Re: extract lines with regexp

[Raymond Wiker]

Ted Zlatanov <tzz@lifelogs.com> writes:

> On Thu, 30 Apr 2009 11:57:06 -0700 (PDT) Xah Lee <xahlee@gmail.com> wrote: 
>
> XL> (let (p1 p2)
> XL>   (save-excursion
> XL>     (goto-char (point-min))
> XL>     (search-forward-regexp "^A.+$") ; begin pattern
> XL>     (setq p1 (point)) ; save cursor pos
> XL>     (search-forward-regexp "theq() :") ; ending pattern
> XL>     (backward-char 8)
> XL>     (setq p2 (point)) ; save cursor pos
> XL>     (setq mytext (buffer-substring p1 p2))
> XL>     )
> XL>   )
>
> I don't think your first patten is exactly what the OP needed.
>
> You can use (forward-line -1) to move the point back to the previous
> line, and (beginning-of-line -1) to move to the beginning of the
> previous line.  Also, you don't need search-forward-regexp the second
> time, just search-forward will work.  Plus, of course, (backward-char 8)
> is just asking for trouble.
>
> Anyhow, regular expressions can handle multiple lines just fine:
>
> A
> theq() :
> non
> B
> theq() :
>
> (save-excursion
>   (goto-char (point-min))
>   (while (re-search-forward "\\(.*\\)\ntheq() :" nil t)
>     (message (match-string 1))))
>
> will produce "A" and "B"

        A slightly more elaborate version (not necessarily more
correct, but with a few more functions to ponder :-)

;;; -------------------------------------------------
(defun collect-all-before (pattern)
  (interactive "sPattern: ")
  (let (ret)
    (while (re-search-forward pattern nil t)
      (save-excursion
        (if (zerop (forward-line -1))
            (push (buffer-substring (point)
                                  (progn 
                                    (forward-line 1)
                                    (point)))
                  ret))))
    (with-output-to-temp-buffer "*tmp*"
      (set-buffer "*tmp*")
      (dolist (elt (nreverse ret))
        (insert elt)))))
;;; -------------------------------------------------

        (push ...) and (dolist ...) are from a package that tries to
make Emacs-Lisp a bit more like Common Lisp, and can be trivially
replaced with native Emacs-Lisp constructs.