17

On the Code Golf Stack Exchange site today, I found this answer in Clojure to the question "Get all links on a webpage".

(->> (slurp "http://www.stroustrup.com")
     (re-seq #"(?:http://)?www(?:[./#\+-]\w*)+"))

Without the fancy macro, it's just this:

(re-seq #"(?:http://)?www(?:[./#\+-]\w*)+" (slurp "http://www.stroustrup.com"))

This returns the list:

("http://www.morganstanley.com/" "http://www.cs.columbia.edu/" "http://www.cse.tamu.edu" ...)

Can I do something similar in Emacs Lisp?

Perhaps a function like (re-seq regexp (buffer-string)) that returns '(firstmatch secondmatch thirdmatch ...)?

  • This is what M-x occur does, but I'd look inside for more low-level functions to do that. – wvxvw Jan 7 '15 at 21:35
  • @wvxvw That's a good point, I didn't even think about occur. I'll have to look through its source. – nanny Jan 7 '15 at 21:42
  • I looked inside, and oh woe, that code does too much and it's not easy to repurpose it, not at all. My next candidate would be s.el, but maybe there's more out there. Here: github.com/magnars/s.el#s-match-strings-all-regex-string how about this? – wvxvw Jan 7 '15 at 22:00
14

Here is how you can do it based on strings, as requested.

(defun re-seq (regexp string)
  "Get a list of all regexp matches in a string"
  (save-match-data
    (let ((pos 0)
          matches)
      (while (string-match regexp string pos)
        (push (match-string 0 string) matches)
        (setq pos (match-end 0)))
      matches)))

; Sample URL
(setq urlreg "\\(?:http://\\)?www\\(?:[./#\+-]\\w*\\)+")
; Sample invocation
(re-seq urlreg (buffer-string))
  • That looks not quite complete, could you expand this to a fully working answer? – wasamasa Jan 7 '15 at 22:05
  • 1
    The code was complete, but I also added an example of usage. What else would you like to see? – Alan Shutko Jan 7 '15 at 22:11
  • This solution is too simple, unfortunately. Try (re-seq "^.*$" ""). Valid regexp, valid string, but it never terminates. – Phil Lord Dec 1 '17 at 23:19
6

It's probably worth noting that invoking occur with the universal argument causes it to populate the *Occur* buffer with only matches — no file names, line numbers or header information. When combined with a capture group, this allows one to extract whatever pattern is desired.

For example, C-u M-x occur followed by \"\(.*\)\" will prompt the user for which capture group to collect (default \1), and then place the content of every quoted string into the *Occur* buffer.

5

I have an emacs lisp answer to that question posted: https://codegolf.stackexchange.com/a/44319/18848

Using the same (while (search) (print)) structure you could modify it into a function to push matches in a buffer to a list and return it like this:

(defun matches-in-buffer (regexp &optional buffer)
  "return a list of matches of REGEXP in BUFFER or the current buffer if not given."
  (let ((matches))
    (save-match-data
      (save-excursion
        (with-current-buffer (or buffer (current-buffer))
          (save-restriction
            (widen)
            (goto-char 1)
            (while (search-forward-regexp regexp nil t 1)
              (push (match-string 0) matches)))))
      matches)))
  • Nice answer, note you may want to replace match-string with match-string-no-properties so the syntax highlight isn't extracted. You may want to pass a regexp-group-index to use so you can choose which text is stored. As well as reversing the order of searching (current list is last-to-first). See this answer which includes a modified version emacs.stackexchange.com/a/38752/2418 – ideasman42 Feb 12 '18 at 0:53
2

Using s.el this would've been shorter, but, unfortunately, it gives too many matches:

(defun all-urls-in-buffer ()
  (s-match-strings-all
   "\\(?:http://\\)?www\\(?:[./#+-]\\w*\\)+"
   (buffer-string)))

If this is ok (the regex for URLs isn't perfect anyway), this just might be shorter, and if not, then I don't think I could make it shorter than Alan Shutko's answer.

2

Let me just mention why I think this is not implemented in the core. Simply for efficiency reasons: there's no need to copy, create lists, pass them around and garbage collect them. Instead, store the whole string as the buffer, and operate with integer match bounds. That's how occur works, for instance: it matches one string at a time and inserts the match into *occur*. It doesn't match all the strings at once, make them into the list, loop on the list to insert into *occur* and garbage collect the list and its strings.

Just like you wouldn't write (do (def x 1) (def x (+ 2 x))) in Clojure, you shouldn't by default try to have Elisp behave like a functional language. I'd love it if it was, but we have to make due with what we've got at the moment.

1

If I may be allowed a plug, take a look at my "m-buffer" library.

(m-buffer-match buffer "foo")

Returns a list of markers to matches to foo.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.