thinrope / fixed_sensor_visualization Goto Github PK

Currently executing make distclean publish doesn't work due to parallel build.
Workaround: make distclean; make publish

Discussions:
http://gnu-make.2324884.n4.nabble.com/order-only-prerequisites-don-t-behave-as-I-d-have-expected-after-reading-the-documentation-tt3787.html#a15681

Sensors can have different update windows, by design.
We should detect the usual value(s) and report if we miss data.

When a sensor changes update window (e.g. 10 -> 1 min and back), there should be a way to understand it.

In the graph, this will be shown as missing data (as opposed to 0) preferably. Maybe 0 is acceptable, given the color coding. These missing values should be computed after the statistics, to avoid pulling down the average and stdev.

Probably getting the data processing in a stand-alone script is required (SMA is now computed inside gnuplot).

On error in the makefile, the script stop plotting. One error happens if the cvs file is empty. Can be avoided if the scripts detect empty file and disregards the device .

rob

There a few places where complex system() calls are used within the .gnuplot scripts... This creates problems such as hard to check failures (e.g. undefined DRE2CPM).

Better to move all such processing to outside and pass (yet another) envvar?

Currently I use 1h meta-equv tag, so depending on the time the page is accessed 1st, a user can me almost 1h behind the newest updates always ...

A better solution is to use either CGI/SSI or some nice Apache setting to calculate that expire header at the time of access.

http://www.gnuplot.info/ReleaseNotes_5_0.html

• use better enhanced mode for text (bold, *2)
• Secondary axes (x2, y2) can be locked to the primary axis via a mapping function. In the simplest case this guarantees that the primary and secondary axis ranges are identical. In the general case it allows you to define a non-linear axis, something that previously was only possible for the special case of log scaling.
• Hypertext labels in the interactive terminals including web display using the HTML canvas or svg terminals.
• use "set {xy}tics time" rather than by "set {xy}data time"
• confirm epoch changes do not affect us (2000 ->1970)

After some additional testing, the plotted data was found to be 9h after the requested start (PLOT_SINCE parameter), due to bug in cacher.pl
I became obvious when trying to plot smaller intervals (e.g. 6 hours).

1. Implement broken axis in gnuplot (ask gnuplot devs?)
2. Add visual cue that scale is not "default"

1. Detect sensors that have no data during the plotted period and make a picture with "No DATA!" or something, but keep sensor QTH, etc.
2. Change icon in Fusion table (or not)

A few places use JST as timezone, use +0900 instead (for internal), add timezone in graphs as label somehow.
http://en.wikipedia.org/wiki/ISO_8601#Time_zone_designators

The Makefile references a command suc that is not common: