I'm observing the python grammar listed in the manual and considering the outputs of their form of EBNF, specifically with varargslist:
varargslist: (vfpdef ['=' test] (',' vfpdef ['=' test])* [',' [
'*' [vfpdef] (',' vfpdef ['=' test])* [',' ['**' vfpdef [',']]]
| '**' vfpdef [',']]]
| '*' [vfpdef] (',' vfpdef ['=' test])* [',' ['**' vfpdef [',']]]
| '**' vfpdef [',']
Though I'm specifically interested in this section:
['*' [vfpdef] (',' vfpdef ['=' test])* ]
Which I interpret as:
[ [ non-terminal1 ] ( non-terminal2) ]
I realize that both
non-terminal1 (non-terminal2)
(non-terminal2)
Are valid options in this form, but does this include:
non-terminal1
as well? The wiki page for EBNF states
That is, everything that is set within the square brackets may be
present just once, or not at all
but does this group everything within the square brackets as one entity that may appear only once, or is the option selective, for example:
[ [non-terminal1] [(non-terminal2)] ]
If
['*' [vfpdef] (',' vfpdef ['=' test])* ]
is
[ [ non-terminal1 ] non-terminal2 ] -- parentheses deleted as redundant
then non-terminal2 represents
non-terminal3 *
which is nullable by definition. (That is, it might be empty.)
So, strictly speaking, once you've done the transform
non-terminal1
is not a valid outcome. The parse must be
non-terminal1 non-terminal2
where non-terminal2 has matched an empty string.
But the actual parse logic is more likely to want to use the formulation
[ [ non-terminal1 ] non-terminal3... ] -- Not EBNF syntax, but I hope you get the idea
in which non-terminal2 has been eliminated as a distraction from the resulting parse. In this case, since the 0-or-more repetition can be 0 repetitions, correct outcomes would include
-- nothing :-)
non-terminal1
non-terminal3
non-terminal1 non-terminal3
non-terminal3 non-terminal3
and so on.