Mandatory background story

I had an end-of-day job that was taking too much time, and it turns out that the problem was retrieving the closing time values from several textual log files, using a sequential scan.

I knew the answer was to do a binary search, but I was fairly certain someone else HAD to have run into this problem before, so I typed text file binary search in the Search box here, and got Binary Searches on Sorted Text Files.

runrig's reply in that thread pointed me to File::SortedSeek.

Module overview

File::SortedSeek implements binary searches through large sorted text files, using string, numeric, or timestamp keys.

It's ideal for finding a particular time in large logs, for instance.

The module was authored by Dr. James Freeman, better known around here as tachyon.

Usage example

The files I have to deal with use a non-standard integer timestamp. Each line looks like

{timestamp} {value} {value} {value} {value}

Using the module to find the right line couldn't be easier!

#!/usr/bin/perl -w
use strict;
use File::SortedSeek ':all';

my $targetTS=1234567789;  # for instance
open(FILE,"<data.in") or die "Can't open data.in, error $!";

# File::SortedSeek::numeric looks for a numeric key
# It takes an optional sub ref argument to "massage" the
# line from the file to extract the key.
numeric(*FILE,$targetTS,sub {$_[0]=~/^\s*(\d+)/ or return 0; return $1
+});

# this will return the line that matches $targetTS, or the first line 
+>
$line=<FILE>;

close(FILE);
[download]

Conclusion

tachyon rules!!! :)