Blank graph of live data

[GoogleCodeExporter]

What steps will reproduce the problem?
1. Show Live Data
2.
3.

What is the expected output? What do you see instead?
A graph with a red data line. 
A graph is shown with changing axes but no data (red) line.

What version of the product are you using? On what operating system?
0.9.29

Please provide any additional information below.
0.9.28b works just fine.

Original issue reported on code.google.com by john.h.m...@gmail.com on 1 Dec 2009 at 5:31

[GoogleCodeExporter]

What version are you using - the compiled Windows exe, or the real Python 
script?

I'm asking because I compiled 0.9.29 using a newer version of matplotlib 

Prior to this, I was using matplotlib 0.98.5.2 

For 0.9.29, I used matplotlib 0.99.1 

To test whether the newer version of matplotlib is causing you a problem, I've 
compiled CurrentCost-
0.9.28b using matplotlib 0.99.1 -  
http://currentcostgui.googlecode.com/files/currentcost-0.9.28b-matplotlib-0.99.1
.zip

If you would be willing to test this and let me know if it still causes a 
problem I can look into what might 
be wrong with the newer matplotlib

Original comment by dale.l...@gmail.com on 1 Dec 2009 at 7:57

[GoogleCodeExporter]

Yep, same problem here. Using Windows exe 0.9.29 on WinXP SP3.
Just tested 0.9.28b using matplotlib 0.99.1 and that is graphing fine.

Original comment by bamz...@gmail.com on 1 Dec 2009 at 9:58

[GoogleCodeExporter]

thanks for that - very helpful

in this case, to diagnose why my code is failing on your 
machine, please could you collect diagnostics (ideally from 
before launching the live tab)

http://code.google.com/p/currentcostgui/wiki/Debug

Original comment by dale.l...@gmail.com on 1 Dec 2009 at 10:13

[GoogleCodeExporter]

Same result here - your 0.9.28b with matplotlib 0.99.1 works just fine.

I've attached the logfile for 0.9.29

Original comment by john.h.m...@gmail.com on 2 Dec 2009 at 9:28

Attachments:

• currentcostdiagnostics.log

[GoogleCodeExporter]

Sorry - I'm stumped. 

I can't recreate it on either of my Windows machines - the live graph works 
fine for 
me. 

And there is nothing in the log - no errors or exceptions - to explain a 
problem.

In the meantime, a test copy of 0.9.29 is available from 
http://currentcostgui.googlecode.com/files/currentcost-0.9.29-testbld.zip

This has:
- additional diagnostics
- the only new significant code change (use of something called 'SpanSelector') 
removed, in case that is a problem

Any feedback on this version would be much appreciated

Thanks very much

D

Original comment by dale.l...@gmail.com on 3 Dec 2009 at 12:34

[GoogleCodeExporter]

Hi, I don't have Live Data either. I'm a new user and have 0.9.29 on XP

Original comment by rolandpa...@gmail.com on 5 Dec 2009 at 12:35

[GoogleCodeExporter]

Please see my message above - I can't recreate this or figure out why it is 
happening, so will need the help of people who do have this problem to make it 
possible for me to investigate. 

For now - that means trying 
http://currentcostgui.googlecode.com/files/currentcost-
0.9.29-testbld.zip and letting me know what happens, and providing me with a 
copy of 
the debug output (http://code.google.com/p/currentcostgui/wiki/Debug). 

Sorry to fob this stuff off to users, but it's either that or this is unlikely 
to be 
fixed.

Original comment by dale.l...@gmail.com on 6 Dec 2009 at 1:33

[GoogleCodeExporter]

0.9.29-testbld is the same, graph axis changes but no graph line drawn, debug 
output 
attached

Original comment by rolandpa...@gmail.com on 7 Dec 2009 at 11:18

Attachments:

• currentcostdiagnostics.log

[GoogleCodeExporter]

Hi - here's the log from your new debug version.  I've just installed on a 
completely
seperate XP machine - same problem.

Original comment by john.h.m...@gmail.com on 7 Dec 2009 at 11:36

Attachments:

• currentcostdiagnostics.log

[GoogleCodeExporter]

[deleted comment]

[GoogleCodeExporter]

Many thanks for the updated logs. 

I am slightly confused that Google is still reporting a download count of 0 for 
my test 
build (see <http://code.google.com/p/currentcostgui/downloads/list?
can=4&q=testbld&colspec=Filename+Summary+Uploaded+Size+DownloadCount> - 
Downloads page)

At any rate, I've gone through the code again trying to guess why you're seeing 
this (and 
all the machines I've tried have not)... 

I've got a couple of ideas, but to be honest, they are shots in the dark. For 
example, I 
can see how there might be a potential problem for users with computers with 
the timezone 
not set to UTC. 

I've put in some fixes for these, and added a ton of new diagnostics output to 
create a 
new test build:
<http://currentcostgui.googlecode.com/files/currentcost-0.9.29-testbld-2.zip>

Please can someone try this and collect diagnostics again (apologies again for 
being a 
pain!), ideally starting the diagnostics from before the live graph is created

Kind regards

D

Original comment by dale.l...@gmail.com on 7 Dec 2009 at 8:17

[GoogleCodeExporter]

Nailed it! Looks like your UTC woes were the problem - live data now graphs 
correctly.

I've attached the diagnostics file in case it reveals anything else.

John

Original comment by john.h.m...@gmail.com on 7 Dec 2009 at 9:16

Attachments:

• currentcostdiagnostics.log

[GoogleCodeExporter]

Many thanks for that.

The test builds still had the latest functionality (the ability to select from 
the 
live graphs) disabled. 

A final test build with all the new code restored is available at 
http://currentcostgui.googlecode.com/files/currentcost-0.9.29-testbld-3.zip

If this still works, then we can (finally!) call this fixed. :-)

Original comment by dale.l...@gmail.com on 7 Dec 2009 at 9:40

[GoogleCodeExporter]

Looks good to me! :-) :-)

No apparent problems.

Original comment by john.h.m...@gmail.com on 7 Dec 2009 at 9:45

[GoogleCodeExporter]

Just noticed that that the time displayed in the live graph is not my time.  It 
looks
like it's showing UTC not AEST (UTC +11 at present)

Original comment by john.h.m...@gmail.com on 7 Dec 2009 at 9:50

[GoogleCodeExporter]

I'm also seeing odd data on the average weekday graph.  

Only data for 2100 2300 0100 is showing - no other time slots.

I have just switched onto a new machine (18 hours ago roughly) - so only have a 
few
hours of data stored.

Screenshots attached.

Original comment by john.h.m...@gmail.com on 7 Dec 2009 at 10:01

Attachments:

• Hourly.gif
• [Avg Day.gif](https://storage.googleapis.com/google-code-attachments/currentcostgui/issue-15/comment-16/Avg Day.gif)

[GoogleCodeExporter]

So the original problem was that the red line was being drawn in one place, 
then the 
graph was zoomed to a different place - the difference between the two places 
being 
the difference in time zone. Because I didn't realise you weren't in the UK. 

Which explains why every machine I tried it on (all of which were in the UK) 
didn't 
have this problem!

So... in my effort to make the two places consistent, I settled on UTC... so 
it's at 
least confirmed the bug, but not really a "fix". 

I think that's enough for one night, though... I'll come back to this another 
time. 

Thanks again for your help with this.

Original comment by dale.l...@gmail.com on 7 Dec 2009 at 10:12

• Changed state: Started

[GoogleCodeExporter]

mine works now, I'm in Australia

Original comment by rolandpa...@gmail.com on 8 Dec 2009 at 10:02

[GoogleCodeExporter]

in addition to the graph working for "live", I do note that the time scale is 
wrong, 
the power monitor shows 2102, but the graph shows 1006

Original comment by rolandpa...@gmail.com on 8 Dec 2009 at 10:06

[GoogleCodeExporter]

After spending much time investigating this, I can't figure out how to work out 
what 
timezone the app is being run in - sorry. 

For now, I'm leaving it displaying the live graph data using UTC timestamps. 
This at 
least means that the graph line itself is visible, which is an improvement on 
the 
original bug. 

If anyone else wants to work out how to do it, the source - as always - is all 
at 
http://code.google.com/p/currentcostgui/source/browse/#svn/trunk 

Original comment by dale.l...@gmail.com on 8 Dec 2009 at 5:52

• Changed state: Fixed

[GoogleCodeExporter]

Hi Dale - I'm interested why your code needed to change to be worried about time
zones in the first place. Prior to 0.9.29 it seemed to be working just fine - 
even in
Oz! ;-)

John

Original comment by john.h.m...@gmail.com on 11 Dec 2009 at 6:24

[GoogleCodeExporter]

Python has two types of datetime object - timezone-aware and timezone-naive. 

I was using timezone-naive objects.

I'm now using timezone-aware objects. I needed timezone-aware objects to be 
able to do the comparisons 
needed to implement the feature for 
[http://code.google.com/p/currentcostgui/wiki/MeasuringEnergyUsage 
measuring area under the live graph curve].

Original comment by dale.l...@gmail.com on 11 Dec 2009 at 8:18

[GoogleCodeExporter]

Hi - I'm pretty sure you will have already found these - but just in case, here 
are a
few useful discussions about timezones in Python (which do seem to be an 
enormous
pain!).  Unfortunately, I'm not a Python man otherwise I would weigh-in & give 
you a
hand.

http://scienceoss.com/timezones-in-python/
http://www.python-forum.org/pythonforum/viewtopic.php?f=1&t=10040
http://bugs.python.org/issue1647654

John

Original comment by john.h.m...@gmail.com on 14 Dec 2009 at 1:24

[GoogleCodeExporter]

Thanks for those - I had already read them. The problem is that creating a 
datetime 
object for timezone X is do-able. Knowing that you're in timezone X is 
seemingly the 
hard bit. 

Original comment by dale.l...@gmail.com on 15 Dec 2009 at 4:10