997Flask – [Errno 48] Address already in use


Stopping the flask server with ⌃-Z (CRTL-Z), instead of the correct ⌃-C (CRTL-C), results in the server still bound to port 5000:

OSError: [Errno 48] Address already in use

Problem: After Flask crash, the port can still be occupied.


Check who is using port 5000

sudo lsof -i:5000
kill $PID
kill -9 $PID


[2]  + killed     flask run

829Digging a tunnel to a remote SFTP Server, Part 2

Back in Part 1, I was using a tunnel and a SOCKS proxy to connect to a remote server. While this might work in certain conditions, other circumstances might call for other solutions.

Take this case.

Recently our Sysadmins changed to login procedure to our servers. Instead of just logging on like this...

localmachine:~ me$ ssh username@host

...we are now forced to jump through another hoop:

localmachine:~ me$ ssh username@host
username@host's password: •••••••
Last login: Tue Aug 14 10:39:17 2012

[username@host]$ ssh username@anotherHost

I have to guess, that for reason's of security, it's not possible to log-in directly to anotherHost from my local machine: Any connection has to go through the host machine.

While some like to edit their text files with the very powerful, but slight archaic vi or Emacs, some others prefer more a more civilised approach and use tools like TextMate.

But if the text files are on a remote server, which needs to be logged-in from another server, how can I use TextMate to edit files there?

Enter the Tunnel

Cyberduck, the ever-popular SFTP application, can lend us a hand in editing the text files. A simple cmd-J, and the text file on the server opens in TextMate, cmd-S saves the modified file back to the server.

But how exactly do we open the tunnel? How do we start the port-forwarding?

ssh usename@host -L 1024:anotherHost:22

Let's dissect this command:

ssh usename@host open a connection to the host with our username. Of course you'll be prompted for your password.

-L 1024:anotherHost:22 is where the magic happens. It specifies the local port on your machine (in this case: 1024) to which the remote port (22) of anotherHost is mapped. And all this happens by way of the host machine.

Other switches which could be helpful for this command are: -f send ssh to the background -N stops the execution for remote command.

Check the ssh man pages for more details.

If you prefer a GUI for opening the tunnel, have a look at Fugu.

It does basically the same, only with a graphical user interface. If that removes or adds confusion is for you to decide.

If we look at the ssh command that Fugu is generated, we see that we were basically doing the same before:

/usr/bin/ssh -N -L1024:anotherHost:22 username@host -oPort=22 -v

737Processing Math convenience methods in pure Javascript

Like everybody else, I am huge fan of Processing, and like everybody else I find myself working more and more with Javascript. There is a implementation port of Processing (aptly called ProcessingJS) by John Resig of jQuery fame. It’s still lacking the 3D features, but not because they would be hard to implement in JS, but rather the OpenGL bindings in the browsers don’t exists wide-spreadly (yet). Let’s hope WebGL is going to change that soon. After working a lot with Processing, one becomes quite spoilt with it’s methods. The two best kept secrets of Processing are on the one hand the matrix state operations – pushMatrix(), and popMatrix() – and on the other hand the very handy map() method, basically mapping a number from one range to another. In Javascript’s canvas you have the pushMatrix() popMatrix() equivalents as save() and restore(). The map() methods does not exists as such, but it is easy implemented. I took a look at the Processing source and rewrote the Java methods in Javascript.
var p5 = {};

p5.sq = function(a) {
	return a*a;

p5.constrain = function(amt, low, high) {
  return (amt < low) ? low : ((amt > high) ? high : amt);

p5.degrees = function(radians) {
  return radians * (180/Math.PI);

p5.mag = function(a, b) {
  return Math.sqrt(a*a + b*b);

p5.dist = function(x1, y1, x2, y2) {
  return Math.sqrt(p5.sq(x2-x1) + p5.sq(y2-y1));

p5.lerp = function(start, stop, amt) {
  return start + (stop-start) * amt;

p5.norm = function(value, start, stop) {
  return (value - start) / (stop - start);

p5.map = function (value, istart, istop, ostart, ostop) {
	return ostart + (ostop - ostart) * ((value - istart) / (istop - istart));
The code including comments and a minified version is also up on GitHub, show you love and fork it. https://github.com/trembl/p5.Math.js/
Update – Extending Math
Rather than creating a dedicated object, and as all the function are kind of related to Math, I thought it would be a good idea to extend the Math object itself. (Well, it’s a good idea for my project – your mileage, especially if you are flying with a lot of libraries might vary). And how to you extend a Javascript build-in object? With prototype, right?
Math.prototpe.sq = function(a) {
	return a*a;
Wrong. Because Math does not have a constructor, it therefore hasn’t been constructed and therefore does not have the prototype method. It has been instantiated and you add methods (or functions?) like this:
Math.sq = function(a) {
	return a*a;
You get the idea. The new code also lives at GitHub, check it out. https://github.com/trembl/p5.Math.js/