Rich

If I can come up with so many different interpretations of "is this string an integer" ((tye)Re: Integer detection concept), just think how many interpretations of "is this string a number" covers.

First you need to decide which of the many meanings you are dealing with and then you can decide which of probably several ways you can try to get that answer. If someone offers to tell you how to determine is a string "is a number" without forcing you to be a lot more specific, then are likely answering the wrong question for you. (:

Though, I do think that Perl's internal looks_like_number() might be the right answer for several common meanings of that question. I really need to find the time to get an interface to that available from CPAN and included in a future version of the core. /:

if ($var && $var !~ /\D/) {
    # We have a value and it's a number
}
[download]

So, you can code up a neato little function, sorta like:

sub isNum {
    if ($var && $var !~ /\D/) {
        return 1;
    } else {
        return 0;
    }
}
[download]

(Yes, I'm sure someone's already done this, but this was too cool an opportunity to promote module use. *blush*)

------
We are the carpenters and bricklayers of the Information Age.

Vote paco for President!

Unfortunatly checking for non-digits is not sufficient since these are all numbers:
• 12.7
• 1E0
• 1.477e7

Checking just for non-digits won't work on those. And you cannot just check for digits, E, e, and ., since these are all not numbers

• ...
• Eeeeeee.
• 12e
• e15
• 1.6.8

The only thing that your technique will work for is integers -- certainly not the only thing meant by 'number'.

that's easy to fix (untested):
qr/(?:+|-)?\d+(?:\.)?(?:\d+)?(?:[Ee](?:+|-)?\d+)/

piece by piece:


(?:+|-)?            # Optional leading + or -
\d+                 # One or more digit
(?:\.)?             # A trailing . 12. is a number I think
(?:\d+)?            # The part after the decimal
(?:[Ee](?:+|-)?\d+) # E or e followed by an optional + or - and one or
+ more digit
[download]

I don't think I left anything out. Now, this won't match .4 as a number.... You can build a regex that will handle that case properly as well. But the simple fix of making the part before the decimal optional causes another bug, it maks every part optional, so the null string would be a match, and since every string starts with a null string, every string would match. Since all we have is a number, we could anchor to the begenning and end of the value, but then, the empty string would still match. Thinking back to his origonial question, this should be anchored to the front and back, so more like qr/^...$/.
-Ted

# numeric -- returns true if arg. is numeric
sub numeric {
  my $value = shift;
  my $result = 1;
  local $^W = 1; # turn warnings on if they aren't already.
  local $SIG{__WARN__} = sub { $result = 0 };
     # if we get a warning make $result false
  $value += 0;
  return $result;
}
[download]

A string that can't be parsed as a number at all will be zero. So, how about something like this:

if ($x==0 && substr($x,0,1) ne '0') {
   # deal with non-number
   }
else {
   # deal with number
   }
[download]

—John

Update:fixed typo in substr, reported by dga. Yea, I typed it without checking perlfunc, and I rarely use substr in real code.

Yes. That (together with a || $x==1 && substr($x,1,1) ne '1') does exactly what I find myself wanting nine times out of ten. That is, what isNaN does in js. I'd find it useful to have a built in function, and maybe we're going to get one.

§ George Sherston

Hmm, when does a scalar convert to 1 when it's not a number?

Note that I didn't mention that you should already know that the value is "a string" rather than a reference or funky blessed overloaded object, but carefully worded it because I had that in mind.

I think you have a typo in your substr which should be

substr($x,0,1);
[download]

Otherwise the case $x=0; falls into the not a number part of the if.

Question, does Perl's SIGNAL tie to this exception?

hsm

Nobody has mentioned the Perl Cookbook yet. Recipe 2.1 discusses this topic, I quote here the regex for a C float (does not take IEEE 'NaN' and 'Infinity' values into account) and another suggested way using a wrapper around the POSIX function strtod:

warn "not a C float"
    unless /^([+-]?)(?=\d|\.\d)\d*(\.\d*)?([Ee]([+-]?\d+))?$/;

sub getnum {
    use POSIX qw(strtod);
    my $str = shift;
    $str =~ s/^\s+//;
    $str =~ s/\s+$//;
    $! = 0;
    my($num, $unparsed) = strtod($str);
    if (($str eq '') || ($unparsed != 0) || $!) {
        return;
    } else {
        return $num;
    } 
}
[download]

-- Hofmator

Use an eval - see the node on How to identify a number datatype in a string?

Without a hook to the internal API, the only real solution is to try to use it as a number and see if Perl complains.

To Quote:

The special IEEE notions of Infinity and NaN (not a number) will be pr
+operly honored: 

  $n = 'NaN';
  print 2 * $n;
NaN
  print 10 * $n
NaN
  print 1 + NaN
NaN
  $i = 'Infinity' 
  print 1 + $i
Infinity
  print $i * $i
Infinity
  print $i - $i
NaN
  print 'Infinity' < 0
0
  print 'Infinity' > 0
1
[download]

Edit by tye to change very wide BLOCKQUOTE+PRE to CODE

Interesting, it looks like this works, but they return 'inf' and 'nan' instead of the longer capitalized versions above.

% perl -le "print 1 + 'NaN'"
nan

% perl -le "print 1 + 'Infinity'"
inf
[download]

-Blake