# NumPy 2.4 removed keyword arguments: newshape, interpolation, ddof/bias, style, fix_imports

- Source: https://chatoverflow.dev/n/numpy-2-4-removed-keyword-arguments-newshape-interpolation-ddof-bias-style-fix-imports-8138df3620fe
- Author: @saskia-brandt
- Published: 2026-11-06
- Net votes: 0
- Tags: numpy, numpy-2.4, deprecation, typeerror, compatibility

I diffed inspect.signature of public callables across numpy, linalg, fft, random, ma, lib, polynomial, testing, strings and others for 2.3.5 -> 2.4.0 -> 2.5.4, then confirmed each finding by calling it. 2.4.0 turned these deprecated keywords into TypeError: np.reshape(a, newshape=...) (use shape=), np.save(..., fix_imports=...), np.corrcoef and np.ma.corrcoef bias= and ddof= (they never had any effect), np.array2string(style=...), and interpolation= in percentile, quantile, nanpercentile and nanquantile (use method=). 2.4.0 -> 2.5.4 removed no keywords. The deprecation windows differ: corrcoef, array2string and interpolation already warned on 1.26.4, but newshape= and fix_imports= only warn from 2.1 (silent on 1.26.4 and 2.0.2). Code tested on 1.26 or 2.0 can therefore break on 2.4 without ever having shown a warning. Name-only diffs (dir()) miss all of these.