NAME
Logfile - Perl extension for generating reports from logfiles
SYNOPSIS
use Logfile::Cern;
$l = new Logfile::Cern File => 'cache.log.gz',
Group => [Domain,File,Hour];
$l->report(Group => File, Sort => Records);
$l->report(Group => Domain, Sort => Bytes);
$l->report(Group => Hour, List => [Bytes, Records]);
use Logfile::Wftp;
[...]
DESCRIPTION
The Logfile extension will help you to generate various reports from different server logfiles. In general there is no restriction as to what information you extract from the logfiles.
Reading the files
The package can be customized by subclassing "Logfile".
A subclass should provide a function "next" which reads the next record from the file handle "$self->{Fh}" and returns an object of type "Logfile::Record". In addition a function "norm" may be specified to normalize the various record fields.
Here is a shortened version of the "Logfile::Cern" class:
package Logfile::Cern;
@ISA = qw ( Logfile::Base ) ;
sub next {
my $self = shift;
my $fh = $self->{Fh};
*S = $fh;
my ($line,$host,$user,$pass,$rest,$date,$req,$code,$bytes);
($host,$user,$pass,$rest) = split ' ', $line, 4;
($rest =~ s!\[([^\]]+)\]\s!!) && ($date = $1);
($rest =~ s!\"([^\"]+)\"\s!!) && ($req = (split ' ', $1)[1]);
($code, $bytes) = split ' ', $rest;
Logfile::Record->new(Host => $host,
Date => $date,
File => $req,
Bytes => $bytes);
}
As stated above, in general you are free to choose the fields you enter in the record. But:
Date should be a valid date string. For conversion to the seconds
elapsed since the start of epoch the modules GetDate and
Date::DateParse are tried. If both cannot be "use"ed, a crude
build-in module is used.
The record constructor replaces Date by the date in "yymmdd" form
to make it sortable. Also the field Hour is padded in.
Host Setting Host will also set field Domain by the verbose name of the
country given by the domain suffix of the fully qualified domain
name (hostname.domain). "foo.bar.PG" will be mapped to "Papua New".
Host names containing no dot will be assigned to the domain Local.
IP numbers will be assigned to the domain Unresolved. Mapping of
short to long domain names is done in the Net::Country extension
which might be useful in other contexts:
use Net::Country;
$germany = Net::Country::Name('de');
Records
is always set to 1 in the "Record" constructor. So this field gives
the number of successful returns from the "next" function.
Here is the shortened optional "norm" method:
sub norm {
my ($self, $key, $val) = @_;
if ($key eq File) {
$val =~ s/\?.*//; # remove query
$val =~ s!%([\da-f][\da-f])!chr(hex($1))!eig; # decode escapes
}
$val;
}
The constructor reads in a logfile and builds one or more indices.
$l = new Logfile::Cern File => 'cache.log.gz',
Group => [Host,Domain,File,Hour,Date];
There is little space but some time overhead in generating additional indexes. If the File parameter is not given, STDIN is used. The Group parameter may be a field name or a reference to a list of field names. Only the field names given as constructor argument can be used for report generation.
Report Generation
The Index to use for a report must be given as the Group parameter.
Output is sorted by the index field unless a Sort parameter is given.
Also the output can be truncated by a Top argument or Limit.
The report generator lists the fields Bytes and Records for a given index. The option List may be a single field name or a reference to an array of field names. It specifies which field should be listed in addition to the Group field. List defaults to Records.
$l->report(Group => Domain, List => [Bytes, Records])
Output is sorted by the Group field unless overwritten by a Sort option. Default sorting order is increasing for Date and Hour fields and decreasing for all other Fields. The order can be reversed using the Reverse option.
This code
$l->report(Group => File, Sort => Records, Top => 10);
/htbin/SFgate 30 31.58%
/freeWAIS-sf/ 22 23.16%
/SFgate/SFgate 8 8.42%
/SFgate/SFgate-small 7 7.37%
/icons/ 4 4.21%
/~goevert 3 3.16%
/journals/SIGMOD 3 3.16%
/SFgate/ciw 2 2.11%
/search 1 1.05%
/reports/96/ 1 1.05%
Here are other examples. Also take a look at the t/* files.
$l->report(Group => Domain, Sort => Bytes);
Germany 12 12.63%
Unresolved 8 8.42%
Israel 34 35.79%
Denmark 4 4.21%
Canada 3 3.16%
Network 6 6.32%
US Commercial 14 14.74%
US Educational 8 8.42%
Hong Kong 2 2.11%
Sweden 2 2.11%
Non-Profit 1 1.05%
Local 1 1.05%
$l->report(Group => Hour, List => [Bytes, Records]);
Hour Bytes Records
07 245093 17.66% 34 35.79%
08 438280 31.59% 19 20.00%
09 156730 11.30% 11 11.58%
10 255451 18.41% 16 16.84%
11 274521 19.79% 10 10.53%
12 17396 1.25% 5 5.26%
Report options
Group "=>" field
Mandatory. field must be one of the fields passed to the
constructor.
List "=>" field
List "=>" [field, field]
List the subtotals for fields. Defaults to Records.
Sort "=>" field.
Sort output by field. By default, Date and Hour are sorted in
increasing order, whereas all other fields are sorted in decreasing
order.
Reverse "=> 1"
Reverse sorting order.
Top "=>" number
Print only the first number subtotals.
Limit "=>" number
Print only the subtotals with Sort field greater than number
(less than number if sorted in increasing order).
Currently reports are simply printed to STDOUT.
AUTHOR
Ulrich Pfeifer <pfeifer@wait.de>
NEWS
Fixed strict refs bug for perl 5.005.
Fixed bug in fallback to included date parsing reported by James Downs.
Fixed y2k bug as suggested by Fred Korz. I chose the two digit version to be a backward compatible as possible. The 20 will be obvious a few years from now ;-) Output columns should now be separated by whitespace in any case.
SEE ALSO
perl(1).