Strange result on perl regexp - end string anchor & ungreedy at once.

Question

I have a very simple substitution:

my $s = "<a>test</a> <a>test</a>";
$s =~ s{ <a> .+? </a> $ }{WHAT}x;

print "$s\n";

that prints:

WHAT

But I was expecting:

<a>test</a> WHAT

What do I misunderstand about "end string anchor" in interaction with ungreedy option?

---

So, I was wrong about regexp engine. Indeed, dont humanize code - it doing rightly what you wrote, not you "think do".

Its just find first <a>, then find </a>$. First lockup are positive, pattern matched.

Right pattern must be something about:

$s =~ s{ <a> (?! .* <a> ) .* </a> }{WHAT}x;

thats give me correctly

<a>test</a> WHAT

because now I really asked regexp for last <a>.

I think its less efficient [^<]+, but more flexible.

Answer 1

The non-greedy modifier (and regexes in general) works from left-to-right, so in essence what is happening here is that it tries to find the shortest string that matches after the first <a> until the next </a> that is at the end of the string.

This does what you would expect:

my $s="<a>test</a> <a>test</a>";
$s =~ s#<a>[^<>]+</a>$#WHAT#;

print "$s\n";

What is the problem you're trying to solve?

Answer 2

This is one of the reasons you don't use a regex to match HTML. Try using a parser instead. See this question and its answers for more reasons not use a regex, and this question and its answers for examples of how to use an HTML parser.