Question

I have a file of data fields, which may contain comments, like below:

id, data, data, data
101 a, b, c
102 d, e, f
103 g, h, i // has to do with 101 a, b, c
104 j, k, l
//105 m, n, o
// 106 p, q, r

As you can see in the first comment above, there are direct references to a matching pattern. Now, I want to capture 103 and it’s three data fields, but I don’t want to capture what’s in the comments.

I’ve tried negative lookbehind to exclude 105 and 106, but I can’t come up with a regex to capture both.

(?<!//)(\b\d+\b),\s(data),\s(data),\s(data)

This will capture all but exclude capture of 105, but to specify

(?<!//\s*) or (?<!//.*)

as my attempt to exclude a comment with any whitespace or any characters invalidates my entire regex.

I have a feeling I need a crafty use of an anchor, or I need to wrap what I want in a capture group and make a reference to it (like with $1) in my lookbehind.

---

If this is another case of “regular expressions don’t support recursion” because it’s a regular language (a la automata theory), please point that out.

Is it possible to exclude the comments in 103, and lines 105 and 106, using a regular expression? If so, how?

Answer 1

The easy way out is to replace \s*//.* with the empty string before you begin.

This will remove all the (single-line) comments from your input and you can go on with a simple expression to match what actually you want.

The alternative would be to use look-ahead instead of look-behind:

^(?!//)(\b\d+\b),\s(data),\s(data),\s(data)

In your case it would even work to just anchor the regex because it is clear that the first thing on a line must be a digit:

^(\b\d+\b),\s(data),\s(data),\s(data)

Some regex engines (the one in .NET, for example), support variable-length look-behinds, your’s does not seem to be capable of this, this is why (?<!//\s*) fails for you.