Matthieu Gomez

@nalimilan Syntax 2 is just a better version of R multipart formula. In these multipart formulas, all arguments specifying variable names are put together and separated by `|`. I just...

The macro would just capture all the expressions and pass them to the function. The function would then specify which argument should be evaluated. Is there a problem with that?

Correct me If I'm wrong, but I think we can avoid this issue. `@fit(fe = z, maxiter = n)` would call `_fit(fe = :z, maxiter = :n)` which would return...

The keyword arguments are specific to my package, no? And how is this problem the same as the one regarding the function scope?

More generally it would be nice to have a way to regress on all variable satisfying certain conditions (a regex, or `dplyr`-like conditions like `starts_with`, `contains` etc).

I know that `@btime` returns a smaller time, but that's not the point. My point is that it takes way too much time to create a formula in the REPL...

That's what I was going to answer. But I am not sure the function g evaluates the macro the second time?

```julia function g() @time @formula(y ~ log(x) + log(y) + log(z)) @time @formula(y ~ log(x) + log(y) + log(z)) end ``` still gives 0.3 for both. So actually I don't...

Using a type that does not need to be specialized over and over again would be awesome! Or maybe use `@nospecialize` everywhere.

> Unfortunately Tables.Columns has a type parameter for the wrapped table type so I don't think it'll solve the problem in all cases, although it may help with sources that...