convert_to_numeric() in a dataset of 500,000+ rows took 2.5 minutes

[avallecam]

running cleanepi::convert_to_numeric() in a dataset of 500,000+ rows took 2.5 minutes

wondering if this may be an expected scenario to happen and if this may require refactoring at an appropriate time to use data.table or dtplyr.

library(rio)
library(cleanepi)
library(tidyverse)
library(tictoc)

covid <- rio::import(
  "https://raw.githubusercontent.com/Joskerus/Enlaces-provisionales/main/data_limpieza.zip",
  which = "datos_covid_LA.RDS"
) %>% 
  cleanepi::standardize_column_names()

tictoc::tic()
covid %>% 
  dplyr::select(numero_de_hospitalizaciones_recientes) %>% 
  cleanepi::convert_to_numeric(
    target_columns = "numero_de_hospitalizaciones_recientes",
    lang = "es")
#> # A tibble: 502,010 × 1
#>    numero_de_hospitalizaciones_recientes
#>                                    <dbl>
#>  1                                     0
#>  2                                     0
#>  3                                     0
#>  4                                     0
#>  5                                     0
#>  6                                     0
#>  7                                     0
#>  8                                     0
#>  9                                    NA
#> 10                                     0
#> # ℹ 502,000 more rows
#> # ℹ Use `print(n = ...)` to see more rows
tictoc::toc()
#> 150.42 sec elapsed

cc: @Joskerus @lgbermeo

[Bisaloo]

Could you give epiverse-trace/numberize#14 a go please? If the performance is still not sufficient, I have a couple of other ideas.