A model that needs an externally-owned parameter injector – a registered
par-loader, e.g. a package's neural-network weight loader – flags that
injector's "<package>:<function>" name here. At solve setup only the loader
with that name runs, so an injector never overwrites an unrelated model's
par_ptr merely because it happens to be registered. The flag is stored on the
ui (a sticky item, so it survives model piping) like rxForcedPars(); set it to
NULL to clear.
Examples
# \donttest{
mod <- function() {
ini({
k <- 0.1
})
model({
d/dt(A) <- -k * A
})
}
u <- rxode2(mod)
#>
#>
#> ℹ parameter labels from comments are typically ignored in non-interactive mode
#> ℹ Need to run with the source intact to parse comments
rxParLoader(u) <- "myPkg:myLoader"
rxParLoader(u)
#> [1] "myPkg:myLoader"
rxParLoader(u) <- NULL
# }
