should we validate cohorts at each step?

[edward-burn]

@nmercadeb @catalamarti I think we did discuss this before and decided not to validate the cohort at each step to be more performant, but maybe we should think about this again. Below for example I can pass dates out of observation, but maybe the default should be to validate and error? I suppose in the example below, we have the choice whether this should be no error (like now), cause an error (which would happen if we passed cohort validation internally), or the record should be dropped.

library(CohortConstructor)
cdm <- mockCohortConstructor(tables = list(
"cohort" = dplyr::tibble(
  cohort_definition_id = 1,
  subject_id = c(1, 2, 3, 4),
  cohort_start_date = as.Date(c("2000-06-03", "2000-01-01", "2015-01-15", "2000-12-09")),
  cohort_end_date = as.Date(c("2001-09-01", "2001-01-12", "2015-02-15", "2002-12-09")),
  date_1 = as.Date(c("2001-08-01", "1900-01-01", "2015-01-15", "2002-12-09")),
  date_2 = as.Date(c("2001-08-01", NA, "2015-02-14", "2002-12-09"))
)
))
cdm$cohort <- cdm$cohort |> 
  entryAtFirstDate(dateColumns = c("date_1", "date_2"))

cdm$cohort |> 
  omopgenerics::checkCohortRequirements() 
#> Error:
#> ! 1 observation outside observation period.

^{Created on 2024-06-03 with reprex v2.1.0}

[nmercadeb]

only use in functions that modify cohort start and end