Make the current group size a variable updated by reference

[DavisVaughan]

This PR:

• Provides access to the current group size through $get_current_group_size()
• Uses this helper in n(), which makes it a little faster

This also gives us an easy private way to access the current group size for #6685 (comment)

It also gets rid of the need to do parent.env(chops) internally, which always felt a little clunky to me.

library(dplyr)

set.seed(123)

# ~100k groups, so that many evaluations of `n()`
x <- sample(1e5, 1e6, replace = TRUE)
df <- tibble(x = x)
df <- group_by(df, x)
df
#> # A tibble: 1,000,000 × 1
#> # Groups:   x [99,995]
#>        x
#>    <int>
#>  1 51663
#>  2 57870
#>  3  2986
#>  4 29925
#>  5 95246
#>  6 68293
#>  7 62555
#>  8 45404
#>  9 65161
#> 10 46435
#> # … with 999,990 more rows

bench::mark(summarise(df, n = n()), iterations = 100)

# Main
#> Warning: Some expressions had a GC in every iteration; so filtering is
#> disabled.
#> # A tibble: 1 × 6
#>   expression                  min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>             <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 summarise(df, n = n())    919ms    1.14s     0.873    7.15MB     13.8

# This PR
#> Warning: Some expressions had a GC in every iteration; so filtering is
#> disabled.
#> # A tibble: 1 × 6
#>   expression                  min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>             <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 summarise(df, n = n())    596ms    742ms      1.35    7.16MB     15.1

^{Created on 2023-02-15 with reprex v2.0.2.9000}

[DavisVaughan]

We previously created an environment sort of like env_current from inside dplyr_lazy_vec_chop_impl and set it as the parent env of chops, but that meant we would always access that end on the R side using parent.env(private$chops) and that was a little clunky.

Creating the env on the R side and passing it in seems a little cleaner, and means we have easy access to env_current and its "current" group information variables for use in n() and cur_group_id().

[lionel-]

Nice improvement. I agree this makes the code clearer.

[lionel-]

Can we find a more informative name? current_group_info?

[DavisVaughan]

Important to note that these return the exact reference to the object we modify by reference at the C level.

This is why we duplicate() the result before returning from n() and cur_group_id(). If we just returned the variable without duplicating then it could get modified by reference accidentally

[lionel-]

Nice improvement. I agree this makes the code clearer.

[lionel-]

Can we find a more informative name? current_group_info?