bug#20897: 25.0.50; [python] sexp-movement are confusing

[Andreas Röhler]

Am 25.06.2015 um 18:46 schrieb Rasmus:
> Hi,
>
> Andreas Röhler <andreas.roehler@easy-emacs.de> writes:
>
> > > Sexp movement in python.el are very confusing to me.  I know that logical
> > > sexp movements outside lisp are subjective,
> > Not more as anything else editor-related. Languages are composed by
> > elements, which a syntax may describe. Even if an editor must not be
> > the slave of a syntax, it should be aware of.
> That is probably true.  I really just want to be able to bent python.el to
> work with my internal "fast" logic which controls how I type on keyboard.
>
> > >    and I know that the behavior
> > > isn't wrong.
> > It behaves arbitrary WRT Python syntax, that's wrong.
> OK.  I'm not sure.  I can somehow imagine parentheses that would justify
> the movements cf. below.  But in lisp it would not go from one "nesting"
> to another, which is essentially what bugs me.
>
>
> > Python is composed by expressions.
> > If inside an expression C-M-f should to to its end.
> >
> >  From end to next end same level if existing - or level up, or next
> > top-level-form, or nil at EOB
> >
> > Backward and forward needs to be consistent.
> I'm not sure I understand.
>
> The way I think about it is like the following.  If I'm at point 2 and
> move backward I really want to be a point 1.

Don't think so WRT picture below. From end of symbol "defun" it move 
back to beginning "d".
Because "defun" is an element of the list.
Next C-M-b should go up backward - but fails. A design flaw IMO.

>   But in python point 0 and 1
> is the same, so it assumes the outer level which has end point 4.  But I
> at most want to go to point 3, closing the "nearest" "sexp".
>
> 01     2          3
> vv     v          V
> ((defun name (arg))
>   ...)
>       ^
>       4
>
> Rasmus
Not sure if I understand the double paren at picture. Rules are simple 
anyway: Either there is an element of same level, reach that. Or go up. 
Or go the end.

You could try py-forward-expression, py-backward-expression.
Just to study the behavior.