Sunday 3 October 2010

TIOBE index decomposed

This post presents a different way to look at the TIOBE index. Of the top 20 languages, I exclude non-generic languages, like SQL or MATLAB. The next is to group the languages by performance characteristics:

Compiled languages that produce very fast apps:
* Ada
* C
* C++
* Delphi, Pascal
* Objective-C

Garbage-producing languages:
* C#/VB.net
* Google Go
* Java

Languages that produce slow apps:
* Javascript
* Lisp
* Perl
* PHP
* Python
* Ruby

When I look at products, which I consider successful, fast and slick, I see:
* Apple uses Objective-C
* Linux kernel and tools (incl. Android, Google servers etc.) use C/C++
* Webkit, Google Chrome use C/C++
* Microsoft uses C/C++
* Several successful vendors with great software in my industry use Delphi

I see problems with:
* Microsoft .net and Java apps tend to be bloated and slow
* Java apps on Android are not as slick as Objective-C apps on iPhone
* C#/VB.net programming requires that you are willing to bet your investment on Microsoft, who has lost major market share in many form factors
* Large apps or frameworks built on PHP or similar, simply fail to deliver

Looking into the future, where we will need one language to write one app that runs on multiple CPUs with each their own RAM, or NUMA, the compiled languages will do just fine, but the garbage languages need to be replaced with something that handles memory allocations differently.