Position in seek() confusion

[stevieb]

I've never had a need to use seek() in my Perl years, so after I got looking at a piece of code BrowserUK posted earlier today, I thought I'd ask about it.

In the below code in the seek $fh, $_ *80, 0; line, I would believe that it would set the cursor to column 80 (times the multiple in the while(...)), but it then proceeds to print 78 chars to the file. When I look at the file, all lines start at column 0.

#! perl -slw
use strict;
use threads;

open my $fh, '>', 'junk.dat' or die $!;
$_->join for map {
    async {
        seek $fh, $_ *80, 0;
        print $fh $_ x 78;
    };
} 1 .. 4;
close $fh;

open $fh, '<', 'junk.dat' or die $!;
print while <$fh>;
close $fh;

perldoc -f seek doesn't help it click for me. Can someone advise what I'm missing in the above code? I did modify the code to try to understand, but the reason why things start at col 0 elude me.

-stevieb

[BrowserUk]

That was a crude demonstration knocked up in haste. If you run it, you'll find that in addition what you've noticed, it also prints the lines double spaced.

This is due to the use of -l on the shebang line. Newlines are added when the lines are printed to the file and then again when they are read back and printed to STDOUT.

I took the liberty of discarding the spaces when posting. If you want a better version, try:

#! perl -sw
use strict;
use threads;
use threads::shared;

my $sem :shared;

open my $fh, '>:raw', 'junk.dat' or die $!;
$_->join for map {
    async {
        for my $i ( 0 .. 999 ) {
            lock $sem;
            seek $fh, $i*320 + ($_-1)*80, 0;
            print $fh $_ x 79, "\n";
        }
    };
} 1 .. 4;
close $fh;

open $fh, '<', 'junk.dat' or die $!;
print while <$fh>;
close $fh;

__END__
C:\test>t-write.pl
1111111111111111111111111111111111111111111111111111111111111111111111
[phone]
2222222222222222222222222222222222222222222222222222222222222222222222
[phone]
3333333333333333333333333333333333333333333333333333333333333333333333
[phone]
4444444444444444444444444444444444444444444444444444444444444444444444
[phone]
1111111111111111111111111111111111111111111111111111111111111111111111
[phone]
2222222222222222222222222222222222222222222222222222222222222222222222
[phone]
3333333333333333333333333333333333333333333333333333333333333333333333
[phone]
4444444444444444444444444444444444444444444444444444444444444444444444
[phone]
1111111111111111111111111111111111111111111111111111111111111111111111
[phone]
2222222222222222222222222222222222222222222222222222222222222222222222
[phone]
3333333333333333333333333333333333333333333333333333333333333333333333
[phone]
4444444444444444444444444444444444444444444444444444444444444444444444
[phone]
1111111111111111111111111111111111111111111111111111111111111111111111
[phone]
2222222222222222222222222222222222222222222222222222222222222222222222
[phone]
3333333333333333333333333333333333333333333333333333333333333333333333
[phone]
4444444444444444444444444444444444444444444444444444444444444444444444
[phone]
1111111111111111111111111111111111111111111111111111111111111111111111
[phone]
2222222222222222222222222222222222222222222222222222222222222222222222
[phone]
3333333333333333333333333333333333333333333333333333333333333333333333
[phone]
4444444444444444444444444444444444444444444444444444444444444444444444
[phone]
1111111111111111111111111111111111111111111111111111111111111111111111
[phone]
2222222222222222222222222222222222222222222222222222222222222222222222
[phone]
3333333333333333333333333333333333333333333333333333333333333333333333
[phone]
4444444444444444444444444444444444444444444444444444444444444444444444
[phone]
1111111111111111111111111111111111111111111111111111111111111111111111
[phone]
2222222222222222222222222222222222222222222222222222222222222222222222
[phone]
3333333333333333333333333333333333333333333333333333333333333333333333
[phone]
4444444444444444444444444444444444444444444444444444444444444444444444
[phone]
1111111111111111111111111111111111111111111111111111111111111111111111
[phone]
2222222222222222222222222222222222222222222222222222222222222222222222
[phone]
3333333333333333333333333333333333333333333333333333333333333333333333
[phone]
4444444444444444444444444444444444444444444444444444444444444444444444
[phone]
1111111111111111111111111111111111111111111111111111111111111111111111
[phone]
2222222222222222222222222222222222222222222222222222222222222222222222
[phone]
3333333333333333333333333333333333333333333333333333333333333333333333
[phone]
4444444444444444444444444444444444444444444444444444444444444444444444
[phone]
1111111111111111111111111111111111111111111111111111111111111111111111
[phone]
2222222222222222222222222222222222222222222222222222222222222222222222
[phone]
3333333333333333333333333333333333333333333333333333333333333333333333
[phone]
4444444444444444444444444444444444444444444444444444444444444444444444
[phone]
...

Note: I truncated some of the output when posting.

[Anonymous Monk]

Ok, Deparse shows

BEGIN { $^W = 1; }
BEGIN { $/ = "\n"; $\ = "\n"; }
use threads;
use strict;
die $! unless open my $fh, '>', 'junk.dat';
$_->join foreach (map {async sub {
    seek $fh, $_ * 80, 0;
    print $fh $_ x 78;
}
;} 1..4);
close $fh;
die $! unless open $fh, '<', 'junk.dat';
print $_ while defined($_ = <$fh>);
close $fh;

So on windows "\n" becomes the bytes "\r\n" when you print, so 78 + 2 = 80, thats the lines

so seek to 1*80 from zero print "1" 78 times followed by "\r\n"

so seek to 2*80 from zero print "2" 78 times followed by "\r\n"

so seek to 3*80 from zero print "3" 78 times followed by "\r\n"

so seek to 4*80 from zero print "4" 78 times followed by "\r\n"

Then when reading the file the first 80 chars are null ("\0") then its 1111...222...333...4444

Thats Basic debugging checklist for you

[Anonymous Monk]

It's not at all clear to me how B::Deparse makes it obvious for people not familiar with seek... OTOH, Devel::Peek does. This is on Linux:

open my $fh, '>', 'junk.dat' or die $!;

for ( 1 .. 4 ) {
    seek $fh, $_ * 80, 0;
    print $fh $_ x 78, "\r\n";
}

open $fh, '<', 'junk.dat' or die $!;

while ( <$fh> ) {
    use Devel::Peek;
    Dump $_;
}

output:

SV = PV(0x189fd70) at 0x18c0a38
  REFCNT = 1
  FLAGS = (POK,pPOK)
  PV = 0x18a4610 "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0
+\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\
+0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0001111111111111111111111111111
+11111111111111111111111111111111111111111111111111\r\n"\0
  CUR = 160
  LEN = 200
SV = PV(0x189fd70) at 0x18c0a38
  REFCNT = 1
  FLAGS = (POK,pPOK)
  PV = 0x18a4610 "2222222222222222222222222222222222222222222222222222
+22222222222222222222222222\r\n"\0
  CUR = 80
  LEN = 200
SV = PV(0x189fd70) at 0x18c0a38
  REFCNT = 1
  FLAGS = (POK,pPOK)
  PV = 0x18a4610 "3333333333333333333333333333333333333333333333333333
+33333333333333333333333333\r\n"\0
  CUR = 80
  LEN = 200
SV = PV(0x189fd70) at 0x18c0a38
  REFCNT = 1
  FLAGS = (POK,pPOK)
  PV = 0x18a4610 "4444444444444444444444444444444444444444444444444444
+44444444444444444444444444\r\n"\0
  CUR = 80
  LEN = 200

So, when you seek past the end of file, it's as if the filesystem fills that part of file (0 .. 79 bytes in our case) with zero bytes (actually it creates a 'hole' in the file, which, when read, returns zeros - apparently NTFS also does that?). These zero bytes are then actually printed, just not displayed on the terminal:

$ perl -E 'say "\0\0\0ABC"
ABC
$ perl -E 'say "\0\0\0ABC"' | perl -nE 'printf "%vx\n", $_'
0.0.0.41.42.43.a

[Anonymous Monk]

Use B::Deparse and it will become obvious

[sundialsvc4]

You folks seriously told the OP to “de-parse” a Perl subroutine into its underlying bytecode-tree, assuring him or her that thereby it would be “obvious?”   Really??   ... :-/ ... Gee, you people really are nerds!

The one and only thing to remember about seek(), in any programming language at all, is that it works strictly on byte-positions within the file.   It does not know about newline sequences (in any of their one- or two-byte flavors ...), nor does it know about Unicode.   The file consists of nothing more and nothing less than a collection of zero-or-more bytes, and seek() sets the read/write cursor to an absolute or relative byte position within that collection.

[Anonymous Monk]

You seriously replied to the OP indirectly? Not every context is a beginner's context. In this case, the OP was looking for deep knowledge and not just the Cliffs Notes. Can you not see this?

[Anonymous Monk]

sundialsvc4:

You folks seriously told the OP to “de-parse” a Perl subroutine into its underlying bytecode-tree, assuring him or her that thereby it would be “obvious?”   Really??   ... :-/ ... Gee, you people really are nerds!

The one and only thing to remember about seek(), in any programming language at all, is that it works strictly on byte-positions within the file.   It does not know about newline sequences (in any of their one- or two-byte flavors ...), nor does it know about Unicode.   The file consists of nothing more and nothing less than a collection of zero-or-more bytes, and seek() sets the read/write cursor to an absolute or relative byte position within that collection.

Poor flushells, you really don't know anything about B::Deparse, print, $/, -l

Why starts at "column 0"? Why 78/80? Because \r\n is added