Bug Report

Description

Selecting all NCA parameters and running NCA causes the app to crash with:

Error in [<-: Can't recycle input of size 996 to size 83.

preceded by:

Warning: Values from `PPSTRES` are not uniquely identified; output will contain list-cols.

Root Cause

When all parameters are selected, PKNCA produces result rows with PPTESTCD = NA and PPSTRES = NA. These are internal dependency calculations that PKNCA runs but doesn't expose as named parameters.

In pivot_wider_pknca_results(), these NA-keyed rows create duplicate keys for pivot_wider() (multiple NA rows per subject/interval), which produces list-columns instead of atomic vectors. Downstream code then crashes trying to assign list-columns into a regular data frame.

Additional finding: missing metadata mappings

Comparing PKNCA::get.interval.cols() with metadata_nca_parameters, 4 PKNCA parameters have no CDISC mapping:

These are not the direct cause of the crash (the NA rows are), but they should be added to the metadata for completeness. When selected, translate_terms() returns the raw PKNCA name unchanged, which could cause issues in CDISC exports.

Debugging evidence

# 24 duplicate rows, all with PPTESTCD = NA
dupes %>% select(USUBJID, PPTESTCD, PPSTRES, start, end, DOSNOA) %>% print(n = 10)
# A tibble: 24 x 6
   USUBJID PPTESTCD PPSTRES start   end DOSNOA
   <chr>   <chr>      <dbl> <dbl> <dbl>  <int>
 1 S1-01   NA            NA     0  24.1      1
 2 S1-01   NA            NA     0  24.1      1
 3 S1-01   NA            NA     0  24.1      1
 4 S1-01   NA            NA     0  24.1      1
 ...

Steps to Reproduce

1. Upload data with multiple study types
2. Go to Parameter Selection -> click "Select all"
3. Click "Run NCA"
4. App crashes

Fix

Filter out NA PPTESTCD rows before pivoting in pivot_wider_pknca_results(). See PR #1232.

Separately, consider adding the 4 missing parameters to metadata_nca_parameters for completeness."