standardize "from"

[Aariq]

In all of the get_() functions, the argument to specify what type of identifier the query is is called from, but in a couple of functions, this argument is type.

• ci_query(query, type = "inchikey")
• aw_query(query, type = "cas")

A couple of functions don't have query and from because they only accept one identifier type in queries. Maybe these are OK as is?

• fn_percept(CAS)
• cts_compinfo(inchikey)

In addition, the possible values that from accepts are mostly standardized, except for:

• aw_query, which uses "commonname" instead of "name"
• get_chebiid() which is more complicated:

character; type of input, can be one of 'ALL', 'CHEBI ID', 'CHEBI NAME', 'DEFINITION', 'ALL NAMES', 'IUPAC NAME', 'CITATIONS', 'REGISTRY NUMBERS', 'MANUAL XREFS', 'AUTOMATIC XREFS', 'FORMULA', 'MASS', 'MONOISOTOPIC MASS', 'CHARGE', 'INCHI/INCHI KEY', 'SMILES', 'SPECIES'.

It might be nice if this wasn't case-sensitive and if both "inchi" and "inchikey" were options rather than "INCHI/INCHI KEY" and maybe if "name" was a synonym for "ALL NAMES" to better standardize across the package.

[Aariq]

One advantage of standardizing all this stuff is the possibility of something like an "autotranslate" feature:

get_auto <- function(query, from, .f, ...) {
  f <- rlang::as_function(.f)
  pos_froms <- eval(rlang::fn_fmls(f)$from)
  if (from %in% pos_froms) {
    f(query, from, ...)
  } else {
    new_from <- pos_froms[pos_froms != "name"][1] #cts name conversion broken
    warning(glue::glue("{.f} doesn't accept {from}. Attempting to translate to {new_from} with CTS and re-running query "))
    new_query <- cts_convert(query, from = from, to = new_from, choices = 1)
    f(new_query, new_from, ...)
  }
}

get_auto('1071-83-6', "cas", "get_etoxid")
get_auto('XDDAORKBJWWYJS-UHFFFAOYSA-N', "inchikey", "get_etoxid")

Not sure how important this is for the average user, but might be useful in the future for functions that integrate across databases.

[stitam]

Thanks @Aariq, I think it would be great to standardise available from arguments across the package and to consistently use from (and not type) everywhere. Probably simpler for all users and also for us to build integrators later on. Maybe I would separate, and call get_id() the one that returns database specific ids like csid or cid, and convert_id() the one that converts between universal identifiers like cas or inchi. On the other hand, some get_*() functions are already quite complex like get_cid() I wonder how the documentation would look like for a general get() integrator. Also when calling get_*(), we would look up a specific database so there is no point searching the rest. Maybe it's more straightforward to build a universal convert_id() function?

[Aariq]

Yeah, I don't know the final form of this function. It could work kinda like purrr::safely(), where it returns a modified function (one that auto-translates ID's). Or it could be get_id(query, from, database). Or maybe it's something we only use internally for figuring out database coverage for compounds. But it sounds like you're ok with standardizing from wherever it makes sense, so I'll work on that as a start.