What's wrong with this perl matching pattern?

For a svn precommit hook, I want to strip leading spaces (there are surely no tabs present) on lines with certain keywords in them before committing them to the server. For example the line

    MACRO_1 (ABC, "Some String");

shall be matched and the leading spaces before MACRO_1 shall be removed. Currently, I have the following expression to match:

   if($line =~ /^\s+MACRO_1|MACRO_2|MACRO_3|MACRO_4.*/) {
     print "Stripping leading space on line $line\n";
     $line =~ s/^\s*//gsxm;   # strip leading spaces
   }

When I look at the console, I开发者_开发知识库 get the following output:

Stripping leading space on line MACRO_1;
Stripping leading space on line MACRO_2;
Stripping leading space on line     MACRO_3 (ABC, "Some String");
Stripping leading space on line  MACRO_1;
Stripping leading space on line MACRO_2;
Stripping leading space on line MACRO_1(123);
Stripping leading space on line MACRO_2(123);
Stripping leading space on line             MACRO_1;
Stripping leading space on line MACRO_2;
Stripping leading space on line MACRO_1;

This seems a bit confusing to me since there are lines with no leading space that I thought would not be matched by \s+ which is in my opinion one or more spaces.

What's the problem?

The | has the lowest precedence among the regex operators.

So the following is an incorrect regex to match a string containing either only foo or only bar:

^foo|bar$

because it'll be treated as:

(^foo)|(bar$)

which matches any string beginning with foo or ending in bar; so it matches for example foo1 and 1bar.

The correct regex would be ^(?:foo|bar)$

Similarly in your case the correct regex is:

if($line =~ /^\s+(?:MACRO_1|MACRO_2|MACRO_3|MACRO_4).*/) {

Also note that there is no need of the trailing .*.

You can shorten your regex as:

if($line =~ /^\s+MACRO_[1-4]/) {