How keep only duplicate lines?

Question

Emacs 26.1. Windows 10

In buffer I has text like this:

    3333333
    1111111
    $PAC_LTC
    1337_BTC
    1337_LTC
    404_BTC
    42_BTC
    222222
    42_LTC
    808_ETH
    8BIT_BTC
    8BIT_BTC
    8BIT_BTC
    8BIT_DOGE
    8BIT_ETH
    8BIT_LTC
    ACC_DOGE
    ACC_USDT
    ACC_USDT
    ADC_DOGE
    ADC_DOGE
    ADC_DOGE
    ADC_ETH
    ADC_USDT
    8BIT_USDT
    8BIT_USDT
    42_ETH

I need to keep only duplicate lines. The result must be like this:

8BIT_BTC
8BIT_BTC
8BIT_BTC
8BIT_USDT
8BIT_USDT
ACC_USDT
ACC_USDT
ADC_DOGE
ADC_DOGE
ADC_DOGE

P.S. I try by this command: delete-non-matching-lines

But after press Enter whole text delete from buffer.

Answer 1

Here's a command that finds all duplicate lines in the buffer and deletes the rest:

(defun keep-duplicate-lines ()
  (interactive)
  (let (lines dups)
    (save-excursion
      (goto-char (point-max))
      (when (/= (char-after (1- (point-max))) ?\n)
        (newline))
      (goto-char (point-min))
      (while (not (eobp))
        (forward-line 1)
        (push (buffer-substring-no-properties (line-beginning-position)
                                              (line-end-position))
              lines))
      (dolist (line lines)
        (when (and (> (cl-count line lines :test 'equal) 1)
              (not (string= "" line)))
          (push (regexp-quote line) dups)))
      (goto-char (point-min))
      (keep-lines (mapconcat #'identity
                  dups
                  "\\|")))))

Answer 2

The following answer addresses the original version of the question "How keep only duplicate lines?". It shows the duplicates in an auxiliary buffer and keeps the original buffer unmodified.

1. Copy the whole buffer with C-x h M-w.
2. Create a new buffer with C-x b *work* RET. (Replace *work* with a buffer name to your likings.)
3. Yank the text there with C-y
4. Mark the whole buffer *work* by C-x h and sort its lines with M-x sort-lines RET
5. Mark the whole buffer by C-x h and delete all lines that do not occur multiple times in sequence with M-x delete-non-matching-lines RET ^\(.*\)\( C-q C-j \1$\)+ RET.

Note: The sequence C-q C-j inserts a verbatim newline in the minibuffer.
Only the gray sections ^\(.*\)\( and \1$\)+ are intended for copy-paste. You must type the part C-q C-j in the middle of the regexp as key sequence with the Ctrl button pressed.

---

If you want to modify the buffer such that you only keep the replicated lines you can skip steps 1, 2, and 3. Furthermore, use the original buffer instead of *work* in step 4. I.e.:

1. Mark the whole buffer by C-x h and sort its lines with M-x sort-lines RET
2. Mark the whole buffer by C-x h and delete all lines that do not occur multiple times in sequence with M-x delete-non-matching-lines RET ^\(.*\)\( C-q C-j \1$\)+ RET.