Delete duplicate lines, with partial match

Question

Sample text:

This is first line.a
This is over_second line.
This is over_fourth line.
This is third line.
This is over_fourth delete it.
This is over_fourth and one more.
This is over_second with another text.

Need to delete lines where partial match occur, i.e if over_second occurs in another line, then whole line should get deleted. So output will be as follows;

This is first line.
This is over_second line.
This is third line.
This is over_fourth line.

I could only come up with: over_\w+ for selecting section of text. But don't know how to recognize duplicates, and delete whole line.

Is the requirement "for each specific match for the regexp, retain only the first line in the buffer containing that text" ? — phils, Nov 29 '18 at 22:11
Yes. Line with first instance of match should be preserved, while subsequent matches or remaining matches should get deleted. — msinfo, Nov 29 '18 at 22:14
Once this process gets complete for over_second, same should be repeated for over_fourth. — msinfo, Nov 29 '18 at 22:16

score 2 · Answer 1 · answered Nov 29 '18 at 22:42

2

Try delete-duplicate-lines, which is part of distributed Emacs.
Emacs Wiki page Duplicate Lines might help.
- It points to a blog post about it.
- It explains why interactive search-and-replace might not help.
- It explains how to do it with Lisp, in various ways.
- It explains how to do it with the UNIX / GNU/Linux command sort or unique.

answered Nov 29 '18 at 22:42

Drew

75,699
9
109
225

These are good suggestions, but OP was asking about deleting *partial duplicates*, i.e., lines that contain the same string, but aren't full duplicates of each other. I don't think delete-duplicate-lines does that? – Tyler Nov 29 '18 at 22:48
@Tyler: Oh, right. Maybe the other stuff on that page will help in some way... – Drew Nov 30 '18 at 00:40

score 1 · Answer 2 · answered Nov 29 '18 at 22:42

If you don't need to do this all the time, the quickest solution might be using interactive replacement: query-replace-regexp, bound to C-M-%. Start with point at the top of the buffer.

Note, if you want to delete entire lines, you'll need to include a newline in your regexp. You enter this at the prompt with C-q C-j.

So, for over_second, call C-M-%, then enter the regular expression:

C-qC-j.*over_second.*

This will match an entire line that contains the string over_second, and includes the previous new line.

Then enter the empty string (just type <enter>) for the replacement value.

The first line that matches the regexp should now be highlighted. This is the one you want to keep, so type n to tell Emacs to skip it. The next line will be highlighted. You can delete this by typing y (or <space>).

You can keep typing y until all the matches are deleted, or you can type ! to delete all remaining matches at once.

phils · Answer 3 · 2018-11-30T00:12:29.673

I could swear this is a duplicate, but I couldn't find it.

Try this:

(defun my-delete-duplicate-matches (regexp)
  "Delete matching lines, except the first instance of each specific match."
  (interactive (list (read-regexp "Regexp: ")))
  (save-restriction
    (when (use-region-p)
      (narrow-to-region (region-beginning) (region-end)))
    (save-excursion
      (goto-char (point-min))
      (let ((matches (make-hash-table :test #'equal)))
        (save-match-data
          (while (re-search-forward regexp nil :noerror)
            (if (not (gethash (match-string 0) matches))
                (puthash (match-string 0) t matches)
              (forward-line 0)
              (delete-region (point) (progn (forward-line 1)
                                            (point))))))))))

Caveats:

If the same matching text appears twice on the first line in which it is found, the line will be deleted.
More generally, if a line contains both the first instance of a given match, and a repeat instance of an earlier match, the line will be deleted.
Multi-line patterns are not supported.

Delete duplicate lines, with partial match

3 Answers3