bramp.net

Beware the \r\n with grep -f, also perhaps PHP is a better choice?

bramp ♦ June 19, 2011 ♦ Leave Your Comment

I had two files, one large CSV file (10million rows), and another file (600K rows). I wanted to find all the lines in the large CSV file that contained a word in the smaller file. The smaller file was a simple text with one word per line.

I found that grep could do:

grep -f smallfile largefile > results.csv

Which would build a list of patterns from the contents of the smallfile. This seemed simple enough, however it didn’t work. Some investiagtion showed that the smallfile had windows
new lines, and grep assumed the \r was part of the pattern. Using dos2unix fixed my problem.

However, new problem, grep used 100% of my CPU and 2GB of my RAM and took over 5 hours. I actually gave up before I let it finish. I assume it was building a large regexp parse tree.
So I figured to allow it to make a simpler tree I would alter my smallfile to contain “^keyword”, instead of just “keyword”. I could do this because I knew the keyword would always
be at the beginning of the line. So awk ‘{print “^”$1} smallfile allowed me to do that. I tried grep again but it seemed to again be taking a long time.

While waiting for grep to finish, I started to write a PHP script:

<?php
// Prints out each line from a CSV where the first entry is in another list
// By Andrew Brampton March 2011
// php inlist.php bigfile smallfile
// TODO we need better names and more error checking

$bigfile = $argv[1];
$littlefile = $argv[2];

$little = array();
$fp = fopen($littlefile, 'r') or die('Failed to open little file');
while ($line = fgets($fp)) {
        $line = trim($line);
        $little[ $line ] = true;
}
fclose($fp);

$fp = fopen($bigfile, 'r') or die ('Failed to open big file');
while ($line = fgets($fp)) {
        $num = strpos($line, ',');
        $num = substr($line, 0, $num);

        if (isset($little[$num])) {
                unset($little[$num]); // We unset so we can get a count/list of all those not in the list
                echo $line;
        }
}
fclose($fp);
?>

It took me 10 minutes to write this. I quickly ran it, and it completed within 20seconds. Looks like this is a far more efficient way to do it