Check for date-like inconsistencies in character columns

Examines specified character columns of a data frame and tests whether their non-missing values can be parsed as dates under a given format. If at least 80% (but less than 100%) of a column's non-missing values parse successfully, the column is flagged, and the row positions of the non-parsable ("straggler") values are reported for inspection. Unlike check_numeric_like(), this check does not attempt to guess a date format automatically - the caller must supply the expected format for each column to be checked, since date formats are ambiguous (for example, "03/04/2023" could mean March 4th or April 3rd depending on convention) and guessing incorrectly could silently produce misleading results.

Usage

check_date_like(data, date_columns)

Arguments

data: A data frame.
date_columns: A named character vector, where each name is a column in data to check, and each value is the expected date format for that column, using the format codes accepted by base::as.Date() (for example, "%Y-%m-%d" or "%m/%d/%Y"). An error is raised if any name in date_columns does not match a column in data.

Value

A named list, one element per flagged column (named by column name). Each element is itself a list containing:

expected_type: Always "date".
proportion_parsable_to_date: Percentage of non-missing values that were successfully parsed under the given format, rounded to 2 decimals.
bad_rows: Integer vector of row indices whose values could not be parsed under the given format.

Returns an empty list if no columns are flagged.

Examples

df <- data.frame(
  signup_date = c(
    "2023-01-15", "2023-02-20", "not a date",
    "2023-03-10", "2023-04-01"
  ),
  notes = c("a", "b", "c", "d", "e")
)
check_date_like(df, date_columns = c(signup_date = "%Y-%m-%d"))
#> $signup_date
#> $signup_date$expected_type
#> [1] "date"
#> 
#> $signup_date$proportion_parsable_to_date
#> [1] 80
#> 
#> $signup_date$bad_rows
#> [1] 3
#> 
#>