Similar clusters stallling

[midlewis]

When trying to cluster blobs by similarity, the code either takes a very long time to run or needs a fix. Screenshot attached.

[jcgiuffrida]

I think I may know what is happening. This line does not take very long at all - but it directs to the Python program (if you use a Mac) and doesn't finish "executing" until that window is closed. It doesn't bring that window to the front, either, so you have to ALT-TAB over. In an IDE this is probably more seamless.

[midlewis]

Thanks Jonathan I'll try that!

Sent from Cyberspace

On Jun 30, 2015, at 10:18 PM, Jonathan Giuffrida notifications@github.com wrote:

I think I may know what is happening. This line does not take very long at all - but it directs to the Python program (if you use a Mac) and doesn't finish "executing" until that window is closed. It doesn't bring that window to the front, either, so you have to ALT-TAB over. In an IDE this is probably more seamless.

—
Reply to this email directly or view it on GitHub.

[midlewis]

Jonathan - I tried your idea, but there are no other screens open. Is there another IDE that you would recommend I try it with?

[brettjgoldstein]

Need to make sure this isn't ide sensitive

On Sun, Jul 5, 2015 at 1:20 PM, midlewis notifications@github.com wrote:

Jonathan - I tried your idea, but there are no other screens open. Is there another IDE that you would recommend I try it with?

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/1#issuecomment-118648085.

[Makosak]

When I tried this before, it only worked in qtconsole using anaconda's environment, so it did seem sensitive to other options. I think part of this was the matlib-plot function required. Alternatively, the issue here could be that matlib-plot isn't updated in the user environment?

[Makosak]

Okay I think I figured it out, as I was able to replicate Shelly's error (not intentionally, ha).

Make sure you:

And then I had to create an account and get an API key with Plot.ly:
https://plot.ly/settings/api

And within iPy console, enter the following. You'll need to replace 'DemoAccount' and 'lr1c37zw81' with your Plotly username and API key.
```import plotly; plotly.tools.set_credentials_file(username='DemoAccount', api_key='lr1c37zw81')

Then, when I attempted the blobs code again using the notebook tutorial, the cluster code worked quickly:
```cl = blobs.Cluster_blobs(b, blobs_per_cluster=10)

<img width="766" alt="screen shot 2015-07-13 at 12 42 03 pm" src="https://cloud.githubusercontent.com/assets/4277692/8656349/9d601698-295c-11e5-82d9-f5474783e515.png">

Resources: for the second half, https://plot.ly/python/getting-started/

[Makosak]

Udpate: it's not working again. May be best to focus on separating viz and cluster assignment and focus on new viz method.

[jcgiuffrida]

Probably best to have Cluster_blobs() not automatically call self.plot() - that way the user explicitly asks it to plot, and knows if that's what the problem is or not. We can definitely separate more of the visualization from the blobs algorithm to make this less platform-dependent.

[midlewis]

I was running this program on a Mac 10.7 OS, now that I upgraded it to 10.10 I can successfully run it in the Mac terminal and in iPython. The only suggestion I have is to change the line: b.assignments to b.regions which I believe Jonathan may have already been in touch with you both about.

[Makosak]

When and if it doesn't work, this is when the application stops without crashing, this seems to be the issue (& this code only shows up in terminal):

cl = blobs.Cluster_blobs(b, blobs_per_cluster=10)
python(30561) malloc: *** error for object 0x11150f000: pointer being freed already on death-row
*** set a breakpoint in malloc_error_break to debug
python(30555) malloc: *** error for object 0x10ae1bb58: incorrect checksum for freed object - object was probably modified after being freed.
*** set a breakpoint in malloc_error_break to debug

Potential solution:
http://stackoverflow.com/questions/14045208/how-to-set-a-breakpoint-in-malloc-error-break-to-debug

I'm not going to mess with that too much though, per our last discussion, instead focusing on other options, being happy that it worked for Shelly, and so forth.