bug#6893: evaluating `'#'(lambda (b) b)8 ; => 8 with `eval-last-sexp' "C-x C-e" in `emacs-lisp-mode'

[MON KEY <monkey@sandpframing.com>]

On Sun, Aug 22, 2010 at 6:49 PM, Stefan Monnier
<monnier@iro.umontreal.ca> wrote:
>> So what about these?
>
>> `'#'8
>>  ;=> (quote (function 8))
>
> What about it?  C-M-b indeed shows that '#'8 is the sexp.
>
>>  `#'?\x27?\x23
>>  ;=> (function 39)
>
My point is that with these C M-b:

`'#'

`'

point is skipped back beyond the sexp at point to "M-b:".

Isn't there an implicit list after the last quote created by backquote?

It "feels" like there is with these:

`#1=#1#
;=> (nil)

`','#1=#1#
;=> (quote (nil))

`',#''#1=#1#
;=> (quote (quote (nil)))

`',#','#1=#1#
;=> (quote \,)

`,#','#1=#1#
;=> \,

`',#','nil
;=> (quote \,)

`',#',#1=#1#
;=> (quote \,)

>> Where did 35 go?
>
> That's a known shortcoming of syntax tables which aren't able to
> describe correctly the syntax of Elisp char constants.

Does that affect how the range of "last sexp" is allowed to extend beyond the
beginning-of-line only to find some arbitrary word barier a few lines up?

> But again C-M-b
> shows you correctly what C-x C-e considers as "the last sexp".

Yes well, even w/ C M-b hopping its still not a valid sexp for `eval':

`#'?\x27?\x23
=> (function 39)

(eval (function 39))
;=> 39

(eval `#'?\x27?\x23)
;=> Debugger entered--Lisp error: (wrong-number-of-arguments eval 2)

And besides, theres this anomaly too:

#&8"\377"
;=> "\377" (actually on Emacs 23.2 its => "ÿ")

`#&8"\377"
;=> "\377" (actually on Emacs 23.2 its => "ÿ")

`#&8"\x3fffff"
;=> "\377" (actually on Emacs 23.2 its => "ÿ")

Which would make sense except that _both_ of these manages to fail "correctly":

`'#s

`'#s" "
     !string

=> Debugger entered--Lisp error: (invalid-read-syntax "#")

Despite the fact that C M-b for either of this forms:

`'#s(hash-table size 30 data (key1 val1 key2 300))

#s(hash-table size 30 data (key1 val1 key2 300))

puts point just after "#s".

I understand that the hash-table read syntax has a higher priority because its
newer but the cummulative effect of the various sharpsign quoting/evaluation
rules for the immediate sexp before point feels inconsistently biased when
viewed in the aggregate.

>
>
>        Stefan

--
/s_P\