pk.nca.intervals: technical question on the cbind action

[Gero1999]

This is a technical question. I am inspecting the internal functions inside pk.nca and exploring with different custom intervals self-made and custom added columns. However, in my multiple dose study seems that there is something missing that produces the next scale of errors while debugging:

Error during wrapup: arguments imply differing number of rows: 1, 14
Error: no more error handlers available (recursive errors?); invoking 'abort' restart

After investigating I spotted that the problem is indeed in pk.nca.intervals, in the cbind merging each interval with its results through a loop:

  for (i in seq_len(nrow(data_intervals))) {
  
  # rest of the code
  
      ret <- rbind(ret, cbind(current_interval[, c("start", 
        "end", options$keep_interval_cols)], calculated_interval, 
        row.names = NULL))
            }
  }

at this point my datasets look like these, which for me makes sense. it is just the current interval row computed (with two keep_columns for the interval: DOSNO & type_interval) and its associated calculations:

> current_interval[, c("start",  "end", options$keep_interval_cols)]
# A tibble: 1 × 4
  start   end DOSNO type_interval
  <dbl> <dbl> <int> <chr>        
1   1.5   504     1 main         

> calculated_interval
              PPTESTCD      PPORRES exclude
1                 cmax 8.660000e+00    <NA>
2                 tmax 6.000000e+00    <NA>
3                tlast 3.821700e+02    <NA>
4            clast.obs 6.300000e-03    <NA>
5             lambda.z 2.001934e-02    <NA>
6            r.squared 9.985178e-01    <NA>
7        adj.r.squared 9.970357e-01    <NA>
8  lambda.z.time.first 2.575000e+01    <NA>
9    lambda.z.n.points 3.000000e+00    <NA>
10          clast.pred 6.429894e-03    <NA>
11           half.life 3.462389e+01    <NA>
12          span.ratio 1.029405e+01    <NA>
13          aucinf.obs 6.275676e+02    <NA>
14              cl.obs           NA    <NA>

The problem is indeed with the cbind action but I am confused if this is indeed a bug or somehow a bad construction of my interval that I am misising. Aren't these two objects supposed to look like these at this point?

[Gero1999]

I think I found the problem. My original dataset presented attributes (labels) associated to the column names of my datasets, which where included in the keep_columns of my custom intervals. This did not bring any warning nor problem during the use of PKNCAdata(), but at this point in pk.nca() / pk.nca.intervals() it will make incompatible the cbind action...

A potential end-solution will be to prevent the presence of any formatting at this point by:

 current_interval = current_interval %>% mutate(across(everything(), ~ {
    attributes(.) <- NULL
    .
}))

But of course I guess that a better solution would be to either run this action right in all PKNCAdata objects or introduce a warning to the user so other potential issues are prevented.

[billdenney]

R's handling of attributes has often been a big problem for code that I've written (unrelated to PKNCA). Can you please provide a reprex in an issue where this occurs? That way we can find an appropriate solution (which will likely look a lot like what you provided above).

[Gero1999]

I opened an issue with an easy to run example. I will close this discussion now!