Addresses issue of allowing player to go beyond number of actual iterations. Also introduces code to automatically enable/disable player buttons depending on the state of the animation.

Addresses #66.

Let me know if this fix isn't surgical enough. And feel free to nitpick.

Small design decisions:

• I didn't add new WebGL detection logic. That would require creating a dummy canvas, etc.
• The error message comes from near the point of failure, rather than assuming we know what went wrong in main.js.
• The failure is localized to the container that fails to get populated, rather than leaving the exception uncaught. This helps avoid the red herring error messages mentioned in the original issue.
• I didn't try to create a super-user-friendly experience. It just shows the error message in place of the canvas. Good enough for hackers.
• To avoid duplicated try/catch for each visualization, and to avoid hypothetical broken scenarios where one visualization succeeds and the other fails:
  • I created a helper function (addVisualization) with success callbacks.
  • I store an array of historyConsumers rather than having one var per consumer (formally visualization2D, visualization3D).
  • I change the GUISync so that the master can be specified outside of the constructor, so that the main.js code doesn't trip over itself if one of the visualizations fails to create.

Not sure how hard this will be, but it would great to have some buttons that will re-position the camera to pre-defined positions, such as front, top, left, right, bottom. Often times, when manipulating the view, the model can get skewed, or off screen. Clicking one of these buttons would:

• Immediately re-position the model and/or camera so that the view is directly from that position (e.g. from the front), with edge aligned with the bounding box of the view.
• Resize the model and/or camera so that the entire model fits within the view.

Another control that might be nice to have would be a camera focal length slider.

Fixes #61.

Attention: @chetan51

Namespaced all global objects under Cerebro2. Also wrapped main.js code in anonymous functions to prevent global namespace pollution. This fixes the issue described in numenta/nupic.cerebro2#61 where, in the Chrome browser, the "history" object cannot be overwritten.

Hey all,

everytime I try to run cerebro2 (with cerebro2.server set up) I get a list of errors like (excerpt):

127.0.0.1:47916 - - [21/Aug/2014 17:45:35] "HTTP/1.1 GET /_model/states/1/input/active_cells" - 500 Internal Server Error
 127.0.0.1:47919 - - [21/Aug/2014 17:45:35] "HTTP/1.1 GET /_model/states/1/output/predicted_cells" - 200 OK
127.0.0.1:47917 - - [21/Aug/2014 17:45:35] "HTTP/1.1 GET /_model/states/1/output/active_columns" - 500 Internal Server Error
127.0.0.1:47921 - - [21/Aug/2014 17:45:35] "HTTP/1.1 GET /_model/states/1/input/active_cells" - 500 Internal Server Error
Traceback (most recent call last):
  File "/usr/local/lib/python2.7/dist-packages/web/application.py", line 239, in process
    return self.handle()
  File "/usr/local/lib/python2.7/dist-packages/web/application.py", line 230, in handle
    return self._delegate(fn, self.fvars, args)
  File "/usr/local/lib/python2.7/dist-packages/web/application.py", line 420, in _delegate
    return handle_class(cls)
  File "/usr/local/lib/python2.7/dist-packages/web/application.py", line 396, in handle_class
    return tocall(*args)
  File "/home/mark/nupic.cerebro2.server/api/model_api.py", line 46, in GET
    return jsonResponse(readJSON(filepath))
  File "/home/mark/nupic.cerebro2.server/api/model_api.py", line 68, in readJSON
    raise error
IOError: [Errno 2] No such file or directory: u'/tmp/cerebro2/model/states/1/output/active_columns.json'

127.0.0.1:47920 - - [21/Aug/2014 17:45:35] "HTTP/1.1 GET /_model/states/1/output/active_cells" - 200 OK
 127.0.0.1:47916 - - [21/Aug/2014 17:45:35] "HTTP/1.1 GET /_model/states/1/output/predicted_cells" - 200 OK
Traceback (most recent call last):
  File "/usr/local/lib/python2.7/dist-packages/web/application.py", line 239, in process
    return self.handle()
  File "/usr/local/lib/python2.7/dist-packages/web/application.py", line 230, in handle
    return self._delegate(fn, self.fvars, args)
  File "/usr/local/lib/python2.7/dist-packages/web/application.py", line 420, in _delegate
    return handle_class(cls)
  File "/usr/local/lib/python2.7/dist-packages/web/application.py", line 396, in handle_class
    return tocall(*args)
  File "/home/mark/nupic.cerebro2.server/api/model_api.py", line 46, in GET
    return jsonResponse(readJSON(filepath))
  File "/home/mark/nupic.cerebro2.server/api/model_api.py", line 68, in readJSON
    raise error
IOError: [Errno 2] No such file or directory: u'/tmp/cerebro2/model/states/1/output/proximal_synapses.json'

Do you have any idea what the problem might be? Cheers, Mark

Added camera controls to position camera in various default views. As a part of this, added an equation to calculate a reasonable distance for the camera based on the size of the model. All of this necessitated some modifications to the display of the dat.GUI elements in order to make room for the extra buttons.

Opted not to include a Field of View slider. While this is possible, the way a particle system is built, caused the size of the particles to scale strangely, and there did not seem to be a great benefit in offering the feature. If there is interest in this, I can add the code to enable this easily.

8080 is the default port for a web.py application, so we should point to that from the Cerebro 2 client instead. That way, just running python server.py in Cerebro 2 server will work, rather than having to run python server.py 9090.

Also update Cerebro 2 server README to say just python server.py instead of python server.py 9090.

I am getting this error after my most recent merge from upstream. It looks like something might have changed recently with the history object? Length no longer seems to be a method? I will try to look into this more.

In order to keep the slider up to date, we remove the slider and re-instantiate it after each snapshot. This could possibly cause problems in situations in which there is a very long stream, and the snapshot updates are happening very rapidly, and the user tries to move the slider as it is being updated, but this seems to work pretty well in my local tests. The only other way to approach this seemed to be to modify dat.GUI code directly, to allow for a dynamic change to be made to the max value of the slider.

Under the Iteration scrubber, there should be a Play button that automatically steps through iterations with a certain speed. Under the Play button (disabled if not playing) should be a scrubber that controls the playback speed.

@chetan51 Is cerebro obsoleted?

@david-ragazzi are you reusing the code? or providing same functionality?

or is there a chance of merging the two projects?

We need automated unit tests for Cerebro2. Questions:

• What should our code coverage target be?
• What framework to use? Jasmine? Mocha?
• Will coveralls.io work with this project?
• Do we want other types of tests, such as end-to-end tests?

Setup a build flow, and use something like RequireJS [1] to combine all the necessary Javascript files into one minified file, so that users can just do

<script src="cerebro2.min.js"></script>

to embed Cerebro2 in their webpage.

[1] http://requirejs.org/docs/optimization.html#wholeproject

Move controller-manipulation functions from AbstractVisualization into a separate, utility file (maybe called dat.gui.utils.js), so they're not causing clutter.

@jefffohl

I find myself often wanting to reset the animation to the beginning. It would be nice to have a "first" button in the iteration controls that resets to the first iteration.