From the same document a few paragraphs up from the General Examples section:

Open returns nonzero on success, the undefined value otherwise. If the open involved a pipe, the return value happens to be the pid of the subprocess.

Even if the command isn't on your system the pipe is opening a subprocess and returning the pid. Here's what I see on Windows 8.

use warnings;
use strict;

print "here it goes...\n";
my $return_val = open(my $article, "-|", "caesar <article");
#    or die "Can't start caesar: $!";  
print "tried it and got [$return_val]\n";
[download]

c:\usr\pm>openprog.pl > openprog.txt 2>&1
here it goes...
The system cannot find the file specified.
tried it and got [14048]
[download]

use warnings;
use strict;
use Win32::Process::List;

$!=0;
print "here it goes...<$!>\n";
my $return_val = open(my $article, "-|", "caesar <article");
#    or die "Can't start caesar: $!";  
print "tried it and got [$return_val] <$!>\n";

print "="x75,"\n";
my %list = Win32::Process::List->new()->GetProcesses();        #return
+s the hashes with PID and process name
foreach my $key ( keys %list ) {
    # $list{$key} is now the process name and $key is the PID
    print sprintf("%30s has PID %15s", $list{$key}, $key) . "\n" if $k
+ey == $return_val;
}
print "="x75,"\n";
[download]

here it goes...<>
The system cannot find the file specified.
tried it and got [8664] <Inappropriate I/O control operation>
======================================================================
+=====
                       cmd.exe has PID            8664
======================================================================
+=====
[download]

Sorry, I feel a bit dense here, but what do I do with the PID once I have it? I get a PID regardless of whether or not the command succeeded.

pipe.pl

#!/usr/bin/perl
use 5.018;
use strict;
use warnings;

my $pid = open (my $ARTICLE, "-|", "caesar") or die "Can't start caesa
+r: $!\n$^E";

my $read = <$ARTICLE>;
say "[$read][$pid]";
[download]

Results:

perl pipe.pl
'caesar' is not recognized as an internal or external command,
operable program or batch file.
Use of uninitialized value $read in concatenation (.) or string at pip
+e.pl line 9.
[][1236]
[download]

You asked in the OP what is different about your command and why isn't the 'or die' clause working in your open command. I provided documentation that shows if you open a pipe it returns the pid instead of just non-zero or zero like with opening a file. That is the difference and the answer to your question.

I wasn't telling you to do anything with the pid, only that it is the return value and the reason why your code didn't work as expected.

I don't think this example should be included in the documentation for open since it doesn't work as expected for pipes.

The problem is that perl is starting cmd.exe that is then asked to run the non-existent command. So the pid is for the shell instance. (Watch the taskmanager to see this is happen.)

To avoid that problem you can try passing the fully qualified pathname of the command you want to run.

Without the command being fully qualified, 'caeser' might be caeser.exe or caeser.bat or caeser.cmd or caeser.pl or caeser.vb etc. And it might be in the current directory, or somewhere in the path or.... Rather than Perl having to emulate all of the possibilities, unqualified commands are passed to the shell to do what it does.

Thank you. That certainly makes sense. I feel though I should probably back up and clarify the root problem though.

I am working with a collection of PDF files and am using the pdftohtml.exe program to convert the PDF into an XML stream in order to extract text of interest with XML::Twig:

open (my $XML, "-|", "e:\\path\\to\\pdftohtml.exe -xml -zoom 1.4 -stdo
+ut $PDF_FILE") or die "pdftohtml failed:\n$!\n$^E";

my $t = XML::Twig->new(
    twig_handlers => {
        '/pdf2html/pagetext[(@top >= 180 and @top <= 190) and (@left >
+= 100 and @left <= 111)]' => \&RouteTo,
        '/pdf2html/pagetext[(@top >= 215 and @top <= 225) and (@left >
+= 260 and @left <= 270)]' => \&InvoiceSort,
        '/pdf2html/page' => sub { $_[0]->purge; 1; }, # free memory af
+ter every page
    },
    comments   => 'drop',  # remove any comments
    empty_tags => 'normal',# empty tags = <tag/>
);

$t->parse($XML);

close $XML;
[download]

The problem is that if I fat-finger the open command (e.g. type "-zom" instead "-zoom" in the command arguments), or if "$PDF_FILE" could not be found, the program merrily continues on its way, unaware that $XML is undefined. I've been working around this by wrapping the "$t->parse" in an eval block to catch this, but I was wondering if there was a better way.

1. The problem is that if I fat-finger the open command (e.g. type "-zom" instead "-zoom" in the command arguments)

   You ought to detect that kind of error the first time you test your script; so correct the typo.

2. or if "$PDF_FILE" could not be found, the program merrily continues on its way, unaware that $XML is undefined.

   This kind of depends on what the executable does in that situation. I'll assume it does the sensible thing of outputting an error message then exits with a non zero exit code.

   Normally, if you were reading the pipe yourself, the first time you attempted to read it would get a end of file (with a pipe abandoned status) and you could then call waitpid on the pid returned by the open, and check $? to obtain the exit code and status.

   As you are passing the filehandle into a module, the simplest check would be to call eof on the filehandle before you give it to XML::Twig; and if there's nothing to read, don't pass it on; just waitpid and check $?

It can get more complicate if the executable is one of those that tries to be 'helpful' and hangs around rather than just exiting on error; but let's assume it's not :)

---

With the rise and rise of 'Social' network sites: 'Computers are making people easier to use everyday'

Examine what is said, not who speaks -- Silence betokens consent -- Love the truth but pardon error.

"Science is about questioning the status quo. Questioning authority". I knew I was on the right track :)

In the absence of evidence, opinion is indistinguishable from prejudice.

...eval block to catch this, but I was wondering if there was a better way.

From the XML::Twig docs:

safe_parse ( SOURCE [, OPT => OPT_VALUE ...])


This method is similar to parse except that it wraps the parsing in an eval block. It returns the twig on success and 0 on failure (the twig object also contains the parsed twig). $@ contains the error message on failure.

Note that the parsing still stops as soon as an error is detected, there is no way to keep going after an error.

        “The sources of quotes found on the internet are not always reliable.” — Abraham Lincoln.3; cf.

If I'm understanding correctly, you're basically wanting to call another program from your Perl code and capture the STDOUT and STDERR of that program so that your code can determine if the program ran successfully or encountered errors. Is that correct?

If the description above is correct, then my approach would be to leverage Capture::Tiny instead of using the piped open construct.

Can't solve your problem, but for what it's worth it does fail properly on my OS X system. Interestingly though, if I call it like you did with an argument to the non-existent program, *that* error is raised first:

[502]► perl -E'
open ( FOO, "-|", "caesar" ) or die $!;
'
No such file or directory at -e line 2.

[503]► perl -E'
open ( FOO, "-|", "caesar <article" ) or die $!;
'
sh: article: No such file or directory
[download]

I got similar results: The shell throws an error, but it never makes it to Perl

Pipe.pl

#!/usr/bin/perl
use 5.018;
use strict;
use warnings;

open (my $ARTICLE, "-|", "caesar") or die "Can't start caesar: $!\n$^E
+";

my $read = <$ARTICLE>;
say "[$read]";
[download]

Result:

perl pipe.pl
'caesar' is not recognized as an internal or external command,
operable program or batch file.
Use of uninitialized value $read in concatenation (.) or string at pip
+e.pl line 9.
[]
[download]

open (my $ARTICLE, "-|", $shell_command)
[download]

open (my $ARTICLE, "-|", "cmd", "/x", "/c", $shell_command)
[download]

open successfully executed cmd as requested. No error occurred, so it returned the pid of the child.

Presumably, cmd will return an error for being unable to execute ceasar. Of this you will be notified when you close the handle. It will return false with a true value for $? >> 8.