spcr_make_data_bundle() errors if empty a_data is supplied

[francisbarton]

Reprex:

library(lubridate, warn.conflicts = FALSE)

test_data <- list(
  month = tibble::tribble(
    ~ref, ~measure_name, ~comment
  ),
  events = tibble::tribble(
    ~ref, ~measure_name, ~comment, ~event_date_or_datetime,
    "1", "Test data", "Reprex for duplicate date", as_date("2023-12-18"),
    "1", "Test data", "Reprex for duplicate date", as_date("2023-12-18")
  )
)

test_rc <- tibble::tribble(
  ~ref, ~measure_name, ~domain, ~spc_chart_type, ~aggregation, ~report_comment,
  "1", "Test data", NA_character_, "t", "none", NA_character_
)

test_mc <- tibble::tribble(
  ~ref, ~measure_name, ~unit, ~data_source, ~data_owner, ~accountable_person, ~improvement_direction, ~target, ~target_set_by, ~data_quality, ~rebase_dates, ~rebase_comment,
  "1", "Test data", "integer", "My brain", "Me", "nobody", "decrease", 0, "Me", "GGGG", "", ""
)

test_data_bundle <- SPCreporter::spcr_make_data_bundle(
  test_data, test_rc, test_mc
)
#> ℹ check_for_optional_columns: Optional column 'allowable_days_lag' is missing.
#> Adding it.
#> Error in `tidyr::pivot_longer()` at SPCreporter/R/helper_functions.R:38:3:
#> ! `cols` must select at least one column.
#> Backtrace:
#>      ▆
#>   1. └─SPCreporter::spcr_make_data_bundle(test_data, test_rc, test_mc)
#>   2.   ├─dplyr::bind_rows(lengthen_measure_data(a_data_df), e_data_time_between) at SPCreporter/R/spcr_make_data_bundle.R:51:3
#>   3.   │ └─rlang::list2(...)
#>   4.   └─SPCreporter:::lengthen_measure_data(a_data_df)
#>   5.     ├─dplyr::arrange(...) at SPCreporter/R/helper_functions.R:38:3
#>   6.     ├─dplyr::mutate(...) at SPCreporter/R/helper_functions.R:38:3
#>   7.     ├─tidyr::pivot_longer(...) at SPCreporter/R/helper_functions.R:38:3
#>   8.     └─tidyr:::pivot_longer.data.frame(.data, !any_of(init_cols), names_to = "date", values_drop_na = TRUE)
#>   9.       └─tidyr::build_longer_spec(...)
#>  10.         └─cli::cli_abort(...)
#>  11.           └─rlang::abort(...)

If only a data frame that contains no rows is supplied within measure data, then spcr_make_data_bundle()'s check and shape process will error at the lengthen_measure_data() step.

[ThomUK]

Human Input Needed

The current check_a_data() function already contains an assertion ncol(x) > length(required_columns) that produces a helpful error when no date columns are present in the aggregated data. This appears to address the case in the reprex (where month has only required columns and no date columns).

Please verify whether the current code already handles this case with an informative error message. If the desired behaviour is to silently skip the empty data rather than error, please clarify that intent.

[ThomUK]

Although the check does produce an error, in the specific case of the reprex we want to be able to produce t charts. The general case should be that a_data is only required when one or more xmr charts are requested, and e_data is only required when one or more t charts are requested.

For clarity, the reprex now results in this error:

> library(lubridate, warn.conflicts = FALSE)
> test_data <- list(
+   month = tibble::tribble(
+     ~ref, ~measure_name, ~comment
+   ),
+   events = tibble::tribble(
+     ~ref, ~measure_name, ~comment, ~event_date_or_datetime,
+     "1", "Test data", "Reprex for duplicate date", as_date("2023-12-18"),
+     "1", "Test data", "Reprex for duplicate date", as_date("2023-12-18")
+   )
+ )
> test_rc <- tibble::tribble(
+   ~ref, ~measure_name, ~domain, ~spc_chart_type, ~aggregation, ~report_comment,
+   "1", "Test data", NA_character_, "t", "none", NA_character_
+ )
> test_mc <- tibble::tribble(
+   ~ref, ~measure_name, ~unit, ~data_source, ~data_owner, ~accountable_person, ~improvement_direction, ~target, ~target_set_by, ~data_quality, ~rebase_dates, ~rebase_comment,
+   "1", "Test data", "integer", "My brain", "Me", "nobody", "decrease", 0, "Me", "GGGG", "", ""
+ )
> test_data_bundle <- SPCreporter::spcr_make_data_bundle(
+   test_data, test_rc, test_mc
+ )
ℹ check_for_optional_columns: Optional column 'allowable_days_lag' is missing. Adding it.
Error in `map2()`:
ℹ In index: 1.
ℹ With name: month.
Caused by error:
! measure_data: No date columns found in the 'month' sheet or dataframe. The data must contain at least one and probably more date column(s) (which will contain the data to be plotted).
Run `rlang::last_trace()` to see where the error occurred.

[ThomUK]

Implementation Plan

When all requested chart types are 't', empty a_data (the 'month' sheet) should be silently skipped rather than erroring. Conversely, empty e_data should be skipped when no 't' charts are requested.

Approach:

• In spcr_make_data_bundle(), before calling lengthen_measure_data(a_data_df), check if a_data_df has no rows (empty); if so, skip to an empty tibble rather than attempting to pivot
• Similarly, guard the e_data processing path so an empty events data frame is handled gracefully when no t-charts are present
• The check should be based on whether the data frame has 0 rows (not 0 columns), as the schema/columns are still provided

Tests:

• Calling spcr_make_data_bundle() with an empty month data frame and valid events data for a 't' chart type succeeds without error
• Calling spcr_make_data_bundle() with valid month data and no events data for an 'xmr' chart type succeeds without error
• Calling spcr_make_data_bundle() with empty month data when a non-t chart type is requested still raises an informative error