+ ,ieaR t0tRt^RIHtHt^RIHtHtH t .t ]^kRR/Rlt RR/RltRR/R ltRR/R lt RR lt RR ltR#)a Extra testing functions that forbid 0d-input, see #21044 While the xp_assert_* functions generally aim to follow the conventions of the underlying `xp` library, NumPy in particular is inconsistent in its handling of scalars vs. 0d-arrays, see https://github.com/numpy/numpy/issues/24897. For example, this means that the following operations (as of v2.0.1) currently return scalars, even though a 0d-array would often be more appropriate: import numpy as np np.array(0) * 2 # scalar, not 0d array - np.array(0) # scalar, not 0d-array np.sin(np.array(0)) # scalar, not 0d array np.mean([1, 2, 3]) # scalar, not 0d array Libraries like CuPy tend to return a 0d-array in scenarios like those above, and even `xp.asarray(0)[()]` remains a 0d-array there. To deal with the reality of the inconsistencies present in NumPy, as well as 20+ years of code on top, the `xp_assert_*` functions here enforce consistency in the only way that doesn't go against the tide, i.e. by forbidding 0d-arrays as the return type. However, when scalars are not generally the expected NumPy return type, it remains preferable to use the assert functions from the `scipy._lib._array_api` module, which have less surprising behaviour. )array_namespaceis_numpy)xp_assert_closexp_assert_equalxp_assert_lessxpNc RpVf \V4pVPV4p\V4'dVPR8XgR#RpVP V4'gQV4hR#)TNaResult is a NumPy 0d-array. Many SciPy functions intend to follow the convention of many NumPy functions, returning a scalar when a 0d-array would be correct. The specialized `xp_assert_*` functions in the `scipy._lib._array_api_no_0d` module err on the side of caution and do not accept 0d-arrays by default. If the correct result may legitimately be a 0d-array, pass `check_0d=True`, or use the `xp_assert_*` functions from `scipy._lib._array_api`.)rasarrayrshapeisscalar)actualdesiredrkwargs__tracebackhide___msgs&&$, =/usr/lib/python3/dist-packages/scipy/_lib/_array_api_no_0d.py _check_scalarr#sd z V $jj!G RLLW]]b0 OD ;;v  $$ check_0dFc JRpV'g \W3/VB\W3RV/VB#Tr)rxp_assert_equal_baser rrrrs&&$, rrr;s1 f00  M( Mf MMrc JRpV'g \W3/VB\W3RV/VB#r)rxp_assert_close_basers&&$, rrrFs- f00  M( Mf MMrc JRpV'g \W3/VB\W3RV/VB#r)rxp_assert_less_basers&&$, rrrOs- f00 v L LV LLrc T^R^ V),,re\W.VO5RVRVRRRR/VB#zPBackwards compatible replacement. In new code, use xp_assert_close instead. g?atolrtol check_dtypeF check_shaperr rdecimalargskwdsr!r s&&&*, rassert_array_almost_equalr)XUCgX&$ 6 *  * $ *+/ *=B *PU *$( **rc T^R^ V),,re\W.VO5RVRVRRRR/VB#rr$r%s&&&*, rassert_almost_equalr,ar*rcVV^8dQh/^\9d\\,;R&#)__all__)__conditional_annotations__liststr)formats"r __annotate__r4s">c?r)))r0__doc__scipy._lib._array_apirrrrrrrrr/rr)r,r4)r0s@rr9sf4<JJ%%0NNNNMM**r