Unsustainable Software Development and its Causes

Woah, this hits close to home. I've seen this type of behavior first hand, more than once. I've often wondered why it's sometimes easier to rewrite a program than trying to improve it. When it happens on projects that I join there's often a good excuse: like "we need to rewrite in a modern computer language", "that software it's based on is no longer supported", "the software was written without concern for portability", etc.. But after reading this article I've realized that these are often just excuses, rewriting from scratch is very expensive.
What's needed, in my opinion is:
  • 90%+ automated unit test coverage
  • 90%+ automated funcional test coverage
  • Automated builds
  • Good programmers and good continuous monitoring of these programmers (code reviews, etc.)
Automated tests can fix a lot of woes. Plus you can make major changes to the underlying code (refactoring) without fear. If done right the automated tests can be used as documentation as well.

Comments

@TheSandyWalsh said…
Whenever I have a developer on my team yell "it has to be rewritten" I always tell them that it ain't gonna happen ever. They quickly learn that they need to come up with alternatives. You can't refactor without tests. Period. The Scaffolding approach works very well in the capacity. Start with functional large-grained tests and slowly work your way down to unit tests when working with legacy code. We successfully did this recently on a large project and was able to eventually replace the legacy product while continuing to add functionality to it and sell it to new customers. After a while, the old program simply was "refactored" out of existence. Sweet.
1:27 am
Post a Comment

Popular posts from this blog

Shortest Sudoku solver in Python

Mark Byer's site has some more Sudoku solvers here's a short one in Python that's only 178 bytes long: def r(a):i=a.find('0');~i or exit(a);[m in[(i-j)%9*(i/9^j/9)*(i/27^j/27|i%9/3^j%9/3)or a[j]for j in range(81)]or r(a[:i]+m+a[i+1:])for m in'%d'%5**18] from sys import*;r(argv[1]) On the site is also shown a longer Perl version at 185 bytes: use integer;sub R{for$i(grep!$A[$_],@x=0..80){%t=map{$_/27-$i/27|$_%9/3-$i%9/3&&amp;amp;amp;$_ /9-$i/9&&($_-$i)%9?0:$A[$_]=>1}@x;R($A[$i]=$_)for grep!$t{$_},1..9;return$A[$i]=0} die@A}@A=split//,<>;R I think the Python version is slightly easier to understand, although they are both quite cryptic. I'm not sure why I find these programs fascinating. I think it reminds me that any program can be written in such a way that it's cryptic, even in Python. In addition, the code is small enough that you should be able to figure out how it works, and may learn something the process. Related is
Read more

Seven Segment Display in Inkscape

Image
Today I got a little sidetracked while working on my key-train (keyboard training) project.  I thought I might spruce it up a little by using a seven segment display for the digits (say for the words per minutes), similar to what kTouch has.  I found an open source font, but it appears to be difficult to just use a ttf font in GTK. So I figured, it's only 10 images, I'll just make them in svg. I vaguely remembered a heathenx inkscape podcast  describing how to make an LCD display and decided to use that at my base. Unfortunately, they didn't provide any source files in the show-notes (as far as I could see) so I had to watch and redo the drawings myself! By naming each of the seven segments I could create the 10 digits simply by deleting a few elements in one template svg file .  I can create the template as beautiful as I like and quickly create the 10 derived images with these programs . Unfortunately, both ImageMagick and GraphicsMagick don't handle
Read more