Neal Fultz

I think each *step* should have it's own attribute of which packages were used when it was declared. On Tue, Jul 17, 2018 at 12:30 AM macartan wrote: > Makes...

I ran this in the debugger, here are some notes: warning is triggered here: https://github.com/DeclareDesign/DeclareDesign/blob/679bd32fc9bb0e5e6e1ff318d227c95c09ef39af/R/simulate_design.R#L204 Note that the merge is doing an *outer join* - the check on the cardinality...

For sequential designs, I think it's easier / cleaner to generate the whole thing up front, and use `subset` options on estimator/estimands instead of declare_sampling - once rows are dropped...

`attach`/`detach` are verboten on CRAN - I think to make this work, you can use overscoping when you evaluate the step. The wallet should also be declared on the `results`...

> Thanks @nfultz. Any hints on what we would do to use overscoping when eval'ing the next step? Tried a number of combinations of making data masks and creating environments...

Instead of modifying the step, you can control the evaluation. something along the lines of current_df step body(step) { eval_tidy(if (is.null(data) & is_implicit_data_arg(dots)) quoNoData else quoData, data = list(data =...

Note that in the above example, you would still expect `f()` itself to fail - the deep copy should only happen internally to the design.

Try splicing the value of `means` with rlang - `outcome_means = !!means` On Wed, Jan 9, 2019 at 3:04 PM Graeme Blair wrote: > Either identical or similar issue: >...

I can take a look at this next week.

Like you mentioned, this is somewhat complicated by the fact that there's multiple layers of code doing slow things. I do think this could be very helpful, although we could...