(v3.0.1.9132) Make translating around 50 times faster and coercion of unknown input in as.ab() up to 70 times faster with identical results, use DOIs instead of PubMed URLs, shorten an example line, and reword cran-comments.md

This commit is contained in:
dr. M.S. (Matthijs) Berends committed 2026-10-14 20:17:28 +02:00
1 parent 4115db67fa
commit 1a1a3abed1
12 files changed
+164 -112

No files matched your search

+28 -2
View File
@@ -37,11 +37,15 @@
# 2. Coercion caches, `AMR_env$ab_previously_coerced` and `AMR_env$mo_previously_coerced`, that remember the
# outcome of earlier coercions. Both have the same structure: a `key` (the cleaned input together with the
# settings that affect the outcome), the original `input`, and the resulting code as `value`.
# 3. For microorganisms, the mapping from (outdated) codes to codes of currently accepted names, which is
# 3. Translated names of antimicrobials and microorganisms, built once per language, see get_ab_translated_index() and
# get_mo_translated_index().
# 4. For microorganisms, the mapping from (outdated) codes to codes of currently accepted names, which is
# memoised per unique code, see synonym_mo_to_accepted_mo().
#
# All of these are invalidated with reset_ab_cache() or reset_mo_cache(), which must be called whenever the
# lookup tables change, e.g. in add_custom_antimicrobials() and add_custom_microorganisms().
# The prepared translation tables of translate_into_language(), see get_translation_table(), are kept for the whole
# session, as the internal translations never change.
new_coercion_cache <- function() {
data.frame(
@@ -92,6 +96,7 @@ coercion_cache_add <- function(type, keys, inputs, values) {
reset_ab_cache <- function() {
AMR_env$AB_index <- NULL
AMR_env$AB_translated <- NULL
AMR_env$custom_ab_id_keys <- NULL
AMR_env$ab_previously_coerced <- new_coercion_cache()
}
@@ -111,16 +116,37 @@ get_ab_index <- function() {
atc_ab <- rep(ab, lengths(AMR_env$AB_lookup$atc))
synonyms <- tolower(unlist(AMR_env$AB_lookup$synonyms, use.names = FALSE))
synonyms_ab <- rep(ab, lengths(AMR_env$AB_lookup$synonyms))
# generalised LOINC codes, synonyms, and abbreviations as flat vectors, with the row number in `AMR_env$AB_lookup`
flat <- function(col) {
vals <- AMR_env$AB_lookup[[col]]
list(value = unlist(vals, use.names = FALSE), row = rep(seq_along(vals), lengths(vals)))
}
gen_syn <- flat("generalised_synonyms")
# synonyms of at least 5 characters, for the Levenshtein distance in as.ab()
gen_syn_long <- which(!is.na(gen_syn$value) & nchar(gen_syn$value) >= 5)
AMR_env$AB_index <- list(
atc = atc[!is.na(atc)],
atc_ab = atc_ab[!is.na(atc)],
synonyms = synonyms[!is.na(synonyms)],
synonyms_ab = synonyms_ab[!is.na(synonyms)]
synonyms_ab = synonyms_ab[!is.na(synonyms)],
generalised_loinc = flat("generalised_loinc"),
generalised_synonyms = gen_syn,
generalised_synonyms_long = list(value = gen_syn$value[gen_syn_long], row = gen_syn$row[gen_syn_long]),
generalised_abbreviations = flat("generalised_abbreviations")
)
}
AMR_env$AB_index
}
get_ab_translated_index <- function(language) {
# translated and generalised names of all antimicrobials, used by as.ab() for input that could not be matched otherwise
# this is built once per language
if (is.null(AMR_env$AB_translated[[language]])) {
AMR_env$AB_translated[[language]] <- generalise_antibiotic_name(translate_AMR(AMR_env$AB_lookup$name, language = language))
}
AMR_env$AB_translated[[language]]
}
get_mo_index <- function() {
add_MO_lookup_to_AMR_env()
if (is.null(AMR_env$MO_index)) {