Bug: Rstudio completion with "mrgmod" S4 class object

FelicienLL commented 2 years ago

Hello, First thank you for these awesome packages. Just reporting a minor issue: Rstudio completion does not work on "mrgmod" S4 class objects when used inside a %>% pipe. However, because it works with the base |> pipe, so I assume the problem is related to magrittr, but maybe I'm wrong.

library(mrgsolve)
mod <- modlib("irm1")
mod |> mrgsim(
# +tab: completion works
mod %>% mrgsim(
# +tab: completion returns error messages and stops

These are quite special S4 objects, since they have names() and a $ extraction method.

I tried to debug Rstudio completion method with debug(search_envs()[["tools:rstudio"]]$.rs.getRChainCompletions). Just noticed that:

with %>%, the value chainObjectName was equal to "mod", which triggers a [[ ]] extraction of each element in names, which is the origin of the error messages. Completion aborts.
with base |>, chainObjectName was NULL, thus the aforementioned extraction step is skipped (if(length(chainObjectName)) statement), and the completion is provided by an other part of the code. Named elements never got evaluated and the proposed completion was accurate (i.e. returned the arguments of the mrgsim function, but it is the same for any function).

Was this difference between |> and %>% expected at this point? Is it only related to the non-standard construction of these "mrgmod" objects, or could something be fixed on your side? Or do you think it is more related to Rstudio?

Many thanks in advance

Félicien

lionel- commented 2 years ago

This is more of an RStudio issue I think.

SimonDedman commented 2 years ago

HI, I'm not sure if this related (this is all at the edge of my mental pay grade!) but recently magrittr assignment pipe (%<>%) doesn't work - instead of assigning the chain of operations to the LHS object, it instead returns the result in the console.

However, I have also recently started using the base R pipe (|>) instead of the magrittr pipe (%>%).

When I replace base pipes with magrittr pipes, the assignment pipe works as expected. Is this something that can/should be fixed at your (magrittr team) end, or something for the base R folks? Thanks!

lionel- commented 2 years ago

The |> rewrites the input in nested form at parse time:

quote(a %<>% b() |> c())
#> c(a %<>% b())

So there is no appealing way to make it work in a %<>% pipeline.

FelicienLL commented 1 year ago

Will be fixed in the next version of rstudio. Thanks for your comment. https://github.com/rstudio/rstudio/issues/10538

tidyverse / magrittr

Bug: Rstudio completion with "mrgmod" S4 class object #253