Question

There are two forms of Perl filename wildcard command: <> and glob. But I found there is difference between the effect of these two forms:

I want to retrieve all the files with similar names, using following code:

my @files = <"rawdata/*_${term}_*.csv">;   #(1)

and another format:

my @files = glob "rawdata/*_${term}_*.csv";  #(2)

I expect to get the same result using these two codes. But there is difference: if the $term is a string without spaces (or to say, one word), then (2) works well, but (1) doesn’t work; if the $term is a string with spaces (or to say, several words), then (1) works well, (2) doesn’t work.

Is there any difference between these two expressions?
Thanks a lot.

Answer 1

<SomeStuff> is equivalent to glob "SomeStuff" (apart from all the ambiguities with <> also being used for reading from file handles — see perldoc perlop and look for I/O Operators there). Therefore your examples aren’t equivalent. You should use

my @files = glob "\"rawdata/*_${term}_*.csv\"";  #(2)

instead.

However, as to why space in the pattern makes a difference: perldoc -f glob tells the story. The normal glob (and therefore <> which is implemented via glob) treat whitespace as a pattern separator. The documentation also mentions File::Glob and its function bsd_glob which does not treat spaces as pattern separators. Therefore consider using this instead:

use File::Glob ':glob';

my $term1  = "some stuff";
my @files1 = glob "dir/${term1}*";
my $term2  = "more";
my @files2 = glob "dir/${term2}*";

print join(' :: ', sort @files1), "\n", join(' :: ', sort @files2), "\n";

Possible output with some files I just created:

[0 mosu@tionne ~/tmp] ~/test/test1.pl
dir/some stuff is betther than other stuff.doc :: dir/some stuffy teachers.txt
dir/more beer.txt :: dir/more_is_less.csv