On Sunday, January 26, 2014 7:22:04 AM UTC-5, Hans W Borchers wrote: > > How can I see a list of all functions in Julia Base, say with signature > and a one line description? >
Well, there is http://docs.julialang.org/en/latest/stdlib/, although the descriptions are not one-line. e.g. quadgk is in the section "Numerical Integration". How can I see the signature of these three methods? In quadgk.jl I only > found two function definitions for quadgk(). > methods(quadgk). (There are three definitions in quadgk.jl, not two ... look for "function quadgk" in the file. But they all basically look the same, and are just for different numeric types.) > > Calling the function (in Julia 0.2.0, on Mac OSX) I get > > julia> quadgk(sin, 0, pi) > Warning: Possible conflict in library symbol dstev_ > (2.0000000000000004,1.7896795156957523e-12) > > What does this warning mean? > This warning is due to two different versions of LAPACK being linked; it happened on some versions of Julia in the past on MacOS because of an extra LAPACK library being pulled in via Apple's "Accelerate" framework (see e.g. https://github.com/JuliaLang/julia/issues/1642). It is harmless, I think, but, the problem seems to have been fixed. > I would like to understand though in Julia 0.3.0 (on Ubuntu Linux) this > warning seems to have gone. By the way, Julia here returns a more exact > result 2.0, which I guess was not reached by improving the algorithm. > The difference between 2.0 and 2.0000000000000004 is probably just due to rounding errors and I wouldn't attach much significance to it. > Seeing the accuracy of the result, the error term is much too large. But > having written my own version of Gauss-Kronrod (in R) I know it's not > obvious how to improve on that. > Yes, the error bounds provided by embedded quadrature rules are generally very conservative for analytic functions. > > In helpdb.jl, quadgk has a distinctly longer description than most other > entries. How can I better exploit this "help database" (online)? > http://docs.julialang.org/en/latest/stdlib/base/#numerical-integration
