enquos() and enexprs() should have attribute for auto_names

[hadley]

So that we can tell what names were user supplied and what names were autogenerated.

(This is will help when unpacking unnamed tibble columns)

[lionel-]

Maybe this should be a different function?

quos <- quos(foo = bar(), baz(quux), !!letters)

auto_label(quos)
#> # A tibble: 3 x 3
#>   expr       name      generated
#>   <quos>     <chr>     <lgl>
#> 1 bar()      foo       FALSE
#> 2 baz(quux)  baz(quux) TRUE
#> 3 <chr>      <chr>     TRUE

[hadley]

Yeah, I think that would be fine too

[lionel-]

This would still be helpful in packages like dplyr or tibble I believe, right @romainfrancois @krlmlr?

[krlmlr]

Yes, dplyr:::dplyr_quosures() currently calls enquos() twice for that reason, it would be significantly faster if called only once.

Perhaps unrelated: Are you planning to move endots() to C?

library(rlang)

basic <- function(...) enquos(...)
ignore_empty <- function(...) enquos(..., .ignore_empty = "all")
named_ignore_empty <- function(...) enquos(..., .named = TRUE, .ignore_empty = "all")

bench::mark(
  basic(a = 1),
  ignore_empty(a = 1),
  named_ignore_empty(a = 1)
)
#> # A tibble: 3 × 6
#>   expression                     min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>                <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 basic(a = 1)                20.2µs   21.4µs    46119.        0B     46.2
#> 2 ignore_empty(a = 1)         30.1µs   31.8µs    30842.    12.4KB     49.4
#> 3 named_ignore_empty(a = 1)   38.6µs   40.8µs    23991.    21.2KB     48.1

^{Created on 2022-05-01 by the reprex package (v2.0.1)}

[krlmlr]

We now have exprs_auto_name(), and ?dots_list documents that this is equivalent to enquos(.named = TRUE) . Do we still need to take action here?

[lionel-]

Are you planning to move endots() to C?

Yes.

We now have exprs_auto_name(), and ?dots_list documents that this is equivalent to enquos(.named = TRUE) . Do we still need to take action here?

Yes because we need a way of determining which names are user-supplied and which names are auto-labelled.

[krlmlr]

Can't we get that info from the names? Non-empty names are user-supplied, empty names should be auto-labeled? What am I missing?

[lionel-]

As I understand this issue is specifically about making the pattern easier without having to use enquos(.named = TRUE) and exprs_auto_name() in two steps. It is a low priority feature and maybe not needed in rlang.

[krlmlr]

Thanks. Did you mean enquos(.named = FALSE) and exprs_auto_name() ?

[lionel-]

yes sorry