in reply to Unique visits - Webserver log parser

2.8GB huh? Thems some pretty monstrous log files. Gonna take you a while to churn through them...

Well, I suspect what you'll find is that this isn't the only statistic you'll need, so why not shove the data into a database instead? Yes, this is going to be more time consuming in the short run, but it's fun to play with alternative suggestions...

So here's some code that shoves a Combined log format into an SQLite table. Of course you'll double your hard disk requirements, and it'll take probably a good couple of hours to create the database, but once you've done that you can pull off some pretty neat queries.

use strict; use DBI; use Time::Piece; use Fatal qw(open close); # parsing: # 213.20.65.52 - - [01/Jan/2002:09:30:38 +0000] "GET /img/bg.gif HTTP/ +1.1" 200 268 # "http://www.axkit.org/" "Mozilla/4.0 (compatible; MSIE 6.0; Windows + NT 5.0; Q31 2461)" use constant IP => 0; use constant AUTH => 1; use constant DATE => 2; use constant REQUEST => 3; use constant STATUS => 4; use constant BYTES => 5; use constant REFERER => 6; use constant UA => 7; use constant METHOD => 8; use constant URI => 9; my $logfile = $ARGV[0] || die "Usage: $0 filename\n"; open(LOG, $logfile); my $dbh = DBI->connect("dbi:SQLite:$logfile.db","","", { AutoCommit => 0, RaiseError => 1 }); print "Dropping old table...\n"; eval { $dbh->do("DROP TABLE access_log"); }; print "Done\n"; $dbh->do(<<EOT); CREATE TABLE access_log ( when datetime not null, host varchar(255) not null, method varchar(10) not null, url varchar(500) not null, auth, browser, referer, status integer default 0, bytes integer ) EOT my $sth = $dbh->prepare(<<EOT); INSERT INTO access_log VALUES ( ?,?,?,?,?,?,?,?,? ) EOT my $line = 0; while (<LOG>) { chomp; # superfluous, but we do it anyway $line++; my @vals; # adjust the regexp depending on your log format if (/^([^ ]*) [^ ]*? ([^ ]*?) \[([^\]]*?)\] "(.*?)" ([^ ]*?) ([^ ] +*?) "(.*?)" "(.*?)"$/) { @vals = ($1,$2,$3,$4,$5,$6,$7,$8); } else { warn("Corrup log line: $_\n"); next; } eval { $vals[DATE] = Time::Piece->strptime($vals[DATE], '%d/%b/%Y:%H: +%M:%S +0000')->datetime; }; if ($@) { die "Failed to parse $vals[DATE] on line $line\n"; } if ($vals[REQUEST] =~ /^(\w+) ([^ ]+)/) { $vals[METHOD] = $1; $vals[URI] = $2; } else { warn "Couldn't parse: $vals[REQUEST] on line $line\n" if $vals[REQUEST] ne '-'; $vals[METHOD] = "INVALID_METHOD"; $vals[URI] = "-"; } # print join(':', @vals), "\n"; $sth->execute(@vals[DATE,IP,METHOD,URI,AUTH,UA,REFERER,STATUS,BYTE +S]); unless ($line % 1000) { print "Completed $line lines, committing.\n"; $dbh->commit; } } close LOG; $sth->finish; $dbh->disconnect;
Comments on the code welcome.

Replies are listed 'Best First'.
•Enumerated lists (was Re: Re: Unique visits - Webserver log parser)
by merlyn (Sage) on Feb 27, 2002 at 15:07 UTC
    In place of
    use constant IP => 0; use constant AUTH => 1; use constant DATE => 2; use constant REQUEST => 3; use constant STATUS => 4; use constant BYTES => 5; use constant REFERER => 6; use constant UA => 7; use constant METHOD => 8; use constant URI => 9;
    you can put this (after installing enum):
    use enum qw(IP AUTH DATE REQUEST STATUS BYTES REFERER UA METHOD URI);
    Much nicer and less chance of making a mistake.

    -- Randal L. Schwartz, Perl hacker

      True enough... Why isn't enum.pm shipping with Perl 5.8??? It seems an ideal candidate to me. I didn't use it because I don't have it installed, though I have been looking at it lately because the code I'm refactoring has a lot of enum-like constants in it.

      Time to fire off an email to jarkko methinks.

      Update: cool, Jarkko has put it on the TODO list for 5.9.