TIOBE Index for April, 2014: F# Booming, Scala Busting, etc.

I check it out more for language trends rather than fixed position; i.e. is language X on an upward or downward trend.

Scala, for example, is my daily dev. so not thrilled to see it trending downward o_O

Assumed Haskell would be trending upward along with F#

Food for thought regardless...

I couldn’t find them. and I guess they are not available. So the only possibility is to guesstimate. This can make a few of my friends that really know statistics cry and never speak to me again. So don’t take my numbers too seriously.

Fist we will assume that all the variation in the top 20 is due to noise, not due to a tendency in the real data. This exaggerates the estimated noise. The table of relative change = change / raiting is:

  Programming Language    Relative Change
  C                          -1.3%
  Java                       -1.9%
  PL/SQL                      9.0%
  Delphi/Object Pascal      -12.1%
  Ruby                      -13.2%
  JavaScript                 13.7%
  Lisp                      -13.8%
  Assembly                   16.4%
  Objective-C                25.5%
  C#                        -27.6%
  MATLAB                     31.6%
  (Visual) Basic            -36.6%
  Visual Basic .NET          37.2%
  Transact-SQL               38.5%
  D                          55.1%
  C++                       -58.3%
  F#                         85.9%
  PHP                       -95.6%
  Python                   -122.9%
  Perl                     -127.6%

(Too many %. Example: Language: PL/SQL. Raiting: 0.782%. Change 0.07%. Relative Change = Change / Raiting 9.0%)

There are 2 languages (Java and C) with a RC <2%, they are the bigger languages, so this is possible not an artifact.

There is a group of 6 languages with 9%<RC<16%.

There is a group of 6 languages with 25%<RC<38%.

There are 6 languges with 50%<RC<130%

I think that we can optimistically estimate the error using the first group, and approximate it as a 10% (but a 20% or a 30% is not out of discussion).

(Technical note: Some distributions are very “noisy”, if this were some combination of Poisson distributions I’d expect a noise of 100%. But an easy model for the Poisson distribution is that each developer choices each month a language with a fixed probability, independently of the previous choice, this assumes that developers have no memory. This is very different of the TIOBE measurement, because they essentially measure the accumulated number of webpages of each language, so this is very different from the no memory model.)

-----

Back to Scala: Did it drop from #31 to #48 in one month or it’s a multi month “trend”?

This month, the ranking of ABAP (#31) is 0.367% and Scala (#48) is 0.0219. I don’t have the Scala raiting of last month, so I’ll use 0.367%. The RC “is” 67%. It looks big for a month, but if the change was in a few months I’d think it’s mostly noise.