Dealing with the "observationTags" column of observations

[ddachs]

When dealing with custom attributes one has to extract them from the observationTags column. Otherwise the column is useless.

IMO the ideal behaviour of the observations() function would be to split the observationTags n columns, where n beeing the number of unique tags. At least the option would be nice to have e.g. ( observations(x, splitObservationTags = T)

Something like this:

convert_customTags <- function(x) {

observation_tags <-
  separate_wider_delim (
    x %>% filter(!is.na(observationTags)) %>%
      select(observationID, observationTags),
      observationTags,
      names = LETTERS[1:10],
      too_few = "align_start",
      delim = "|",
      cols_remove = T
                                         ) %>%
  pivot_longer(cols = (A:J),
                        names_to = "letter",
                        values_to = "text") %>%
  filter(!is.na(text)) %>%
  select(-letter) %>%
  separate_wider_delim(cols = "text",
                                     delim = ":",
                                     names = c("tag", "value"),
                                     too_few = "align_start") %>%
  pivot_wider(names_from = tag, values_from = value) %>%
  as.data.table()

y <- merge.data.table(observations, observation_tags, on= "observationID", all.x = T)

return(y)
}

Although this has the flaw, that it only works up to 10 custom attributes. Maybe someone has a better idea?

[peterdesmet]

That is a good suggestion. The same is true for deploymentTags. Will have to think how to best implement this. Options:

• Unpack tags with dedicated function
• Unpack when calling observations(), optionally with a parameter
• As part of read_camtrapdp() cf. how we assign the taxonomic scope to the observations as separate columns

@sannegovaert @PietrH @damianooldoni thoughts?