Delete all Bash comments

Question

How can I match and delete all comments from the line? I can delete comments starting from new line, or the ones not in quotes using sed. But my script fails in the following examples

This one "# this is not a comment" # but this "is a comment"

Can sed handle this case? if yes what is the regex?

Example:

Input:

This one "# this is not a comment" # but this "is a comment"

Output:
```
This one "# this is not a comment"
```

A comment is a word (obeying bash's convoluted quoting rules) which starts with an unquoted #. That's very tricky to recognize with a regular expression, because determining the range of a quoted expression can only be done with an understanding of bash syntax. And the syntax is recursive: `cmd "$(other_cmd #comment` is legal (though not complete). — rici, Oct 26 '14 at 18:03
I know it is tricky. Spend lots of time trying to figure it out... It means there is no simple solution using sed? — Inventor, Oct 26 '14 at 18:05
There are no simple solutions using `sed`. Not even complicated ones. Perhaps there are insanely complicated ones, though, but nobody even bothered to think about it. — gniourf_gniourf, Oct 26 '14 at 18:09
Thank you very much. Maybe you could recommend some other tools I can use in bash? — Inventor, Oct 26 '14 at 18:10
@user2590816: Can you tell us why you want to strip that comments? Are you parsing a configuration file? — firegurafiku, Oct 26 '14 at 18:38
possible duplicate of [how to remove comments from a bash script](http://stackoverflow.com/questions/25291228/how-to-remove-comments-from-a-bash-script) — rici, Oct 26 '14 at 19:38
@user2590816: The best tool is often bash itself :) See my answer to the question I marked as duplicat. — rici, Oct 26 '14 at 19:40
Many people in this forum are trying to use `sed`,`grep`,`awk` as an interpreter, compiler, parser of XML, CSV or some other format. They are just not made for this task! They can work in some cases but almost every time you can find a little bit more complex input when the solution will fail. — Vytenis Bivainis, Oct 26 '14 at 21:13
@Vytenis Can you suggest some tools that are made for this task? — Inventor, Oct 26 '14 at 21:15

score 1 · Answer 1 · edited Nov 19 '17 at 20:36

1

You can use a lexical analyzer like Flex directly applied to the script. In its manual you can find "How can I match C-style comments?" and I think that you can adapt that part to your problem.

If you need an in-depth tutorial, you can find it here; under "Lexical Analysis" section you can find a pdf that introduce you to the tool and an archive with some practical examples, including "c99-comment-eater", which you can draw inspiration from.

edited Nov 19 '17 at 20:36

rici

234,347
28
237
341

answered Oct 26 '14 at 19:45

Alberto Coletta

1,563
2
15
24

thanks for information. I ended up writing C-style code which traces each character on the line and stops as soon as it sees '#' outside double quotes. Don't think it is efficient solution, but at least it works – Inventor Oct 26 '14 at 20:01

score 1 · Accepted Answer · answered Oct 28 '14 at 18:46

If we assume that # is not a comment when it is in quotes or escaped with backslash, then we can define the following regex:

(ES|RT|QT)*C?

where

ES - escape sequence: \ followed by 1 char

\\.

RT - non-special regular text

[^"\\#]*

QT - text in quotes

"[^"]*"

C - comment starting with unescaped, unquoted hash sign # and ending with the end of line

#.*

The possible solution using sed:

sed 's/^\(\(\\.\|[^"\\#]*\|"[^"]*"\)*\)#.*$/\1/'

Delete all Bash comments

2 Answers2