Odd "Use of uninitialized value", regular expression error - perl

Question

I'm using Strawberry Perl 5 on Windows 10. It does seem like my regular expressions are broken or regex101 won't tell me the truth. I want to catch 'num km'. Even tho my array seems to be the right length it'd often say"Use of uninitialized value".
my $string = "^ˇ~ --_ 12 km aéeklwa 32 km | \|ġ^ 0 km 23-24 km";
if (#szelmatches = $string =~ /\d+(\-\d+)?\s+km/gm) {
my $number_of_elements = scalar(#szelmatches);
print "Elements in the array : $number_of_elements \n";
}
foreach (#szelmatches) {
print "$_\n";
}
OUTPUT:
Elements in the array : 4
Use of uninitialized value $_ in concatenation (.) or string at C:\misc\perlek\wttr\szel.pl line 16.
I've ran defined() checks but it seems like my array elements are all defined. Changing \- to .{1} occasionaly worked but it is quite annoying to write like this. regex101.com and regexr.com tells me everything is allright.
I know you could write it simpler/shorter/better/faster/nicer etc, but i honestly think this should work. Do you guys have any idea what am i doing wrong?

Answer 1

Firstly, I had to fix a syntax error in your code before I could run it (the closing ) was missing from your if statement). Please cut and paste code, rather than retyping it.
If Perl tells you that it's finding undefs then it's almost certainly right. Using Data::Dumper can show us what is going on.
use warnings;
use Data::Dumper;
my $string = "^ˇ~ --_ 12 km aéeklwa 32 km | \|ġ^ 0 km 23-24 km";
if (#szelmatches = $string =~ /\d+(\-\d+)?\s+km/gm) {
my $number_of_elements = scalar(#szelmatches);
print "Elements in the array : $number_of_elements \n";
}
print Dumper \#szelmatches;
foreach (#szelmatches) {
print "$_\n";
}
That gives us the following:
$VAR1 = [
undef,
undef,
undef,
'-24'
];
So, yes, there are three undefs in your results. Can we work out why?
Well, here's your match operator.
/\d+(\-\d+)?\s+km/gm
It's looking for digits followed by an optional dash and more digits. But it's only that optional part that you're capturing (as it has parentheses around it). And in the first three cases, that optional section doesn't appear. So you get undef for those first three matches.
Let's actually match what you want (the whole digits section, I think) by putting more parentheses around the whole thing.
/(\d+(\-\d+)?)\s+km/gm
Now we get this result:
$VAR1 = [
'12',
undef,
'32',
undef,
'0',
undef,
'23-24',
'-24'
];
That's better. We get all of the matches we want, alongside the original ones. So, that's twice as many matches as we want. That's because we now have two sets of parentheses for each match. We need the first set to match and capture the digit section and the second set to join together the "-" and the "\d+". But we don't need the second set to capture its contents.
If you read the section on "Extended Patterns" in the perlre manual page, you'll see that we can create non-capturing parentheses with (?:...). So let's use that.
/(\d+(?:\-\d+)?)\s+km/gm
And that gives us:
$VAR1 = [
'12',
'32',
'0',
'23-24'
];
Which is, I think, what you wanted.
Update: Re-reading your question, I realise that you wanted the 'km' as well. So I've moved the closing parentheses past that.
/(\d+(?:\-\d+)?\s+km)/gm
And that gives us:
$VAR1 = [
'12 km',
'32 km',
'0 km',
'23-24 km'
];

Answer 2

The warning you see is because $_ is undefined. In Perl, you can have variables that have no value at all. That's undef.
The first thing you want to do in this case is inspect your array. The core Data::Dumper module is good for that. Or you can install Data::Printer from CPAN, which I prefer.
print Dumper \#szelmatches;
foreach (#szelmatches) {
print "$_\n";
}
This will output
$VAR1 = [
undef,
undef,
undef,
'-24'
];
Clearly there are some undefs in the array. This is because you have a capture group (\-\d) that is optional ?. Each time the string gets matched successfully via the /g modifier, it will put all of the capture group results into your array. But the only group you have is optional, so the pattern matches even if there is no -\d going on.
You can visualise this on Debugex. If you want to have a more detailed play-around with it, try the Regexp::Debugger module, which will allow you to step-by-step debug your regex right in your terminal.
You will have to tell us which numbers you actually want to capture.
If all you are after is the second one after the dash (which you do not have to escape, it has no special meaning), then you should not make that capture group optional.

Answer 3

Two problems.
When a capture is conditional (e.g. (...)?) and it doesn't match anything, it captures undef.
When there's one or more captures, the match returns the capture text rather than the entire text matched.
The solution is to remove the useless and problem-causing capture. Replace
if ( my #szelmatches = $string =~ /\d+(\-\d+)?\s+km/g )
with
if ( my #szelmatches = $string =~ /\d+(?:\-\d+)?\s+km/g )