One thing I learned today is the huge importance of Perspectives and Views in Eclipse. When Kelvin first said “right, stop using PhpEd, and use this Eclipse thing”, I was overwhelmed by the feature-richness, and my knee-jerk reaction was to just set it up as close to the way I used PhpEd. Thats the first mistake I made, because although I got it close, I missed out on all the glorious features I could’ve been using.

Our Eclipse configuration at work uses Eclipse, PDT, Mylyn, SVN and CollabNet - which integrates wonderfully into our Sourceforge 5.0 server. Mylyn and CollabNet allow task management and allows me to work on Sourceforge artefacts within the Eclipse environment. A rather cool feature is contexts, which allow me to basically attach a list of files to the SF artefact, and when I activate the task in my Task List (note, not “Tasks”, “Task Lists”, a part of Mylyn), it opens all the files I was working on.

Because of our heavy reliance on SourceForge (everything we do has an associated artefact), it makes sense to make use of the CollabNet perspective. If I want to go into “task management mode”, i.e. line up a load of work that has been assigned to me, I switch to the CollabNet perspective. I open an artefact from my assigned artefacts list in the CollabNet view, and it automatically adds it to my Mylyn Task list. I can then say “right, I’m working on this one”, and click a little blob which turns blue, and then I’m working on it. If I’ve been previously working on it, and I attached the context, it automatically opens the files I was working on.

This was meant to be a short homage to Eclipse, but has turned into a little rant, but now that Kelvin set up the SVN repository on Friday properly, we’re actually FINALLY using source control, after the months of me (and Chris) nagging him to get it set up. It’s a little step, but we’re working on the same checkout at the moment, but eventually all that will change. I’ll start bugging him for that soon…

Anyway, if you’re a PHP developer, on a Mac, Windows or Linux, and you’re not using Eclipse PDT yet, I strongly suggest you do so. It’ll make your productivity increase hugely - it did mine!

We dragged out an old favourite, but varying a little with the Here and Now UK Edition of Monopoly. Playing reminded us both of why we never play it, because it’s such a long game, and boy it drags. Hannah eventually gave up trying and I managed to settle hotels on half the board and own the rest of the board… which I suppose is the aim of the game really.

Fun nevertheless, but it’s off to bed for me, ready for another busy week in the office… and a whole new week’s worth of Twitter updates.

And thinking about it, how much I’ve changed as well! I’ve been dabbling an awful lot more in trying to make things XHTML compliant and using CSS in the proper fashion recently, and after a bit of modification to the Picasa widget, my page is fully XHTML 1.0 Transitional valid. I’ve not bothered making it strict yet, as generally widgets made by other people take a lot of work to make them strict. In fact, up until Matched.co.uk went under, my pages have been non-compliant due to non-compliant Matched.co.uk ad code.

And last, but most certainly not least, the biggest change is of course the pregnancy, which is coming along swimmingly by the way!

Basically, whenever you commit a file to the repository, if you have a checked out workspace on the same server, then it will automatically update to the latest revision of the checked out branch or trunk. It’s handy so you don’t have to “svn up” every time you make a change, especially if you’re making lots of tiny changes.

This guide was original from Marc Grabanski’s blog, so 99% of credit goes to him.

cd /svnroot/hooks/
cp post-commit.tmpl post-commit
chmod +x post-commit

You can make a little program to update all your SVN repositories, so nano svnupdate.c:

#include <stddef.h>
#include <stdlib.h>
#include <unistd.h>
 
int main(void)
{
  execl(
    "/usr/bin/svn",
    "svn",
    "update",
    "/path/to/mywebsite1.co.uk/www/",
    "/path/to/mywebsite2.co.uk/www/",
    "/path/to/mywebsite3.co.uk/www/",
    (const char *) NULL);
  return(EXIT_FAILURE);
}

You can add as many paths as you like there, just copy and paste the line and edit the path to your needs. You need to then compile it, test it.

gcc -o svnupdate svnupdate.c
env - ./svnupdate

Nano your post-commit file and add in the FULL path to your svnupdate program:

/path/to/svnroot/hooks/svnupdate

Ingredients

• Carrots
• 3/4 pint Vegetable or chicken stock
• A white potato or two (depending on how thick you want it - I used 3!)
• Fresh coriander
• A bit of butter
• Optional - double cream

Instructions

Skin potatoes and carrots. Chop coriander up really fine. Remember - the more coriander you have, the more spicy it gets! Put a bit of butter in a large saucepan. Heat them on medium/low for about 5 minutes. Pour in your stock, and simmer for 15 minutes. Add coriander, and if you want, some double cream.

Let it all cool right down to a temperature that won’t melt your blender - you’re not blaming me if you put hot stuff in your blender and it melts! Then put it all in the blender and blend until really smooth. Once blended, pour back into your saucepan and heat… and as always, do not boil it as this may impair the flavour, and it’ll just be downright hot.

Pour into your bowl, and garnish with a coriander leaf. Serve with bread and butter. Yum!

$out = number_format($val,0);
$out = (int) @floor($val);

Interestingly, we ran a real-world test case of 10 million iterations of doing each of these, and it turns out that method #2 is approximately 0.5 seconds quicker. So there you have it, if you want to be 0.5 seconds quicker when removing decimal places from a number 10 million times.. then use the second method.

/usr/bin/ld: cannot find -lgssapi_krb5

After a little research and talking on the #roadsend IRC channel with GRiD (see chat log), I discovered that you just can’t compile a statically linked binary on Ubuntu. The statically linked library was dropped between Feisty and Gusty, and this may also be the case across many other Debian-based distributions.

Read this forum post.

My advice is to try a different distro, GRiD suggested Gentoo, which I may just try tonight.

Update: I recommend Fedora, and dynamically linking your binary, because Gentoo takes way to long to install and configure.

UBUNTU USERS - READ THIS FIRST!!

You might think compiling PHP, a scripted language, is a bit daft, but there are a few really good reasons you might want to compile your source code. In fact, compiled web code has been around for a while now, what with ASP.NET and all that. The main two reasons are (1) protecting your intellectual property if your project is closed-source and (2) potential speed increases. I say potential because that’s just a theory really - I’ve not investigated speed increases (or even decreases) yet, all I’ve done is compiled a test app and got it working.

I used Ubuntu as my compilation OS, because… well it’s awesome really*. I used this and this as the main guide, so read them and I’ll skim over the process I did here.

Firstly, you need to install and compile Bigloo (this will take 10 mins or so):

cd /usr/local/src
sudo wget ftp://ftp-sop.inria.fr/mimosa/fp/Bigloo/bigloo3.0c-4.tar.gz
sudo tar -xvzf bigloo3.0c-4.tar.gz
sudo rm bigloo3.0c-4.tar.gz
sudo chown -R root bigloo3.0c && sudo chmod -R u=rwx,go=rx bigloo3.0c
cd bigloo3.0c
sudo ./configure --prefix=/usr --enable-sqlite
sudo make && sudo make install

Then, install Roadsend (this will take 5 mins or so), ensuring FastCGI support is compiled in:

cd /usr/local/src
sudo wget http://code.roadsend.com/snaps/roadsend-php-2.9.7.tar.bz2
sudo tar -xvjf roadsend-php-2.9.7.tar.bz2
sudo rm roadsend-php-2.9.7.tar.bz2
sudo chown -R root roadsend-php-* && sudo chmod -R u=rwx,go=rx roadsend-php-*
cd roadsend-php-*
sudo ./configure --prefix=/usr/local
sudo make && sudo make install

Add this line into /etc/profile (or ~/.profile):

export LD_LIBRARY_PATH=/usr/local/lib

Go to your “source code” directory (e.g. /var/www/src), and create a script with this in:

#!/bin/sh
 
# Change this value to the name of your binary
APPNAME='testapp'
 
# Change this to your cgi-bin directory
CGIDIR='/usr/lib/cgi-bin'
 
# Compile using PCC
pcc -v --static -O --fastcgi $APPNAME `find ./ -name '*.php'`
 
# Copy fastcgi to
sudo rm $CGIDIR/$APPNAME.fcgi
sudo cp -v $APPNAME.fcgi $CGIDIR

You’ll need to modify the script - replace “testapp” with whatever you want the final binary to be called, and replace “/usr/lib/cgi-bin” with the location of your cgi-bin (you can find this by doing “grep -r ScriptAlias /etc/apache2″ on an Ubuntu system). If you don’t need to sudo to modify files in your cgi-bin directory (i.e. you have permissions in that directory), then you can remove the sudo’s from the beginning of the rm/cp commands. I need this as my cgi-bin is owned by root.

Now all you need to do, in the live directory, create a .htaccess with this in:

RewriteEngine On
RewriteRule ^(.+\.php)$ /cgi-bin/testapp.fcgi/$1
RewriteRule ^$ /cgi-bin/testapp.fcgi/index.php

This forwards requests like http://host/blah.php to http://host/cgi-bin/testapp.fcgi/blah.php, and the FastCGI module magically deals with it just as you would any other normal website… but it’s compiled!

The great thing about this is that all you need to provide in your release packages is:

• The FastCGI module (e.g. testapp.fcgi)
• The data files (images, javascript etc. that go in your web root)
• If you’re nice, some instructions on how to set it up.

In theory, as it’s a FastCGI it should run on a multitude of web servers, and should be cross-platform compatible… but I’ve not tried it, so I can’t vouch for it yet!

* but not awesome at compiling FastCGI statically linked binaries.

The host is called Slicehost, and the provide VMs of 256mb RAM up to 15.5GB RAM (!). Prices for the smallest “slice” is $20 (£13.37) per month, which is pretty reasonable for the muck-around equivalent of your own dedicated server. The beauty of it is that the resources you pay for are reserved, so you won’t find some other “slice” on the same server as you using up your RAM or CPU time when you need it the most. Of course, these VMs aren’t just limited to development servers, upwards from the 1GB slice is the equivalent of your own dedicated server, so is also more than suited to production servers, especially as initially Slicehost was meant for business and production needs. Bandwidth is pretty reasonable as well - 100GB per month for the base package. Personally I wish something like this had come along before I bought 1and1 hosting for 2 years, heh! Oh well.

Check out Slicehost here!

• Grab the nearest book.
• Open it to page 56.
• Find the fifth sentence.
• Post the text of the sentence in your blog along with these instructions.
• Don’t dig for your favorite book, the cool book, or the intellectual one: pick the CLOSEST.

Similarly, a class that wraps an HTTP request (including headers, response codes, and so on) has only one instance per request.

The book that was nearest to me was Advanced PHP Programming by George Schlossnagle (ISBN 0-672-32561-6).

Update: Resetting the Tweet Checking seems to have fixed this so far…

Update 2: Nope, still not working…

Update 3: Alex King has posted a patch that was written by Greg that will hopefully fix the issue. I’m giving it a try, and I’ve also updated my Twitter Tools Mod patch.

Update 4: Looks like it works! The fix was just returning true from the digest post function, instead of some unset variable ($success). Woot!

As of the latest stable release of PHP, this is not the case. And it bugs me when people say “stdClass is the base class of everything in PHP”. stdClass is actually just a basic object with no methods or properties that you can use however you wish. A common use is to create a quick on-the-fly object, e.g.:

$obj = new stdClass();
$obj->Name = 'foobar';
$obj->Id = 24;
 
$json = json_encode($obj);

As you can see, $obj in this context is throwaway and is just used and forgotten about. I’ll re-iterate what David Otton says about it, and is quick and simple proof:

class DoesNotExtend {};
 
class DoesExtend extends stdClass {};
 
$doesNotExtend = new DoesNotExtend();
$doesExtend = new DoesExtend();
 
var_dump($doesNotExtend instanceof stdClass);
var_dump($doesExtend instanceof stdClass);

Outputs:

bool(false)
bool(true)

So please, when someone asks “what is stdClass” etc., don’t say “Oh, it’s the base class for all objects”, because it bloody well isn’t.

* Jaxxed seems to be slightly more well-informed, but that post was back in 2005. Since then, PHP 5 has come out, and classes are implemented much better. He’s still wrong though, it’s not the base class.

After having used Netbasic’s custom MVC for several months now, which has a base class, I believe there can be merits, especially in a framework situation, to extending a base class such as Object throughout all classes. You can add things like debug code into Object, and it propogates through into all child objects. It’s sometimes a good idea to take this ethic further, such as in .NET, classes are arranged from parent to child according to how specific their function is… i.e.

• Object - base object
• Control - a generic control
• WebControl - a generic web-based control
• BaseDataBoundControl
• DataBoundControl
• ListView - the list view object

I suppose how your classes are structured should be dictated primarily on what your application requires, and secondly on how you like them structured… it’s preference at the end of the day!