How to add an empty line after lines starting with a timestamp

Question

The log viewing experience is so bad in my current company. I want to add new line to the logs that I've trimmed for a specific time duration. I want a new line after every date-time.

The date-time format is like this.

2023-10-03 15:34:37

I am aware tr can do this but I'm not sure how. I'll select the answer best anyone who can tell me how to write that new to a new file. Because of performance reasons. I don't prefer using vi, I use less as a logs reader.

Answer 1

In vim, you could do:

:%s/\v^\d{4}(-\d{2}){2} \d{2}(:\d{2}){2}.*/&\r

Note the \v for very magic so you don't need to add backslashes before every {}().

Yes, \r in the replacement adds a newline (splits the line in two using the wording at :help s/\r), not a carriage return as one might have expected.

In standard vi (\d, \v, \r, norm being all vim extensions), the equivalent would look like:

:%s/^[[:digit:]]\{4\}\(-[[:digit:]]\{2\}\)\{2\} [[:digit:]]\{2\}\(:[[:digit:]]\{2\}\)\{2\}.*/&^M/

Where the ^M is entered by pressing Ctrl+V followed by Enter (and again inserts a LF aka newline aka ^J, not CR aka carriage-return aka ^M).

But the vi implementation on CentOS is vim (and has been since at least as far back as CentOS 4) so you should be able to rely on the vim extensions in vi there.

In any case tr is a tool to transliterate, replace characters with other characters or delete or squeeze them, it cannot be used here.

sed, awk or perl could be used (AFAIK, \d as a regex operator originated in perl):

perl -pi -e '$_ .= "\n" if /^\d{4}(-\d{2}){2} \d{2}(:\d{2}){2}/' your-file

To edit the file in place. Or:

perl -p -e '$_ .= "\n" if /^\d{4}(-\d{2}){2} \d{2}(:\d{2}){2}/' < your-file > new-file

To get the result into a new file.

Or here, provided the Regex::Common::time module is available:

perl -MRegexp::Common=time -p -e '
  $_ .= "\n" if m{^$RE{time}{strftime}{-pat=>"%Y-%m-%d %H:%M:%S"}}
  ' < your-file > new-file

Which will do a stricter matching (not match on 2023-13-00 25:61:62 for instance). There's also a $RE{time}{iso} which will match those dates, but also many other variants of ISO8601 dates including some without date or time component, so can't be used here.

You can see the actual regex with:

$ perl -MRegexp::Common=time -le 'print $RE{time}{strftime}{-pat=>"%Y-%m-%d %H:%M:%S"}'
(?:(?:\d{4})-(?:(?:(?=[01])(?:0[1-9]|1[012])))-(?:(?:(?=[0123])(?:0[1-9]|[12]\d|3[01]))) (?:(?:(?=[012])(?:[01]\d|2[0123]))):(?:(?:[0-5]\d)):(?:(?:(?=[0-6])(?:[0-5]\d|6[01]))))

Answer 2

This Command:

Press : and enter g/\v\d{4}-\d{2}-\d{2} \d{2}:\d{2}:\d{2}/normal o

Explanation:

• g: short for global, executes a command on all lines that match a regex
• \v: Makes the regex 'very magic' (actual name of it). Means that you don't have to escape special characters
• /^\d\{4\}-\d\{2\}-\d\{2\} \d\{2\}:\d\{2\}:\d\{2\}/: is the regex. Put it into a tool like https://regexr.com/ or https://regex101.com/, they explain it really good
• normal o: normal command executes arguments in normal mode. In that case it presses o which opens a new line

---

---

Read explanations from the Official Documentation: :help global, :help \v, :help normal