I accidentally deleted an important python script, and so I ran the command

sudo grep --binary-files=text --context=100 'unique string' /dev/sda1 > recover_file

to search for it on my hard drive and save matches to ./recover_file. When I open ./recover_file in Vi ("Vi Unimproved", not Vim) I see that it is ~10800 lines long and contains many versions of my ~200 line file, with some junk between each occurrence, as expected. But there are also hundreds of strange lines with unexpected behavior that I will attempt to describe.

I have line numbers on. If line 19 is the first strange line in the file, upon opening the file I get a message a the bottom of the window saying

Conversion error on line 19

Initially the strange lines appear as empty lines, like the lines displayed at the bottom of a document when there are no more lines in the file to be displayed, with a ~ character at the far left edge of the window, but located between two other lines, not at the end of the file:

    18 junk junk junk
    20 junk junk junk

When I attempt to delete line 19 using dd, nothing happens. If I delete a normal line, then line 19's appearance changes and it looks like any other blank line:

    18 junk junk junk
    20 junk junk junk

But as soon as I move my cursor over it, the line number disappears and it looks just like it did before. If I try to perform any operation on it, such as inserting or appending text, I get

Error: unable to retrieve line 19

If I write the file to the disk, I get

Error: recover_file: Invalid or incomplete multibyte or wide character.

Then, if I close and re-open the file, I see that all lines from 19 onward have been removed, leaving only lines 1-18. I was able to reproduce the situation and copy a recent version of the python file into a new file, after which further digging in ./recover_file produced a segmentation fault and the entire file was lost.


1) For future reference, is there a way I can remove these strange lines so that I can save the file directly without losing important data, or will I always need to highlight and copy from the terminal window?

2) I assume that this behavior is due to the presence of binary code in ./recover_file not corresponding to text characters, which Vi cannot render. If someone could confirm/correct this impression and perhaps provide further explanation, I would be grateful.


I'm not sure if this is relevant, but I'm running lubuntu 18.04 as a virtual machine on VMware Workstation 14 Player.

  • 1
    I would have added | strings at the end of the command, before redirecting to get rid of junk chars. Nonethess, with your commands, you can also get extraneous EOF chars, I think. – Rui F Ribeiro Jul 19 '18 at 17:20
  • 2
    ... or if applicable use extundelete as described here Accidentally deleted a folder – steeldriver Jul 19 '18 at 17:45
  • @RuiFRibeiro putting | strings after the grep command and before the redirect > worked pretty well. If you want to post it as an answer I'll select it. If not, I'll post it in a day or two. – The Ledge Jul 19 '18 at 23:32
  • I will post it and explain why it works in a couple of hours, preparing to go to work now. – Rui F Ribeiro Jul 20 '18 at 6:02
  • 1
    If /dev/sda1 was the disk that contains the filesystem where you wrote recover_file you could easily have overwritten your deleted data with your "recovering" data. This still might be case. This isn't a good way to recover a lost file - but good luck. – roaima Jul 20 '18 at 10:46

From looking at your script, you are dumping and try to edit, searching and line editing, binary files with the vi text editor.

In that way, you will encounter a lot of control characters that will subvert the notion of lines, length of lines and in some stituations, even possibly the end-of-file.

Since you are only interested in text, and you are already somewhat parsing the disk contents, I would add a strings command to it to discard non-text characters.

To be able to handle your output in vi, you may thus change your script to:

sudo grep --binary-files=text --context=100 'unique string' /dev/sda1 | strings > recover_file

I also suspect it will be more efficient to discard those control chars to begin with as in:

sudo strings /dev/sda1 | grep --context=100 'unique string' > recover_file

Though I am not entirely sure this last instruction with give the same results due to being handled as text and not binary.

From man strings

strings - print the strings of printable characters in files.


For each file given, GNU strings prints the printable character sequences that are at least 4 characters long (or the number given with the options below) and are followed by an unprintable character. By default, it only prints the strings from the initialized and loaded sections of object files; for other types of files, it prints the strings from the whole file.

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.