Sun, 08 Apr 2007

Horses For Courses.

A race horse

In my day job, I work with a hardware engineer named Joe. A couple of months ago, he had to do some C coding to talk to a serial port and came to me for pointers. He was basically on the right track, but was using too many global variables, not checking the return values of system calls etc. These weren't horrible problems, but I explained why practices like this can lead to problems later, showed him better solutions to the same problems and introduced him to the gcc warning flags and Valgrind. He was very grateful for my help and was a quick to pick up all the tips I'd given him.

More recently Joe came to me with another programming problem; he had to parse some numerical data out of a plain text log file. He already had about 60 lines of C code that opened a file based on a hard coded filename and he was starting to fiddle around with the fgetc function but was a little stuck on how to go further.

I had a look at his code and since Joe's a nice Irish chap, his predicament brought to mind a joke I once heard:

It is said that there was once an English motorist in Ireland who stopped his car to ask the way to Kilkenny. "Sure and to goodness," replied the Irishman., "If I wanted to go to Kilkenny, I wouldn't be starting from here."

The problem is that C is not a very good language for parsing text data. I told Joe that writing his log file parser in C could certainly be done, but that it would be painful, time consuming and error prone in comparison to other programming languages. So I told him that for this particular task Python would be a much better fit and asked him if he'd like me to teach him the basics of Python. Joe's no dummy; he agreed without hesitation.

First up I showed him the Python Tutorial, the Python Module index and how to used Google Groups advanced search to find Python specific answers from the comp.lang.python Usenet group. I then showed him the basic hello world program in Python: 


  #!/usr/bin/python

  print "Hello world!"

Over the next hour we built up a good portion of his program. We used used the sys module to get the file name of the log file to parse, the built in Python file handling functions and the regular expression module. We even used a list comprehension to remove outliers from his data set.

In the end we had about 30 lines of Python code that was very much closer to Joe's end goal than his original 60 lines of C code. Joe was really, really impressed with how easy Python was in comparison to C. It was at this point that I warned Joe about the Blub Paradox. He was well aware that when he only knew C, C was his first choice for this programming task. However, now that he knows Python as well, he'll be able to pick between C and Python depending on the task. I also told him that many Python programmers see Python as the ultimate programming language and are really Blub programmers even if they don't know it.

In my own programming I'm currently using:

So, with the above languages at my disposal, I can match a programming language to the task at hand. For numerical and mathematical programming I use Ocaml, for low level programming I use C and from now on, for multi-threaded and concurrent programming I will chose Erlang.

More importantly, correctly matching the language to the problem should make the task of developing a solution to the problem far easier than using an inappropriate language.

Posted at: 20:16 | Category: CodeHacking | Permalink