in reply to Peculiar Perl/R Problem - Memory?

Most likely you are running against IPC buffering. The likely case is, that somewhere in between your program and R, all your commands stack in a buffer that waits for more input, and your read call waits for more input as well.

I would look if the R module maybe has some flush command to flush all buffers after your send sequence. If there is no such facility, you will have to either create it yourself, by grabbing into the guts of R::Statistics and changing the file handle used for communication to autoflush, or by tearing down the communication after each batch of commands. Maybe you should simply batch-send all your commands one after another, and only then start reading the results in.

Replies are listed 'Best First'.
Re^2: Peculiar Perl/R Problem - Memory?
by maybeD (Sexton) on Aug 08, 2005 at 13:55 UTC
    OK, I've tried a few more things to test what the effects on the problem would be, now I know what the likely cause is.

    ->Eliminating the read commands did not affect the problem. The script still halts at arbitrary points in the loop even without the 'read from R' element.
    ->Restarting R at the end of the foreach loop (using the command $R -> restartR();) prevented the random halting. However, as you might expect the script is very slow with this workaround in place, since it has to load R anew with each iteration of the loop.

    I am somewhat new to the use of modules in my scripts, and completely so to IPC buffering, so any more guidance you could provide into how to implement the autoflush would be very much appreciated.

    Thanks for your help, M

[reply]
      Update: Although because of the (greatly complicated) Perl module that this script is reliant on I wasn't able to try a lot of the suggestions, a simple modification has reduced the incidence of the problem to a much more acceptable level (now about 1 in 8 of the times the script is run). The modification is as follows: 
         $R_send_0 = "$R_comment_out$R_line0";
    $R_send_1 = "$R_line1$R_line2$R_line3$R_line4$R_line5";
    $R_send_2 = "$R_dblquotation_start$R_line6$R_send_3";

my $r_send_comment = $R_send_0;
my $r_send_string = $R_send_1;
my $r_send_string2 = $R_send_2;
$R->send (qq'$r_send_string');
$R->send (qq'$r_send_string2');
my $ret = $R->read;
push (@R_output_store, $r_send_comment);
push (@R_output_store, $ret);
print $ret;
print "\n\n";

      [download]
      has been changed to: 
      $R_send_0 = "$R_comment_out$R_line0";
    $R_send_1 = "$R_line1$R_line2$R_line3$R_line4$R_line5";
    $R_send_2 = "$R_dblquotation_start$R_line6$R_send_3";
    $R_send = "$R_send_1\n$R_send_2";


my $r_send_comment = $R_send_0;
my $r_send_string = $R_send;
$R->send (qq'$r_send_string');
push (@R_output_store, $r_send_comment);
print $r_send_comment;
print "\n";
my $ret = $R->read;
push (@R_output_store, $ret);
print $ret;
print "\n\n";

      [download]
      The program as I say is still not completely free of problems. The freezing continues to occur at a low frequency, and a warning message of the form: 'Warning Message: cannot open file 'input.4.r' is sometimes displayed at the end of the screen-printed output.
[reply]
[d/l]
[select]
Re^2: Peculiar Perl/R Problem - Memory?
by anonymized user 468275 (Curate) on Aug 08, 2005 at 13:14 UTC
    If a CPAN module is missing an essential method you want, I suggest it is easier to write a new little package that merely:

    - inherits the CPAN module, e.g. using @ISA

    - defines the required new method

    One world, one people

[reply]

In Section Seekers of Perl Wisdom