On 12/08/2021 10.32, Dennis Lee Bieber wrote: > On Wed, 11 Aug 2021 11:59:11 -0300, Jack Brandom <jbran...@example.com> > declaimed the following: >> Where are these production rules coming from? They're not at >> https://docs.python.org/3/reference/grammar.html >> The word ``stride'' doesn't appear in this grammar. > > Possibly from older versions of the grammar, before simplification to > just <expression>. Originally, the "stride" term was added in response to > requests from the numerical library developers (NumPy seems to be the > survivor of that cluster). At the time, it wasn't envisioned to be usable > with regular Python objects. > > https://en.wikipedia.org/wiki/Array_slicing#1991:_Python > """ > The stride syntax (nums[1:5:2]) was introduced in the second half of the > 1990s, as a result of requests put forward by scientific users in the > Python "matrix-SIG" (special interest group). > """ > > As for the grammar... Would you prefer that from 2.7? > > trailer: '(' [arglist] ')' | '[' subscriptlist ']' | '.' NAME > subscriptlist: subscript (',' subscript)* [','] > subscript: '.' '.' '.' | test | [test] ':' [test] [sliceop] > sliceop: ':' [test] > > test: or_test ['if' or_test 'else' test] | lambdef > or_test: and_test ('or' and_test)* > and_test: not_test ('and' not_test)* > not_test: 'not' not_test | comparison
Are you trying to give us nightmares @wulfraed? Yes, the somewhat piecemeal coverage(s) are explained by changes over time, and the different applications folk might apply or differing motivations of the people involved. I've been trying to remember if we had negative-steps in FORTRAN do-loops especially once the capability to define subscripting-ranges came 'in' (but can't be bothered researching further). If it was available, or available within other, older languages, then might this explain Python's adoption? It is quite plain (Ref.Man 6.3.3 and 3.2) that initially the plans for stride/step did not include the negative 'direction'. Perhaps the initial coders thought: 'while I'm doing this, adding ... will be easy to do', aka "famous last words", see also "YAGNI". Negative-stepping comes-across as a bit of a party-trick. I wonder if anyone can offer a (professional) application? Similarly, when adding a __getitem__() to a custom-class, has anyone had need to implement/account for negative-stepping, to achieve some purpose? PS I'm waiting (with bated breath*) for @Chris' bared teeth... * or is that "baited"? -- Regards, =dn -- https://mail.python.org/mailman/listinfo/python-list
