Re: [O] info URL « open at point » patch

[Vincent Belaïche]

I generated the regexp with (regexp-opt '("ftp" "http" "https"
"file")). I admit that the list of protocols is not clear from the
regexp, but it was already done this way before my change, and I tried
to minimize the change.
More concerning in my opinion is how the Info-get-token is written. As
some point of the code one can read the following statement:
      ;; First look for a match for START that goes across POS.
      (while (and (not (bobp)) (> (point) (- pos (length start)))
    (not (looking-at start)))
 (forward-char -1))


Here start is a regexp, so (length start) is just the length of the
string holding the regexp, not the max length over which the regexp
match can span. For instance assume that aaaaaaaaaa is a new new
protocol which you want Info-get-token to catch. This is 10 character
long, but the typical regexp to catch it would be « a\{10\} » which is
only 7 character long.
It would propably be cleaner to provide the value to be used instead of
(length start) as a separate optional argument that would be set to
(length start) if omitted.
Another way would be to have some standard function max-matchable-length
that given some regexp would compute the maximum length of its match (or
output some special value like t if the maximum length is infinite),
thus (length start) could be replaced by (max-matchable-length start)
--- maybe this is already somthing existing. In the same vein there
could be some standard function that given regexp re and some position
pos-in would function position pos-out such that the following
expression would be true:
(save-excursion
  (goto-char pos-out)
  (and (<= pos-out pos-in)
       (looking-at re)
       (>= (match-end 0) pos-in)))
I think that such standard function is already existing, but I can't
remember the package name which provides it…
  V.