A relativelly simple improvements to the current Any-based placeholder annotations would be to replace them with simple unannated function definitions. For the main numpy namespace such enhancements were recently implemented (see below); ideally this can be done as well for all other namespaces before the 1.21 release.

TODO

• np main namespace: ENH: Improve the placeholder annotations for the main numpy namespace #18765, ENH: Improve the placeholder annotations for the main numpy namespace (part 2) #18780 and MAINT: Improve the placeholder annotations for the main numpy namespace (part 3) #18799
• np.core namespace: private namespace; deliberately kept empty
• np.char namespace: ENH: Improve the placeholder annotations within sub-modules #18842
• np.ctypeslib namespace: ENH: Improve the placeholder annotations within sub-modules #18842
• np.emath namespace: ENH: Improve the placeholder annotations within sub-modules #18842
• np.f2py namespace: ENH: Improve the placeholder annotations within sub-modules #18842
• np.fft namespace: ENH: Improve the placeholder annotations within sub-modules #18842
• np.lib namespace: ENH: Improve the placeholder annotations within sub-modules (part 2) #18853
• np.linalg namespace: ENH: Improve the placeholder annotations within sub-modules (part 2) #18853
• np.ma namespace: ENH: Improve the placeholder annotations within sub-modules (part 3) #18909
• np.matrixlib namespace: ENH: Improve the placeholder annotations within sub-modules #18842
• np.polynomial namespace: ENH: Improve the placeholder annotations within sub-modules (part 3) #18909
• np.random namespace: fully annotated prior to this issue
• np.rec namespace: ENH: Improve the placeholder annotations within sub-modules #18842
• np.testing namespace: ENH: Improve the placeholder annotations within sub-modules #18842 and ENH: Add placeholder annotations for two missing np.testing objects #18919
• np.typing namespace: fully annotated prior to this issue