Re: iSeries, web development and perl

Kevin Schroeder skrev:

PHP won't complain until runtime, but the IDE may. I know that Java, C++, C#, heck, even VB developers won't believe this, but even though the PHP compiler won't complain prior to runtime this is very seldom an issue. I have only my own anecdotal experience on it but that experience runs through many different companies (being that I'm in services) and compiler-level issues seldom are problems. If they are, it's usually because someone edited a file manually, in production.

I see. Empirical evidence is much better than "we-think"-bashing :)

Is it common to create modules - like jars - in PHP? Or do you just have all the source available to the application all the time?

Just to be certain - does the PHP _compiler_ know if you write "$userObj = new User()" that $userObj is a User and that only operations provided by the User class is allowed, or is this only checked by the runtime?

PHP is not compiled until runtime and so the type is only checked at that point. But "runtime" is usually the next refresh of the browser during development or the next run of your unit tests and problems are usually detected long before production. Ironically, I probably get just as many Java ClassNotFoundException exceptions as I do invalid classes in PHP when I write code.

Really? I've found that Runnable/Executable jars helps enourmeously with classpath problems. Eclipse 3.5 has an Export -> Runnable Jar option which can bundle a whole project nicely.

Unit tests? Ok, php does unittests. Good to know :)

One thing I should correct myself on is that while you have static type checking for arrays and objects for function and method parameters, and instanceof checks available at any point, you do not have it for return values. However, even the simplest IDEs have the ability to understand PHPDoc and use it similarly to annotations. Variations from what you document in your code may cause a fatal error in PHP, but if you follow even the most basic of documentation standards, the IDE will help you out. This I can understand as a legitimate criticism from the static-typing crowd and would concede that strongly typed return values have value. But, like the previous examples, the actual impact is much lower than what the static typing crowd would expect.

You can state the type for objects and arrays. You don't have to, but you can and should. I probably write my code in a more structured manner than > a lot of PHP developers, but a good portion of them are really starting to take architecture seriously and building well defined class, interfaces and > such... and actually FOLLOWING a defined architecture. PHP has a bit of a reputation as a fly-by-night language. Some of that is deserved but new > applications are being built much better than even just 2 years ago. So, yes. I'm seeing a lot of new development following proper OO patterns, more or > less.

Ok. So you are basically saying that you PRAY that the previous programmer did the program nicely :)

To some degree, yes. But I can also say the same thing about Java, having also worked with it for several years (one of the talks I give is on integrating Java and PHP http://www.slideshare.net/SamHennessy/caffeinated-php-zendcon-2009 ). From my personal experience, variable type is usually down on the issues list in either environment.

Many of the experiences destilled into the Java language are not very visible, and not necessarly things valued by the particular programmers who would integrate Java and PHP (I think :) ).

But to target it back abit. I was not talking about writing programs, but maintaining other peoples program to a degree where major refactoring is needed. Would this still work as efficiently as you describe?