From d1284de05dfa3d43dc0a5cc6a63cc6c6a08233fa Mon Sep 17 00:00:00 2001
From: hjp <hjp>
Date: Mon, 24 Jul 2006 08:43:57 +0000
Subject: [PATCH] Added POD and comments to top_n. Added script
 sensorstemmpplot.

---
 sensorstemmpplot | 66 ++++++++++++++++++++++++++++++++++++++++++++++++
 top_n            | 43 +++++++++++++++++++++++++++++++
 2 files changed, 109 insertions(+)
 create mode 100755 sensorstemmpplot

diff --git a/sensorstemmpplot b/sensorstemmpplot
new file mode 100755
index 0000000..bc8cc74
--- /dev/null
+++ b/sensorstemmpplot
@@ -0,0 +1,66 @@
+#!/usr/bin/perl -w 
+
+=head1 NAME
+
+sensorstempplot - plot temperatures from (lm_)sensors output
+
+=head1 DESCRIPTION
+
+This program expects output from lm_sensors where each line is prepended
+with a timestamp, like this:
+
+  2006-07-22T15:52:01 adm1023-i2c-0-18
+  2006-07-22T15:52:01 Adapter: SMBus I801 adapter at 1100
+  2006-07-22T15:52:01 Board:       +33°C  (low  =   -55°C, high =  +127°C)
+  2006-07-22T15:52:01 CPU:         +35°C  (low  =   -55°C, high =  +127°C)
+  2006-07-22T15:52:01 
+  2006-07-22T15:52:01 max1617-i2c-0-1a
+  2006-07-22T15:52:01 Adapter: SMBus I801 adapter at 1100
+  2006-07-22T15:52:01 Board:       +37°C  (low  =   -55°C, high =  +127°C)
+  2006-07-22T15:52:01 CPU:         +40°C  (low  =   -55°C, high =  +127°C)
+  2006-07-22T15:52:01 
+
+Note that there are two lines for "Board" and "CPU" temperature with 
+different values. We distinguish them by noting the section identifier
+and prepending it to the series names. 
+
+=cut
+
+use strict;
+use TimeSeries;
+use utf8;
+
+binmode STDOUT, ':raw';
+
+my %series;
+my $ns;
+my %data;
+
+my $ts = TimeSeries->new();
+my $section = "";
+while (<>) {
+    chomp;
+    if (/^[-0-9T:]* ([\w-]+)$/) {
+	$section = $1;
+	next;
+    }
+    if (/^([-0-9T:]*) ([\w-]+):\s+([+-]?[0-9.]*)°C /) {
+	my ($timestamp, $series, $value) = ($1, "$section $2", $3);
+	$series{$series} = ++$ns unless ($series{$series});
+	$data{$timestamp}{$series} = $value;
+	next;
+    }
+}
+my @series = sort { $series{$a} <=> $series{$b} } keys %series;
+$ts->legend(@series);
+$ts->legend_position("below");
+$ts->stacked(0);
+$ts->output_format('ps');
+for my $timestamp (sort keys %data) {
+    my %d = %{$data{$timestamp}};
+    my @values = @d{@series};
+    $ts->add_timestring($timestamp, @values);
+}
+
+my $g = $ts->plot();
+print $g
diff --git a/top_n b/top_n
index 8b27b40..6e3acb9 100755
--- a/top_n
+++ b/top_n
@@ -1,6 +1,39 @@
 #!/usr/bin/perl -w 
 use strict;
 
+=head1 NAME
+
+top_n - get top n (of many) time series
+
+=head1 SYNOPSIS
+
+top_n n f v
+
+This script expects a file in tab-separated format containing all series mixed together in 
+"normalized" or "vertical" format,
+e.g., 
+
+	2006-07-22T20:00:00	Wien/Innere Stadt	31.5
+	2006-07-22T20:00:00	Wien/Hohe Warte		29.5
+	2006-07-22T21:00:00	Wien/Innere Stadt	30.9
+	2006-07-22T21:00:00	Wien/Hohe Warte		27.9
+	2006-07-22T22:00:00	Wien/Innere Stadt	30.1
+	2006-07-22T22:00:00	Wien/Hohe Warte		26.2
+	...
+
+n is the number of time series to extract.
+
+f is the number of the column with the series names, starting at 0. In the example above, this would be 1.
+
+v is the number of the column with values to be used for sorting. In the example above, this would be 2
+(there is only one column with values in this example).
+
+All values for a given series are added, and the the n series with the highest sum are selected.
+Finally all records from the selected series and an additional series "OTHER" representing the sum
+of all other series are printed.
+
+=cut
+
 my $n = shift;
 my $f = shift;
 my $v = shift;
@@ -21,6 +54,16 @@ my %index;
 
 for my $i (0 .. $#data) {
     unless ($top_n{$data[$i][$f]}) {
+
+	# this is a bit tricky, as records for the same point in time for 
+	# different series can be scattered all over the place. We assume
+	# that all columns except $f and $v are relevant to the point in 
+	# time (or some other distinguishing criterium), so we save the value, 
+	# set $v and $f to fixed values and concatenate all columns.
+	# This will get a unique value iff all columns except $f and $v are unique.
+	# If we have seen this value before, we add the current value to
+	# the record with this value and undef the current value. Otherwise we
+	# record the index of the current record.
 	my $val = $data[$i][$v];
 	$data[$i][$v] = 0; 
 	$data[$i][$f] = 'OTHER';