Question

I am using the regexec() function in C. I basically am trying to write a regular expression to capture portions of a string for substitution.

So for example, if I have the string “Hello $X” Then I want the regexec to give me the range 6,7 as that is “$X”. But as there can be an arbitrary number of substitutions, I am using the regular expression:

"([^$]*(\\$[A-Za-z][A-Za-z0-9_]*))+"

This should match any arbitrary sequence of text + substitution patterns.

So for example in the string “First=$X, Second=$Y” I need to know that $X occurred at offset 6-7 and and $Y occurred at offset 17-18.

The actual offsets I get from regexec are:
0,19 8,19 17,19

First, I understand that the ending offset is actually one past the the character of the match. So the above offsets correspond to the following parts of the string:

First=$X, Second=$Y
, Second=$Y
$Y

Now I can see what is happening here: the first range is obviously the entire match, and the second is the first entire sub-match of the second sub-expression. But from this point on I am puzzled. Why is it only returning the first sub-match of the second sub-expression and not the first?

I suspect it has something to do with the fact that I have a repeating expression, but I’m not sure what I need to do to fix the problem. How do I get it to return the desired offsets?

Note: I am passing a 128-element regmatch_t to regexec() (nmatch=128), so I should be able to get all matches.

Answer 1

You’re confused about what first and second mean. In this expression:

"([^$]*(\\$[A-Za-z][A-Za-z0-9_]*))+"
 ^_______________________________^    this part

is the first parenthesizes subexpression and

"([^$]*(\\$[A-Za-z][A-Za-z0-9_]*))+"
       ^________________________^    this part

is the second. If a parenthesized subexpression gets used more than once as part of a *, ?, +, or {} repetition operator, it’s the last match that counts.

If you want to match an arbitrary number of instances, than rather than using the + on the end of your regex, you simply need to call regexec multiple times, and use the ending offset of the previous run as your new starting point.