I'm feeling that the -reghdfejl- package and the underlying -julia- package are more solid now, and have bumped their versions to 1.0.0. Both are available on SSC. (reghdfejl mimics reghdfe while calling the Julia package FixedEffectModels.jl for a ~10X speed up on hard problems.)
A new paper about all this is at https://arxiv.org/abs/2404.09309. It is called "Julia as a universal platform for statistical software development." It presents of vision of developing back ends in Julia--once--and then calling them from Stata, R, Python, etc. Julia's a great environment for developing numerical software.
The jl command in the julia package now lets you start an interactive Julia session in Stata. And Mata-like Julia functions such as st_global() and st_data() and st_view() give more convenient access to Stata objects from within Julia.
Example session:
A new paper about all this is at https://arxiv.org/abs/2404.09309. It is called "Julia as a universal platform for statistical software development." It presents of vision of developing back ends in Julia--once--and then calling them from Stata, R, Python, etc. Julia's a great environment for developing numerical software.
The jl command in the julia package now lets you start an interactive Julia session in Stata. And Mata-like Julia functions such as st_global() and st_data() and st_view() give more convenient access to Stata objects from within Julia.
Example session:
Code:
sysuse auto
jl
sample = rand(Bool, st_nobs()) # random Boolean vector defining subsample
v = st_view("price mpg", sample) # view onto subsample of two vars
v ./= 2 # halve these Stata data points
st_numscalar("s", sum(v)) # sum the halved data into Stata scalar s
st_local("m", string(st_numscalar("s"))) # put sum as string in Stata local m
exit()
