Plotting complex polynomials with octave and octaviz (vtk)

I got to wondering what happens when you feed complex numbers into polynomials, so I poked around and found the awesome vtk library binding for the octave language, called octaviz.  (octave is a free implementation of the MATLAB language).

When we plot z = f(x), where x and z are complex numbers, we get 4 dimensions to plot, so I shifted each point in the surface plot N units + or - in the Y direction, where N is the imaginary part of the output.  The color also indicates how far each point was stretched in the + or - Y direction. 

#!/usr/bin/octave

global a = 1
global b = 0
global c = 0

function z = polynomial (i,j)
  global a
  global b
  global c
  x = i + (j * 1j);
  z = a*x*x + b*x + c;
end

[I,J] = meshgrid(-1:0.2:1);
REAL = arrayfun(@real, arrayfun(@polynomial, I, J));
IMAG = arrayfun(@imag, arrayfun(@polynomial, I, J));
vtk_surf(I,J+IMAG,REAL, IMAG);

input ("Orient it the way you want it before I save.");
vtk_print('polynomial.jpg', '-djpeg');
input ("Done!");

Borges

The stoics teach that we should not complain of life--the door of the prison is open.
-august 25, 1983

Things I hate: externalized costs in programming

Bob adds a feature to handle a situation that someday might arise. Alice then discovers a production failure caused by that feature. Good luck removing the feature: doing that might cause a production failure someday!

python-tk: tkinter photo doesn't work inside a function

Wow, python-tk sucks. When I create the photo inside a function, it gets garbage-collected because it's not smart enough to know that the label is using it. So I have to manually keep it from being deallocated (in this case, by making it global). Also, I'd really like to use after_idle instead of after, but it never gets around to actually updating the display.

#!/usr/bin/python

import Tkinter
import ImageTk,Image

root = Tkinter.Tk()
root.geometry('+640+480')

old_label = None
label = None
photo = None
def display_image(root, image):
  global old_label
  global label
  global photo

  root.geometry('%dx%d' % (image.size[0], image.size[1]))

  photo = ImageTk.PhotoImage(image)

  label = Tkinter.Label(root, image=photo)
  label.place(x=0,y=0,width=image.size[0],height=image.size[1])

  if old_label is not None:
    old_label.destroy()

  old_label = label

def image_loop():
  global root
  print '.'
  image = get_image()
  display_image(root, image)
  root.after(1000,image_loop)

image_loop()

root.mainloop()

python plumbing: PIL Image from data string

Today I wanted to use netcat (nc) to send a JPEG image file over a network to a python script so that I can manipulate it with PIL, the python imaging library.

PIL has fromstring and frombuffer methods, but I couldn't get them to work. But I found the ImageFile module, which let me feed the data in as it came off the socket.

To serve up the images, I did this from the shell:

while true ; do nc -q 0 -l -p 12345 < foo.jpg  ; done

Then this python script received and displayed an image:

#!/usr/bin/python

import socket
from PIL import ImageFile

tcp = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
tcp.connect(('127.0.0.1', 12345))
tcp.send("1000 4")

file = open("bar.jpg", "w")
parser = ImageFile.Parser()

while 1:
  jpgdata = tcp.recv(65536)
  if not jpgdata:
    tcp.close()
    break

  parser.feed(jpgdata)
  file.write(jpgdata)

file.close()
image = parser.close()
image.show()

Xorg window redraw: workaround by forcing update

When using Sketchup in wine on my ubuntu linux machine, I find that it doesn't update the screen properly. For instance, if I hit control-A to select everything, I don't see the selection until I do something else. That is, the screen seems to be one update behind.

Here's a workaround that works with Sketchup: in a separate terminal, I run:

$ watch -n 0.3 xrefresh -geometry 1x1+500+500

That uses xrefresh to create a 1x1 pixel x11 window at 500,500 on the screen every 0.3 seconds. That's enough to get sketchup to redraw itself.

If that's not enough, you can run xrefresh without the -geometry option to have it draw a window the size of the entire screen. But that's a lot more distracting.

libgphoto2 / gphoto2 build from svn and install locally

Today I wanted to make a patch against gphoto2 (the command-line interface to libgphoto2) without overwriting the version installed by ubuntu.

So I needed to download libgphoto2 and gphoto2 from subversion and then install it to a local directory.

$ sudo apt-get install automake autoconf libtool gettext pkg-config subversion build-essential
$ mkdir gphoto-svn
$ cd gphoto-svn
$ svn checkout https://gphoto.svn.sourceforge.net/svnroot/gphoto/trunk/libgphoto2
$ cd libgphoto2
$ autoreconf -is
$ mkdir -p /tmp/gphoto2/local
$ ./configure --prefix=/tmp/gphoto2/local
$ make
$ make install
$ # neat, libgphoto2 installed to /tmp/gphoto2/local
$ cd ..
$ svn checkout https://gphoto.svn.sourceforge.net/svnroot/gphoto/trunk/gphoto2
$ cd gphoto2
$ autoreconf -is
$ ./configure --prefix=/tmp/gphoto2/local --with-libgphoto2=/tmp/gphoto2/local
$ make
$ # make install is optional; I just ran ./gphoto2/gphoto2 directly
$ # edited files, then "make" to rebuild