Programming Language Levels

http://www.theadvisors.com/langcomparison.htm

$75 for current version at http://www.spr.com/products/programming.htm

cited by Scripting: Higher Level Programming for the 21st Century (1997) http://citeseer.ist.psu.edu/44642.html

Capers Jones "Programming Languages Table" -- ranking programming languages by "level" (larger is better), typical source lines per function point, and rough numbers for number of function points implemented by man month for each range of language levels.

I would criticize his paper for failing to mention that specialized languages are more effective at a particular task. SQL, for example, is "good" at fetching and updating relational data, but computing primes is not its bag. -- JeffGrigg

[Joe Celko has made a career out of getting SQL to do things it really shouldn't do. Don't be surprised if he hasn't done the SieveOfEratosthenes...]

That wouldn't require Celko's skills, that's easy; almost as easy as in APL. Create a big table of 10k integers 1..10k (which you already have lying around if you follow Celko's advice), and a small table of sqrt(10k) == 100 integers (1..100). Delete entries from big table where value = small table self joined value*value. Remaining big table entries are the primes <= 10k. (Oops, bug, should exclude 1 from the above, it's a special case. Second bug, it excludes primes < 100. Easy to fix, though; start with small table of primes listed by hand, use above to create primes through N^2, repeat to get as big a list as you like) -- dm

Jones has argued that although his measures are simplistic, they are surely better than no measures, and probably better than elaborate measures that can't be understood. I believed this once, but have now realized the flaw. It is the same flaw underlying IntelligenceQuotient (IQ). -- WardCunningham

Right. Write some RPG programs to print reports; it's great. Now implement the Star Trek game in RPG. Ick. (It's something one would only try if one had lots of time to spare, nothing better to do, and no other appropriate tools.) Now, as an exercise to the student, try to implement the Star Trek game in SQL. (HA!)

The point being that specialized languages can be VERY GOOD within their specialized niche, but also be less effective on more general problems -- or they may fail entirely. Thus a single Language Level, independent of the problem you're trying to solve probably doesn't make a whole lot of sense.

If Language Levels, as documented, really made sense, then we should implement all applications in MATHCAD (or maybe spreadsheets) -- that's what his numbers imply. -- JeffGrigg

 From: Doug Bagley (of GreatComputerLanguageShootout fame) [mailto:doug@bagley.org]
 Sent: Wednesday, April 25, 2001 11:33 PM
 To: Grigg, Jeff
 Subject: Re: another link for Language Shootout

[...] It's an interesting point of view. (But why didn't they sort by level? :-)

It would also be nice if they noted when the numbers were actually measured and when they were estimated, as they mention they did for some languages.

[Capers Jones "Programming Languages Table" is sorted by language name. First thing I did when I got it was to put it in a spreadsheet and sort it by level. I also did some interpolation of function points per man month based on place within the language level ranges. -- JeffGrigg]

It's pretty hard to take that table seriously. The table indicates that CeePlusPlus is level 6, with 53 statements per function point, and CommonLisp is level 5, with 64 statements per function point. Having programmed extensively in both, my estimates would be approx 10 statements/FP for lisp, 100 statements/FP for C++ (arbitrary normalization, only the ratio matters. I don't know how to set the scale). If the table is so wrong on the few languages I know well, how can I trust any of the rest on languages I don't know? -- AlainPicard

Oh, and his numbers "validated" our assumption that it would take about 80% as long (per function point) to hack it out in Visual Basic instead of Java. -- JeffGrigg (...with plenty of crazy assumptions embedded in the numbers ;-)

[Mentioned in ComputerLanguageBenchmarksGame; a performance comparison of different programming languages.]

CategoryPaper