/dev/stu

Posted by stu at May 17th, 2010

Seeing”Realtime FFT Graph of Audio WAV File or Microphone Input with Python…”  on python.reddit.com reminded me of one I’d built in python with shoebot.

While it works OK, I feel like I’m missing a higher level audio library (especially having seen Minim, for C++ and Java).

To run it in shoebot:

sbot -w audiobot.bot

audiobot.bot

# Major library imports
import atexit
import pyaudio
from numpy import zeros, short, fromstring, array
from numpy.fft import fft

NUM_SAMPLES = 512
SAMPLING_RATE = 11025

def setup():
    size(350, 260)
    speed(SAMPLING_RATE / NUM_SAMPLES)

_stream = None

def read_fft():
    global _stream
    pa = None

    def cleanup_audio():
        if _stream:
            _stream.stop_stream()
            _stream.close()
        pa.terminate()

    if _stream is None:
        pa = pyaudio.PyAudio()
        _stream = pa.open(format=pyaudio.paInt16, channels=1,
                           rate=SAMPLING_RATE,
                           input=True, frames_per_buffer=NUM_SAMPLES)
        atexit.register(cleanup_audio)

    audio_data  = fromstring(_stream.read(NUM_SAMPLES), dtype=short)
    normalized_data = audio_data / 32768.0

    return fft(normalized_data)[1:1+NUM_SAMPLES/2]

def flatten_fft(scale = 1.0):
    """
    Produces a nicer graph, I'm not sure if this is correct
    """
    for i, v in enumerate(read_fft()):
        yield scale * (i * v) / NUM_SAMPLES

def triple(audio):
    '''return bass/mid/treble'''
    c = audio.copy()
    c.resize(3, 255 / 3)
    return c

def draw():
    '''Draw 3 different colour graphs'''
    global NUM_SAMPLES
    audio = array(list(flatten_fft(scale = 80)))
    freqs = len(audio)
    bass, mid, treble = triple(audio)

    colours = (0.5, 1.0, 0.5), (1, 1, 0), (1, 0.2, 0.5)

    fill(0, 0, 1)
    rect(0, 0, WIDTH, 400)
    translate(50, 200)

    for spectrum, col in zip((bass, mid, treble), colours):
        fill(col)
        for i, s in enumerate(spectrum):
            rect(i, 0, 1, -abs(s))
        else:
            translate(i, 0)

    audio = array(list(flatten_fft(scale = 80)))

Posted by stu at May 13th, 2010

An update on my latest cairo adventures…

When cairo 1.10 comes out we’ll get a RecordingSurface so you can record commands and play them back to another surface, but how to do something similar now ?

Skip to the end if you just want the code, explanation of how I got there ahead:

My first advice was to try using PDFSurface, and set the file object to None.

# Record a red rectangle onto a surface.
#
# Create another surface and draw a background on it
# Draw the first surface onto this surface
#
from cairo import PDFSurface, ImageSurface, FORMAT_ARGB32, Context

recording = PDFSurface(None, 200, 200)
target = ImageSurface(FORMAT_ARGB32, 200, 200)

# Record a red rectangle
cr = Context(recording)
cr.set_source_rgb(1.0, 0.0, 1.0)
cr.rectangle(20, 20, 10, 10)

cr.fill_preserve()
cr.set_source_rgb(0.0, 0.0, 1.0)
cr.stroke()

target_context = Context(target)

# Draw background
target_context.set_source_rgb(1.0, 1.0, 0)
target_context.paint()

# Replay recording to target
target_context.set_source_surface(recording)
target_context.paint()

target.write_to_png('output.png')

That seems to work, except when I tried in Windows, when it complained that the file object was wrong.

OK, we can work round that:

def RecordingSurface(w, h):
    if os.name == 'nt':
        fobj = 'nul'
    else:
        fobj = None
    return PDFSurface(fobj, w, h)

This seems to be working, but my animation seemed slow… time for some benchmarking; I rendered 1000 frames and got a rough wall clock time:
1m48.

Hmm… perhaps SVGSurface will be quicker:
1m28

This is much better, 20 seconds difference just by changing what kind of surface is returned!

Animation not smooth though

The animation still seemed jerky it occured to me that when the surfaces are disposed they will attempt to out their content to the file object !

Luckily, get_similar_surface() comes to the rescue; it returns a surface not associated with a file object. Using this the original surface can be kept around forever, and never output.

Wallclock time:
50 seconds!

And here it is:

_svg_surface = None
def RecordingSurface(*size):
    '''
    We don't have RecordingSurfaces until cairo 1.10, so this kludge is used

    SVGSurfaces are created, but to stop them ever attempting to output, they
    are kept in a dict.

    When a surface is needed, create_similar is called to get a Surface from
    the SVGSurface of the same size
    '''
    global _svg_surface
    if os.name == 'nt':
        fobj = 'nul'
    else:
        fobj = None
    if _svg_surface is None:
        _svg_surface = SVGSurface(fobj, 0, 0)
    return _svg_surface.create_similar(cairo.CONTENT_COLOR_ALPHA, *size)

This is really useful, you can record commands and play them back to other surfaces.

Posted by stu at May 12th, 2010

Often it’s useful to be able to store up drawing commands so you can use them later somewhere else (or even just pass them to another thread).

This is a simple drawing model, implemented in cairo, hopefully somebody will find it useful.

Queue

class DrawQueue:
    '''
    A list of draw commands, stored as callables that, are
    passed a set of parameters to draw on from the canvas
    implementation.
    '''
    def __init__(self, render_callables = None):
        self.render_callables = render_callables or deque()

    def append(self, render_callable):
        '''
        Add a render callable to the queue
        '''
        self.render_callables.append(render_callable)

    def render(self, cairo_ctx):
        '''
        Call all the render callables with cairo_ctx
        '''
        for render_callable in self.render_callables:
            render_callable(cairo_ctx)

The queue just accepts callables (any old function), and calls them when you call render, passing them a cairo context you pass in.

To get useful functions you can call closure functions like these:

def paint_closure():
    def paint(ctx):
        ctx.paint()
    return paint

def fill_closure():
    def fill(ctx):
        ctx.fill()
    return fill

def set_source_rgb_closure(r, g, b):
    def set_source_rgb(ctx):
        ctx.set_source_rgb(r, g, b)
    return set_source_rgb

def moveto_closure(x, y):
    def moveto(ctx):
        ctx.move_to(x, y)
    return moveto

def rectangle_closure(x, y, w, h):
    def rectangle(ctx):
        ctx.rectangle(x, y, w, h)
    return rectangle

Adding commands to the queue is simple:

dq = DrawQueue()
dq.append(set_source_rgb_closure(1, 1, 1))
dq.append(paint_closure())
dq.append(moveto_closure(10, 0))
dq.append(rectangle_closure(0, 0, 20, 20))
dq.append(set_source_rgb_closure(0, 0, 0))
dq.append(fill_closure())

This is the same drawing model I’m using in my branch of shoebot, I’m hoping to expand it to be multithreaded; while a foreground thread adds commands a background thread is executing them.

Here it is all put together in a simple example to draw a black rectangle

from collections import deque
import cairo

class DrawQueue:
    '''
    A list of draw commands, stored as callables that, are
    passed a set of parameters to draw on from the canvas
    implementation.
    '''
    def __init__(self, render_callables = None):
        self.render_callables = render_callables or deque()

    def append(self, render_callable):
        '''
        Add a render callable to the queue
        '''
        self.render_callables.append(render_callable)

    def render(self, cairo_ctx):
        '''
        Call all the render callables with cairo_ctx
        '''
        for render_callable in self.render_callables:
            render_callable(cairo_ctx)

#### drawing closures
def paint_closure():
    def paint(ctx):
        ctx.paint()
    return paint

def fill_closure():
    def fill(ctx):
        ctx.fill()
    return fill

def set_source_rgb_closure(r, g, b):
    def set_source_rgb(ctx):
        ctx.set_source_rgb(r, g, b)
    return set_source_rgb

def moveto_closure(x, y):
    def moveto(ctx):
        ctx.move_to(x, y)
    return moveto

def rectangle_closure(x, y, w, h):
    def rectangle(ctx):
        ctx.rectangle(x, y, w, h)
    return rectangle

#### /drawing closures

dq = DrawQueue()

# Add some commands to the drawing queue
dq.append(set_source_rgb_closure(1, 1, 1))
dq.append(paint_closure())
dq.append(moveto_closure(10, 0))
dq.append(rectangle_closure(0, 0, 20, 20))
dq.append(set_source_rgb_closure(0, 0, 0))
dq.append(fill_closure())

# Create a surface and context
surface = cairo.ImageSurface(cairo.FORMAT_ARGB32, 200, 200)
ctx = cairo.Context(surface)

# run defered rendering
dq.render(ctx)

surface.write_to_png('output.png')

Posted by stu at February 10th, 2010

Update: 26/April/2010

Problems I was having with incomplete images have been fixed in the current version of the web library, available in shoebots mecurial repository.

Using Processing from Jython is a promising idea, so I took the base from this post on backspaces.net where they explained how to use Jython and built on it a little.

This is great as Shoebot/Nodebox have great libraries for data manipulation, while processing is more focused on graphics.

The result is the attached Netbeans project which demonstrates using the nodebox web library and  drawing with processing.

Slowcessing

The glue code is in slowcessing.py

Theres a special version of PApplet (PJApplet), and ‘pj_frame’ which can put this in a JFrame.

The other method is ’shoebot_imports’ adds the shoebot imports to the library path

In case anybody doesn’t want to download the whole project, heres the code:

imagestrip.py

from slowcessing import PJApplet, pj_frame, shoebot_imports
from processing.opengl import *

shoebot_imports()
import web
import thread

class ImageQueue(list):
    """
    Download images in the background and add them to a list
    """
    def __init__(self, search, size):
        list.__init__(self)
        self._search = search
        self._image_size = size
        thread.start_new_thread(self._get_images, ())

    def _image_downloaded(self, path):
        p = PJApplet()
        self.append(p.loadImage(path))

    def _get_images(self):
        for image in self._search:
            image.download(self._image_size, asynchronous=False)
            self._image_downloaded(image.path)

class WebTest (PJApplet):
  def setup(self):
    self.size(400, 400, self.P3D)
    self.images = ImageQueue(web.morguefile.search("sweets", max=1), size='small')

  def draw(self):
    self.background(0);
    y = (self.height * 0.2) - self.mouseY * (len(self.images) * 0.58)
    for image in self.images:
        self.image(image, 20, y)
        y += image.height

if __name__ == '__main__':
    pj_frame(WebTest)

slowcessing.py

from javax.swing import JFrame

from processing.core import PApplet

class PJApplet(PApplet):
  # rqd due to PApplet's using frameRate and frameRate(n) etc.
  def getField(self, name):
      return PApplet.getDeclaredField(name).get(self)

def pj_frame(pj_applet, **kwargs):
    from time import sleep

    frame = JFrame(kwargs.get('title', 'Slowcessing'))
    frame.defaultCloseOperation = kwargs.get('defaultCloseOperation', JFrame.EXIT_ON_CLOSE)
    frame.resizable = kwargs.get('resizable', False)

    panel = pj_applet()
    frame.add(panel)
    panel.init()

    while panel.defaultSize and not panel.finished:
        sleep(0.5)

    frame.pack()
    frame.visible = 1

    return frame

def shoebot_imports():
    """
    Allow import of the shoebot libraries
    """
    ##APP = 'shoebot'
    import sys
    DIR = sys.prefix + '/share/shoebot/locale'
    ##locale.setlocale(locale.LC_ALL, '')
    ##gettext.bindtextdomain(APP, DIR)
    ###gettext.bindtextdomain(APP)
    ##gettext.textdomain(APP)
    ##_ = gettext.gettext

    LIB_DIR = sys.prefix + '/share/shoebot/lib'
    sys.path.append(LIB_DIR)

Problems

There are some things I couldn’t work :

The callback to say that images have been downloaded happens before the whole file is available, for this reason there are grey parts on the images on the first run.

Nodebox web…

While I did manage to fix things to get this working in Jython and get Morguefile working, I had a lot of trouble understanding what was going on here.

Cheers to Tom De Smedt for fixing these the areas of nodebox-web that I couldn’t

Processing…

Some parts of PApplet to do with image loading seem to be static, which may also explain problems I was getting with reentrancy.

Download

If you want to have a go, you’ll need to:

Install Netbeans 6.8

Install Jython (2.5 or higher) by installing the Netbeans python module

Add python to the path (if using Netbeans it’s copy is where Netbeans is installed).

Get nodebox-web by downloading shoebot and install it with:

jython setup.py install

In Netbeans, add all the jars in the processing\lib folder to the Jython classpath, and opengl\library\opengl.jar

Download the PythonOnProcessing (tested on Netbeans 6.8)

Posted by stu at February 3rd, 2010

Cairo is a 2D vector graphics api used by Firefox, Gtk and other desktop projects.

I’m going to show here that it can also be used to generate web content, using Django.

I’m going to port two examples from the Michael Urmans Cairo Tutorial for PyGTK Programmers.

To understand the cairo and it’s drawing model I’d recommend his his Cairo Tutorial for Python Programmers.

Note: In this example I’ll be generating SVGs…  as I.E. (as of 2010) does not support them, you might want to generate PNG or PDF – if you need to do this with cairo, look for one of the many cairo tutorials on the web.

The example django project can be downloaded at the end of the article.

(more…)

Posted by stu at January 25th, 2010

I’ve uploaded my scripts to help workflow to google code, naming the project batch flow.

Heres a summary of some of the things you can do.

Clipboard integration

Go to a directory in the clipboard.

If you have the address bar enabled in explorer:

Copy the location, go to the prompt and enter ‘pcd’ to go to that folder

Go to the location of a setting in netbeans or eclipse:

Copy the location , go to the prompt and enter ‘pcd’ to go to the folder (or folder containing the file).

Get the current directory without dragging:

Enter ‘ccd’

Directory bookmarks

dirsave and dirload let you save named bookmarks.

Using hotkeys

If you use the alternate shell 4nt (or the free TCC/LE) you can use the supplied configuration and aliases to access the bookmark functionality from the keyboard:

In the 4nt or tcc/le prompt enter “option”, under the “TCStart/TCExit” path, change the location to the location where batch-flow is installed + “\conf”, for instance on my computer batch-flow is installed to c:\usr\batch-flow, so I set it to

c:\usr\batch-flow\conf

Now in new TCC/LE prompts F5-F10 are reserved for directories: Ctrl+Fkey to save, and Alt+Fkey to load.  Alt-F12 lists these shortcuts.

Note:  Alt-F12 only lists shortcuts on FKeys, to list these and other shortcuts enter dirload /l

batch-flow comes with other handy hotkeys, use ‘alias’ in TCC/LE to see what they are.

Path manipulation

addpath

It’s annoying after installing a program to have to add it to the path, so there is an ‘addpath’ command to do this.

regpath

This is a more general utility for viewing the registry path, you can list it, validate it, check for the location of files within it.

Also useful is ‘regpath /L’ which sets the local prompts path to the one in the registry.

Further help

Most of the commands have help builtin, which you can access by using the /? option.

Posted by stu at January 21st, 2010

I recently found a neat way of accessing files in a linux VMWare image.   This is really useful, as theres never really a good time to break your VMWare image, this is also handy if you don’t want to run the whole machine, but just access the files inside.

There is one caveat:

• It only works if the filesystem is ext2 (ext3 works, and ext4 probably works).

Install VMWare DiskMount Utility

Accept the EULA, download and install the VMWare DiskMount utility.

For convenience add the utilities folder to the path:

C:\Program Files\VMware\VMware DiskMount Utility

Do this through the Windows Gui, or even use my addpath utility.

At this point you can mount Windows VMWare images.

The usage is:  vmware-mount drive-letter vmdk-image.

Heres how I mount my Ubuntu image to the j: drive

[C:\vmware\Ubuntu]vmware-mount j: ubuntu.vmdk

[C:\vmware\Ubuntu]

No output here indicates success.

At this point everything seems fine, but a crucial piece of the puzzle is still missing; try and view the files and you still can’t:

The next step is to make Windows understand the ext2 filesystem, using a special driver.

Install ext2ifs

Grab ext2ifs from www.fs-driver.org and install.

If the following steps don’t work then you may need to reboot.

Thats it!

You should be now able to access files inside your VMWare image (assuming it’s ext2 and not reiserfs), remount the image and have a go:

In my case I did:

[C:\vmware\Ubuntu]vmware-mount j: ubuntu.vmdk

[C:\vmware\Ubuntu] dir j:

Heres the output – hooray, I can copy my work out of the image !

This is very useful, especially if you do dist-upgrade in ubuntu and can’t access the network.

Bonus tip:

Newer versions of VMWare player let you install VMWare tools from inside the GUI, this is another way to fix the kind of catestrophic problems you can cause yourself by accidentally upgrading the kernel in an image.

Posted by stu at September 6th, 2009

Stoyan of PHPIED.com has found the joy using javascript for scripting in the OS…

I’ve put up a few scripts I find useful in windows (download at the end of the article):

• addpath.js – Add a path to the registry path.
• dt.cmd – Change to desktop folder.
• e.cmd – Open explorer in current or specified folder.
• regpath.js  – Output the path stored in the registry.
• updateenvironment.js – Updates running apps with any changed settings in the registry.

And a couple that use python and pywin32:

• ccwd.py – Copy the current working directory to the clipboard
• cpath.py – Copy the current path to the clipboard

All of have acompanying batch files to run them, I generally have everything in a folder c:\usr\cmd, but they should work from anywhere in the path.

Download  cmd.zip

Posted by stu at August 21st, 2009

I recently came across Java2Python.  As I’m interested in Cairo I thought it would be interesting to try porting one of the example tutorials from ZetCode.

I’ll run through the steps involved in porting then try and reach some conclusions at the end .

1. Get setup

This is easiest in Linux, I’m running Ubuntu (in vmware), and installed

antlr 2.x
python2.5.x
sun java6
pygtk
java-gnome

You can install them like this:

# sudo apt-get install antlr python2.5 sun-java6-bin libjava-gnome-java

Then install Python2Java with easy_install

# sudo easy_install-2.5 java2python

To test if it’s working run j2py -i.  It should complain there is no file:

# j2py -i
Usage: j2py [options]

j2py: error: -i option requires an argument

If you get any other errors your missing some packages.

2. Get the Java Code from the Simple Example.

Save the ’simple.java’ example as ‘GSimple.java’

If java-gnome is running ok, compiling and running it you should see a window:

# javac GSimple.java
# java GSimple

Posted by stu at August 7th, 2009

Just been checking out what Mr Palmer has been up to… quite a lot it seems…   if I could get further than 2 chapters in 6 months with Getting Things Done maybe I could stop procrastination and start GTD too

And I’ve been checking out some online comics…

My 6 monthlyish check on SayUncle comics blog always makes me want to do some drawing.

Also Quite enjoying the Beartato in space storyline in nedroid.