craneium.net

brain matters . . .

  • Increase font size
  • Default font size
  • Decrease font size
Home Python Programming
Python Programming

Open Letter to Leap Motion about Linux

E-mail Print PDF

Dear Misty and broader LEAP community,

I must admit, I am a really annoyed.  I am a software developer who primarily work on image and video processing tools on the Linux platform.  Over the past 8 months I have waited patiently for any word on a Linux SDK for the Leap motion device, or even a hint of progress.

I found none, and hence I decided to move forward with what I have done in the past when waiting on a closed source piece of software.  Figure it out, code it up, and make something awesome that others can use.

When I discovered this morning that my post on your forums had been removed (notably w/o any actual notification to me or others), and another thread asking what happened was somewhat harshly dismissed with a resounding "refrain from commenting on moderation, as this is against the forum rules" I really felt put-down.  

I appreciate your restoration of the thread, but it already feels like Leap is being hostile toward the very users that will enable their platform.  
The purpose of this letter is not just to demonstrate how off-put I am by this, but to more to explain that if you are going to continue to ignore the Linux community (more or less blatantly)  it will find its own way to integrate your device, likely outside your control, just as was done with libfreenect.

If you will not at least update us, expect us to take our own course, silence is a indicator that Linux is not a priority.

I would love to create things with you and thanks.

--Matthew Goodman


=====================
Check Out My Website: http://craneium.net
Find me on LinkedIn: http://tinyurl.com/d6wlch
Last Updated on Wednesday, 06 March 2013 22:18
 

Installing New iPython with Parallel Compute Based on ZMQ

E-mail Print PDF

Step 0 - Get rid of your current ipython.

There are many places this can lurk but running:

sudo apt-get remove ipython

on ubuntu might help you in this goal

First get ZeroMQ:

By repository:

http://johanharjono.com/archives/633

Or from source:

sudo apt-get install uuid-dev g++

wget http://download.zeromq.org/zeromq-2.1.7.tar.gz

tar xfz zeromq-2.1.7.tar.gz

cd zeromq-2.1.7

./configure

make

sudo make install

sudo ldconfig

 

Build pyzmq:

git clone https://github.com/zeromq/pyzmq.git

cd pyzmq

python setup.py config

python setup.py build

sudo python setup.py install

 

Build iPython:

git clone https://github.com/ipython/ipython.git

cd ipython

python setup.py config

(make sure it finds pyzmq.  Mine reports "pyzmq: 2.1dev")

python setup.py build

sudo python setup.py install

 

Check it out:

ipython-qtconsole

 

Last Updated on Wednesday, 17 August 2011 06:39
 

matplotlib/pylab plotting SSH tunnels and backends

E-mail Print PDF

I recently discovered that tunneled connections using matplotlib force X11 to cram the entire figure across a slow SSH connection, even if the figure will never be rendered!  To fix this, simuply use the backend that requires no X11 support:

import matplotlib

matplotlib.use('Agg')

from matplotlib.pyplot import (plot, figure, psd, hist, savefig, close, title, xlabel, ylabel, axis)

from matplotlib.mlab import detrend_linear

This way there is (vitrually) no SSH traffic overhead to plotting and saving figures!!  Yay!  Enjoy!

 

Compiling yourself a Scientific Python Paradise on EC2

E-mail Print PDF

First start by following the Starcluster directions found here:

http://starcluster.scripts.mit.edu/~starcluster/wiki/index.php?title=StarCluster_AMI_Cookbook_Ubuntu_10.04

Notes:

  1. The AMI they reference no longer exist, I used the 64 bit Ubuntu Maverick from here: http://alestic.com/
  2. There are much newer versions ipython, so leave that out of the apt-get install process.  You can download that later at your leisure.
  3. When you are reenabling the root login, removing prefix commands means everything before rsa- . . . .
  4. Also, on the root login section, the line number 67 is no longer valid for current versions.  I hit the one on line number 86, and haven't had problems.
  5. Given the merge of Sun and Oracle, you will need to find the Sun (cough), sorry, Oracle Grid engine, which is highly hidden in the oracle website.  Here are the keys you need to find it: Get to the downlod section, and look in the "Oracle Enterpries Manager" product pack.  "Grid engine media pack."  Thanks to their nazi download system, it is unlikely you can retrieve this to your ec2 instance, instead download it locally, then scp it over. . . .  Weak. At 820 megs, this is more than a little bit annoying.
  6. To get the SGE and drmaa v0.2 (why is this needed?)  Just boot a starcluster instance and pirateer the /opt directory to your new instance.  You can get the old version of drmaa this way too.
  7. The ubuntu package libopenmpi-dev now points at openmpi.  This is fine.  It all works just like the directions say.  I ignored the ubuntu patch for openmpi, looks like it fixes some obscure bugs.

 

Last Updated on Wednesday, 19 January 2011 00:02
 

What I don't Like About MATLAB

E-mail Print PDF

This is a place that I list the various things I come across in MATLAB that annoy me:

Major (Systematic Problems):

  1. Functions must be separate files.  This is not fundamentally a bad thing, but it discourages the declaration of small functions which make code clearer.
  2. Lack of development headers.  Python lets you tear down what you don't like and do as you please.  This is mostly a bad idea, but I would rather have the freedom to blow off my own leg than live without the high-power firearms associated with messing with the interpreter.  See IPython for beautiful examples, and for some bad ones see the awful GOTO/COMEFROM module.
  3. Lack of a linked list class in the default install.  This is really glaring.  There are a ton of problems that are conducive extensible lists of objects.
  4. Lack of clean integration with other languages.   This process is simply a disaster.  Ever tried making anything in c-linked matlab work properly between platforms/compilers?  Don't.  This problem is distinctly linked to problem #2.

 

Minor Annoyances:

  1. The lack of a true outer product.  There is no way to (in one step) project together the dimensions of arbitrary arrays with matching inner dimensions.  AKA The product of a n-dimensional array A with shape (ExFxGx1)  and B ( H ) to get outer(A, B) with shape (ExFxGxH).  For something with such a strong linear algebra focus, this is just really weak sauce.
  2. One based indexing.  Fortran used 1 based indexing.  Need I say more?
  3. MATLAB complains (warnings) when you allocate temporary values.  This leads to really, really, ugly syntax which has to create temporaries anyway . . . the warnings might as well say something like "Warning:Readable code ahead!"

 

Last Updated on Wednesday, 15 December 2010 17:52
 
  • «
  •  Start 
  •  Prev 
  •  1 
  •  2 
  •  3 
  •  Next 
  •  End 
  • »


Page 1 of 3