pjs/tools/page-loader/URLTimingDataSet.pm

262 строки
7.4 KiB
Perl

#
# The contents of this file are subject to the Mozilla Public
# License Version 1.1 (the "License"); you may not use this file
# except in compliance with the License. You may obtain a copy of
# the License at http://www.mozilla.org/MPL/
#
# Software distributed under the License is distributed on an "AS
# IS" basis, WITHOUT WARRANTY OF ANY KIND, either express or
# implied. See the License for the specific language governing
# rights and limitations under the License.
#
# The Original Code is Mozilla page-loader test, released Aug 5, 2001
#
# The Initial Developer of the Original Code is Netscape
# Communications Corporation. Portions created by Netscape are
# Copyright (C) 2001 Netscape Communications Corporation. All
# Rights Reserved.
#
# Contributor(s):
# John Morrison <jrgm@netscape.com>, original author
#
#
package URLTimingDataSet;
use DBI;
use PageData; # list of test pages, etc.
use strict;
sub new {
my $proto = shift;
my $class = ref($proto) || $proto;
my $self = {
dataset => [],
results => [],
sorted => [],
average => undef,
avgmedian => undef, # note: average of individual medians
maximum => undef,
minimum => undef,
};
$self->{id} = shift || die "No id supplied";
$self->{table} = shift || "t" . $self->{id};
$self->{pages} = PageData->new;
bless ($self, $class);
$self->_grok();
return $self;
}
sub _grok {
my $self = shift;
my ($res);
# select the dataset from the db
$self->_select();
for (my $i=0; $i < $self->{pages}->length; $i++) {
my $name = $self->{pages}->name($i);
my $count = 0;
my @times = ();
my $nan = 0;
foreach my $ref (@{$self->{dataset}}) {
next if ($name ne $ref->{content});
$count++;
if ($ref->{c_part} eq "NaN") {
# we bailed out of this page load
$res = "NaN";
$nan = 1;
}
else {
my $s_intvl = $ref->{s_intvl};
my $c_intvl = $ref->{c_intvl};
my $errval = abs($s_intvl-$c_intvl)/(($s_intvl+$c_intvl)/2);
if ($errval > 0.08) { # one of them went wrong and stalled out (see [1] below)
$res = ($s_intvl <= $c_intvl) ? $s_intvl : $c_intvl;
} else {
$res = int(($s_intvl + $c_intvl)/2);
}
}
push @times, $res;
}
my $avg = int(_avg(@times));
my $med = _med(@times);
my $max = $nan ? "NaN" : _max(@times);
my $min = _min(@times);
push @{$self->{results}}, [ $i, $name, $count, $avg, $med, $max, $min, @times ];
}
$self->_get_summary();
$self->_sort_result_set();
}
sub _select {
my $self = shift;
my $dbh = DBI->connect("DBI:CSV:f_dir=./db", {RaiseError => 1, AutoCommit => 1})
or die "Cannot connect: " . $DBI::errstr;
my $sql = qq{
SELECT INDEX, S_INTVL, C_INTVL, C_PART, CONTENT, ID
FROM $self->{table}
WHERE ID = "$self->{id}"
};
my $sth = $dbh->prepare($sql);
$sth->execute();
while (my @data = $sth->fetchrow_array()) {
push @{$self->{dataset}},
{index => $data[0],
s_intvl => $data[1],
c_intvl => $data[2],
c_part => $data[3],
content => $data[4],
id => $data[5]
};
}
$sth->finish();
$dbh->disconnect();
}
sub _get_summary {
my $self = shift;
my (@avg, @med, @max, @min);
# how many pages were loaded in total ('sampled')
$self->{samples} = scalar(@{$self->{dataset}});
# how many cycles (should I get this from test parameters instead?)
$self->{count} = int(_avg( map($_->[2], @{$self->{results}}) ));
warn $self->{count};
# calculate overall average, average median, maximum, minimum, (RMS Error?)
for (@{$self->{results}}) {
push @avg, $_->[3];
push @med, $_->[4];
push @max, $_->[5];
push @min, $_->[6];
}
$self->{average} = int(_avg(@avg));
$self->{avgmedian} = int(_avg(@med)); # note: averaging individual medians
$self->{maximum} = _max(@max);
$self->{minimum} = _min(@min);
}
sub _sort_result_set {
my $self = shift;
# sort by median load time
# @{$self->{sorted}} = sort {$a->[4] <=> $b->[4]} @{$self->{results}};
# might be "NaN", but this is lame of me to be carrying around a string instead of undef
@{$self->{sorted}} =
sort {
if ($a->[4] eq "NaN" || $b->[4] eq "NaN") {
return $a->[4] cmp $b->[4];
} else {
return $a->[4] <=> $b->[4];
}
} @{$self->{results}};
}
sub as_string {
my $self = shift;
return $self->_as_string();
}
sub as_string_sorted {
my $self = shift;
return $self->_as_string(@{$self->{sorted}});
}
sub _as_string {
my $self = shift;
my @ary = @_ ? @_ : @{$self->{results}};
my $str;
for (@ary) {
my ($index, $path, $count, $avg, $med, $max, $min, @times) = @$_;
$str .= sprintf "%3s %-26s\t", $index, $path;
if ($count > 0) {
$str .= sprintf "%6s %6s %6s %6s ", $avg, $med, $max, $min;
foreach my $time (@times) {
$str .= sprintf "%6s ", $time;
}
}
$str .= "\n";
}
return $str;
}
#
# package internal helper functions
#
sub _num {
my @array = ();
for (@_) { push @array, $_ if /^[+-]?\d+\.?\d*$/o; }
return @array;
}
sub _avg {
my @array = _num(@_);
return "NaN" unless scalar(@array);
my $sum = 0;
for (@array) { $sum += $_; }
return $sum/scalar(@array);
}
sub _max {
my @array = _num(@_);
return "NaN" unless scalar(@array);
my $max = $array[0];
for (@array) { $max = ($max > $_) ? $max : $_; }
return $max;
}
sub _min {
my @array = _num(@_);
return "NaN" unless scalar(@array);
my $min = $array[0];
for (@array) { $min = ($min < $_) ? $min : $_; }
return $min;
}
# returns the floor(N/2) element of a sorted ascending array
sub _med {
my @array = _num(@_);
return "NaN" unless scalar(@array);
my $index = int((scalar(@array)-1)/2);
@array = sort {$a <=> $b} @array;
return $array[$index];
}
1; # return true
################################################################################
#
# [1] in looking at the test results, in almost all cases, the
# round-trip time measured by the server logic and the client logic
# would be almost the same value (which is what one would
# expect). However, on occasion, one of the them would be "out of
# whack", and inconsistent with the additional "layout" measure by the
# client.
#
# i.e., a set of numbers like these:
# c_part c_intvl s_intvl
# 800 1003 997
# 804 1007 1005
# 801 1001 1325 <--
# 803 1318 998 <--
# 799 1002 1007
# ...
#
# which looks like the server side would stall in doing the accept or
# in running the mod-perl handler (possibly a GC?). (The following
# c_intvl would then be out of whack by a matching amount on the next
# cycle).
#
# At any rate, since it was clear from comparing with the 'c_part'
# measure, which of the times was bogus, I just use an arbitrary error
# measure to determine when to toss out the "bad" value.
#