3 ms·

Not sure what's happening with Scala, it's been on a steady downward trend of late (from 31 to 48). F#'s meteoric rise, on the other hand, when will it plateau

Not sure what's happening with Scala, it's been on a steady downward trend of late (from 31 to 48). F#'s meteoric rise, on the other hand, when will it plateau, top 10, top 5? Haskell dropped out of the top 50 which was also a surprise. Ruby seems to have stabilized, looks like it won't be evicted from the top 20 anytime soon. Other than that the 4 Cs continue to dominate.

gus_massa 13y ago
After a few years of Physics laboratory. the lack of error intervals make me cry :(. 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% (Ordered by absolute relative change.) (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.
- virtualwhys 13y ago
  Thanks for this. Scala's drop has been over a 3 month period. F#'s rise seems "real" given the consistently strong numbers of late.

gus_massa 13y ago

After a few years of Physics laboratory. the lack of error intervals make me cry :(. 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% (Ordered by absolute relative change.) (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.

virtualwhys 13y ago
Thanks for this. Scala's drop has been over a 3 month period. F#'s rise seems "real" given the consistently strong numbers of late.