the old keyword when the code must also run on SciPy older than 1.15.

xpd check --fix --unsafe-fixes applies the fixes of this rule. Check the result, because an unsafe fix can change what the code does.

The unsafe fix writes rng= for random_state= and seed=. The same integer gives a different stream under rng=, and qmc.Sobol raises for a RandomState instance passed as rng=.

On sparse.random, sparse-matrix-api then advises sparse.random_array, which returns a sparse array.

xpd prints this message with the issue. A name that starts with $ stands for code from your file.

$F takes rng= since SciPy 1.15, and SciPy's docs say new code should pass rng instead of random_state=. The same integer gives different numbers under rng, so expect results to change. Keep random_state= if the code must also run on a SciPy older than 1.15, where rng= raises TypeError.

Browse the other rules of the SciPy pack.