Re: [Matplotlib-users] how would you do this (animated bargraph) (Neal Becker)

[Neal B. <ndb...@gm...>]

David Hoese wrote:

> Neal,
> 
> I do something similar to this where data that I'm plotting in a 2D line
> plot comes from a UDP socket and some memory mapped files.  To
> accomplish the live updating I have a mix of your #2 and #3.  I have a
> main GUI thread that displays the plots, then I have a second thread
> that gets data from a python generator and signals the GUI thread when
> that new data has arrived.  Inside the generator is where I have socket
> communications and select calls.  Hope this helps.
> 

Haven't tried it, but I did notice QSocketNotifier, which sounds like it may be 
what I need if I want to hook into qt event loop to implement as single-thread, 
single-process.

> -Dave
> 
> On 9/30/11 9:10 AM,
> mat...@li...
> wrote:
>> Message: 4
>> Date: Fri, 30 Sep 2011 07:37:49 -0400
>> From: Neal Becker<ndb...@gm...>
>> Subject: [Matplotlib-users] how would you do this (animated bargraph)
>> To:mat...@li...
>> Message-ID:<j649me$ug4$1...@do...>
>> Content-Type: text/plain; charset="ISO-8859-1"
>>
>> I just put together an animated bargraph to show results from a realtime
>> process.
>>
>> I used this as an example:
>> http://matplotlib.sourceforge.net/examples/animation/animation_blit_qt4.html
>>
>> The tricky part for me was that in the original design, there was a realtime
>> process running.  I have to write some information to my hardware (at a rate
>> of
>> about 1/360ms).  To do this, I have a kernel driver that uses 'read' to tell
>> the user space when an interrupt has occurred, which is when the user space
>> should write new information to the hardware.
>>
>> So I don't know how or if this could be hooked into qt event processing. 
>> Just for a quick-and-dirty demo, I just removed the realtime processing from
>> the user-space and put it in the kernel driver, so now my bargraph display
>> can simply update on a periodic timer, and the userspace code has no realtime
>> requirement.  But this is just a kludge.
>>
>> So I wonder how other's would solve this?  I'm thinking it would be either:
>>
>> 1) multi-process
>> 2) multi-thread
>> 3) 1 process, but somehow hook my realtime events into qt's event loop.
>>
>> For #3, my device driver has a filedescriptor, that could use select to
>> detect the interrupt (rather than blocking read call).
>>
>> #1 and #2 seem complicated.
>