as_polars_df() for nanoarrow_array_stream seems slow

[eitsupi]

Slower than as_tibble surprised me.

library(adbcdrivermanager)
library(arrow)
#>
#> Attaching package: 'arrow'
#> The following object is masked from 'package:utils':
#>
#>     timestamp
library(tibble)
library(polars)

polars_info()
#> Polars R package version : 0.15.0
#> Rust Polars crate version: 0.38.1
#>
#> Thread pool size: 16
#>
#> Features:
#> default                    TRUE
#> full_features              TRUE
#> disable_limit_max_threads  TRUE
#> nightly                    TRUE
#> sql                        TRUE
#> rpolars_debug_print       FALSE
#>
#> Code completion: deactivated

db <- adbc_database_init(adbcsqlite::adbcsqlite())
con <- adbc_connection_init(db)

flights <- nycflights13::flights
flights$time_hour <- NULL
flights |>
  write_adbc(con, "flights")

query <- "SELECT * from flights LIMIT 10000"

bench::mark(
  polars_df_1 = {
    con |>
      read_adbc(query) |>
      as_polars_df()
  },
  arrow_table = {
    con |>
      read_adbc(query) |>
      as_arrow_table()
  },
  tibble = {
    con |>
      read_adbc(query) |>
      as_tibble()
  },
  polars_df_2 = {
    con |>
      read_adbc(query) |>
      as_polars_df()
  },
  check = FALSE,
  min_iterations = 5
)
#> # A tibble: 4 × 6
#>   expression       min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>  <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 polars_df_1  158.5ms  158.8ms      5.58    1.13MB     3.72
#> 2 arrow_table   36.5ms     39ms     25.6     1.62MB     2.13
#> 3 tibble        40.8ms   43.2ms     22.9     1.62MB     0
#> 4 polars_df_2  153.8ms  185.8ms      5.38    89.1KB     8.07

^{Created on 2024-03-04 with reprex v2.0.2}

[eitsupi]

I think the content of as_polars_df.nanoarrow_array_stream is bad.
Proof of this is that converting to DataFrame after converting to Series is not slow.

I will rewrite that. (Related to #755)

library(adbcdrivermanager)
library(polars)

polars_info()
#> Polars R package version : 0.15.0
#> Rust Polars crate version: 0.38.1
#>
#> Thread pool size: 16
#>
#> Features:
#> default                    TRUE
#> full_features              TRUE
#> disable_limit_max_threads  TRUE
#> nightly                    TRUE
#> sql                        TRUE
#> rpolars_debug_print       FALSE
#>
#> Code completion: deactivated

db <- adbc_database_init(adbcsqlite::adbcsqlite())
con <- adbc_connection_init(db)

flights <- nycflights13::flights
flights$time_hour <- NULL
flights |>
  write_adbc(con, "flights")

query <- "SELECT * from flights LIMIT 10000"

bench::mark(
  polars_df = {
    con |>
      read_adbc(query) |>
      as_polars_df()
  },
  polars_df2 = {
    con |>
      read_adbc(query) |>
      (\(x) as_polars_series(x)$to_frame())()
  },
  check = FALSE,
  min_iterations = 5
)
#> # A tibble: 2 × 6
#>   expression      min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 polars_df   154.1ms  154.1ms      6.49    1.17MB    26.0
#> 2 polars_df2   44.2ms   47.9ms     20.2     1.64MB     2.02

^{Created on 2024-03-04 with reprex v2.0.2}