Usage for dqrng

[rstub]

It's nice to see dqrng being used. However, I think a slightly different pattern would be appropriate since it is only in Suggests. Right now you have

  requireNamespace('dqrng')
[...]
      #Mixi <- matrix(rnorm(n=x, mean=Means[d], sd=SDs[d]), nrow=AnzPoints[d], ncol=1)   
      #MT 2019/03: das ist die schnelle version fuer den bottleneck
      Mixi <- matrix(dqrng::dqrnorm(n=x, mean=Means[d], sd=SDs[d]), nrow=AnzPoints[d], ncol=1)   

which will fail if dqrng is not installed. Instead you could either go from Suggests to Imports or use the fall-back pattern from Hadley's book

  if (requireNamespace("dqrng", quietly = TRUE)) {
    Mixi <- matrix(dqrng::dqrnorm(n=x, mean=Means[d], sd=SDs[d]), nrow=AnzPoints[d], ncol=1)
  } else {
    Mixi <- matrix(rnorm(n=x, mean=Means[d], sd=SDs[d]), nrow=AnzPoints[d], ncol=1)   
  }

BTW, you might be interested that dqrng also has support for fast sampling. Support for more distribution functions (like lnorm) is planned.

[Mthrun]

Thanks for the suggestion. Fix will be included in next CRAN version.
Btw I am looking for a random generation for the log normal distribution. Do you know where do find such a function?

[rstub]

There is a log-normal distribution function in boost, which I plan to use for extending dqrng. In principle one could do this one-self, see e.g. the examples in the vignette. However, this would be a separate RNG that would require separate seeding.

[rstub]

BTW, I am still planing to add dqlnorm eventually to dqrng. In the meantime, you could follow the approach that I described in https://stubner.me/post/2024-05-15-dqrng-v0-4-0/. Now it is possible to write your own dqr* functions while making use of the global RNG controlled via dqset.seed.