Simon P. Couch

Ah, looks like we _do_ gitignore revdep but track a couple files in that directory still: https://github.com/tidymodels/tune/blob/0ea505c1f35268bbba59f0b0695ae65340b781d1/.gitignore#L14 https://github.com/tidymodels/tune/blob/0ea505c1f35268bbba59f0b0695ae65340b781d1/revdep/README.md#L1 I guess the flow that I actually usually have is not to...

Idea from Hannah: instead of using the name of a function that would generate the class (e.g. `"initial_split"` for `"mc_splits"`), use a set of keywords like `"random"`. That could be:...

Okay, there's a bit more to this than I had wrapped my head around. We take in `method` so that we know how it was that data came to be....

We still haven't solved the problem in the comment above, where knowing `method` doesn't actually give us enough information to actually split data in a principled way (unless the split...

Further self-discussion: https://gist.github.com/simonpcouch/646eaaed213f634efd72a7f5615a9519 Thinking I will move forward with approach c) unless we decide otherwise in further discussion.

After following up with others on this, it's indeed surprising to others that workflows doesn't just rely on recipes and parsnip for print methods (and possibly modifying them a bit...

@ThomasSoeiro A separate issue is good! :)

Synonyms are now supported with `dbListTables()` and `dbExistsTable()`. With `main` ``` r library(DBI) library(odbc) con [1] 0 dbExecute(con, "create table odbc.test (x int)") #> [1] 0 # confirm that we...

Here are those benchmarks with the PR as-is (**EDIT: updated timings below**): ..where the values to the left of `x = 1` are `x = 0` (with a small shift...

Updated benchmarks: Definitely better than before but not negligible. I haven't seen enough real-world DBs to know how close to "worst-case" this is.