Match timestamps in WebVTT files with sed

Question

I have the following PCRE2 regex that works to match and remove timestamp lines in a .webVTT subtitle file (the default for YouTube):

^[0-9].:[0-9].:[0-9].+$

This changes this:

00:00:00.126 --> 00:00:10.058
How are you today?

00:00:10.309 --> 00:00:19.272
Not bad, you?

00:00:19.559 --> 00:00:29.365
Been better.

To this:

How are you today?

Not bad, you?

Been better.

How would I convert this PCRE2 regex to an idiomatic (read: sane-looking) equivalent for sed's flavour of regex?

Answer 1

Using your regex with sed

$ sed -En '/^[0-9].:[0-9].:[0-9].+$/!p' file
How are you today?

Not bad, you?

Been better.

Or, do not match lines that end with an integer

$ sed  -n '/[0-9]$/!p' file
How are you today?

Not bad, you?

Been better.

Answer 2

Your pattern is not a specific PCRE2 pattern, only using sed you have to escape the \+ to make it a quantifier for 1 or more times.

At the positions that you use a dot to match any character (and looking at the example data) there is a digit as well.

You could make the pattern a bit more specific, and omit the quantifier at all. Just prevent the line from printing if the pattern matches.

sed -n '/^[0-9][0-9]:[0-9][0-9]:[0-9]/!p' file

• -n prevents the default printing in sed
• !p prints the line if the pattern does not match

Output

How are you today?

Not bad, you?

Been better.