Less Magical Numbers - A coding standard proposal
- 1. L l o y d M o o r e , P r e s i d e n t L l o y d @ C y b e r D a t a - R o b o t i c s . c o m w w w . C y b e r D a t a - R o b o t i c s . c o m Toward Less Magical Magic Numbers
- 2. In the “dark times” we represented constants like this: double x = 2.718; We eventually started calling this a “magic number”, as it is just a number that has to be correct for the program to work.
- 3. In time it was determined this was bad and we added more information: // The natural log of 1, e const double e = 2.718281; This is better, but for many applications it still isn’t good enough. Information is still missing: Where did this number come from? Why this particular precision? Is there a traceability requirement to the specifications?
- 4. Even when not formally required, having more information makes the number less magical: // The natural log of 1, e // Source: https://en.wikipedia.org/wiki/Natural_logarithm // 12 decimal places required for algorithm foo() // per requirement ABC-33920. const double e = 2.718281828459;
- 5. But what if it is just a number that someone made up?: // Developer discretion const double x = 2.718; Then say so with some type of flag. This will let the reader know the number doesn’t have any particular significance!
- 6. Real World Implications: This presentation is a result of a real world issue from an aerospace project. A constant had two different levels of precision between the flight software and the IV&V software, resulting in a failed verification. It took minutes to find the difference and a month to sort out value which was needed!
- 7. Questions?