Order of output when switching filehandles

Cirollo has asked for the wisdom of the Perl Monks concerning the following question:

I came across a strange problem earlier today: when switching to a new filehandle with select(), sometimes output will go to the new filehandle before it goes to the old one. Here is some relevant code to help explain:

# Create the TEE filehandle, which will print to both STDOUT and to a 
+file
open (TEE, "| tee $name") or die "Teeing off: $!\n";
select(TEE); # Now print goes to the TEE handle by default
$| = 1; # unbuffer TEE

my $stmt = $dbh->prepare("$sql") or die $DBI::errstr; # $sql takes 30-
+40 secs. to run
$stmt->execute() or die $DBI::errstr;

while (my @array = $stmt->fetchrow()) {
    print;
}
$stmt->finish();

select(STDOUT);
&printfooter(); # prints a here-doc, about 10 lines
close(STDOUT) or die "Error closing: $!\n";
[download]

The SQL query that gets printed to the TEE filehandle is pretty huge - it generally takes 35-40 seconds for that block to execute, according to a benchmark. And, it doesn't seem to matter whether the filehandle is buffered. So, can anyone explain why this is happening?

Comment on Order of output when switching filehandles Download Code

Replies are listed 'Best First'.
Re: Order of output when switching filehandles by Shendal (Hermit) on Aug 14, 2000 at 22:44 UTC
Instead of using tee, have you considered just printing directly to the file? I assume you are wanting the results to print to screen as well as to a log file. Why not do something like this: `open(FILE, ">$name") or die "Failed file open: $!\n"; my $stmt = $dbh->prepare("$sql") or die $DBI::errstr; # $sql takes 30- +40 secs. to run $stmt->execute() or die $DBI::errstr; while (my @array = $stmt->fetchrow()) { print; print FILE "$_"; } $stmt->finish(); close(FILE);` [download] I didn't test this, but I'd bet that it's faster than running tee -- not to mention quite a bit easer to customize. Hope that helps, Shendal	[reply] [d/l]
RE: Re: Order of output when switching filehandles by Cirollo (Friar) on Aug 14, 2000 at 22:52 UTC
I ended up using tee because I use more print's than are in this example, and it becomes pretty unwieldy when you have to duplicate them all.	[reply]
(crazyinsomniac) Re: Order of output when switching filehandles by crazyinsomniac (Prior) on Aug 14, 2000 at 22:45 UTC
I'm not sure, but you should close ( TEE ) when you're done with it. "cRaZy is co01, but sometimes cRaZy is cRaZy". - crazyinsomniac	[reply]
RE: Re: Order of output when switching filehandles by Cirollo (Friar) on Aug 14, 2000 at 22:54 UTC
Erk. That seems to have fixed it. Thanks. Just out of curiosity, can any of you gurus explain why that happens?	[reply]
RE: RE: Re: Order of output when switching filehandles by isotope (Deacon) on Aug 14, 2000 at 23:04 UTC
I may be smoking something, but usually $\| = 1; doesn't completely disable buffering, just most of Perl's buffering. You still need to close it to force the OS buffer to be flushed.	[reply]
RE: RE: RE: Re: Order of output when switching filehandles by Cirollo (Friar) on Aug 14, 2000 at 23:12 UTC
Re: Order of output when switching filehandles by lhoward (Vicar) on Aug 14, 2000 at 23:27 UTC
Have you considered using IO::Tee instead of launching tee as separate process? `#!/usr/bin/perl -w use strict; use IO::Tee; open FH,">somefile" or die "error opening $!"; my $tee=IO::Tee->new(\STDOUT,\FH); select $tee; print "foo\n"; print "bar\n"; close FH;` [download]	[reply] [d/l]
RE: Re: Order of output when switching filehandles by Cirollo (Friar) on Aug 14, 2000 at 23:45 UTC
I considered that, but the machine I'm using does not have IO::Tee installed on it, so launching it as a seperate process is easier for all parties involved - it might not be the "right way" to do it, but theres no reason for it to not work and security is not a concern.	[reply]


Perl Monk, Perl Meditation
	PerlMonks