mirror of
https://github.com/msberends/AMR.git
synced 2026-11-16 07:10:50 +01:00
(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:
1 parent
4115db67fa
commit
1a1a3abed1
12 files changed
+164
-112
No files matched your search
+92
-61
@@ -200,9 +200,86 @@ translate_into_language <- function(from,
|
||||
return(from)
|
||||
}
|
||||
|
||||
df_trans <- TRANSLATIONS # internal data file
|
||||
trns <- get_translation_table(lang, only_unknown = only_unknown, only_affect_ab_names = only_affect_ab_names, only_affect_mo_names = only_affect_mo_names)
|
||||
if (is.null(trns)) {
|
||||
return(from)
|
||||
}
|
||||
from.bak <- from
|
||||
from_unique <- unique(from)
|
||||
|
||||
# check if text to look for is in one of the patterns
|
||||
any_form_in_patterns <- tryCatch(
|
||||
# the same as `from_unique %like% trns$any_pattern`, but without validating the long pattern on every call
|
||||
any(grepl(trns$any_pattern, tolower(from_unique), fixed = trns$any_pattern_fixed, perl = !trns$any_pattern_fixed)),
|
||||
error = function(e) {
|
||||
warning_("Translation not possible. Please create an issue at {.url https://github.com/msberends/AMR/issues}. Many thanks!")
|
||||
return(FALSE)
|
||||
}
|
||||
)
|
||||
if (!any_form_in_patterns) {
|
||||
return(from)
|
||||
}
|
||||
|
||||
from_unique_translated <- from_unique
|
||||
# combinations like "Ticarcillin/clavulanic acid" are translated per part, unless the combination itself is in the translations
|
||||
is_combination <- grepl("/", from_unique, fixed = TRUE) & !from_unique %in% trns$pattern_all
|
||||
if (any(is_combination)) {
|
||||
# all parts of all combinations are translated at once
|
||||
parts <- strsplit(from_unique[is_combination], "/", fixed = TRUE)
|
||||
parts_translated <- translate_into_language(unlist(parts, use.names = FALSE), language = lang, only_unknown = only_unknown, only_affect_ab_names = only_affect_ab_names, only_affect_mo_names = only_affect_mo_names)
|
||||
parts_translated <- split(parts_translated, factor(rep(seq_along(parts), lengths(parts)), levels = seq_along(parts)))
|
||||
from_unique_translated[is_combination] <- vapply(FUN.VALUE = character(1), parts_translated, paste, collapse = "/", USE.NAMES = FALSE)
|
||||
}
|
||||
if (any(!is_combination)) {
|
||||
# every pattern is applied to all strings at once, in order of the translation table
|
||||
terms <- from_unique[!is_combination]
|
||||
for (i in seq_along(trns$pattern)) {
|
||||
terms <- gsub(
|
||||
pattern = trns$pattern[i],
|
||||
replacement = trns$replacement[i],
|
||||
x = terms,
|
||||
ignore.case = trns$ignore_case[i],
|
||||
fixed = trns$fixed[i],
|
||||
perl = !trns$fixed[i]
|
||||
)
|
||||
}
|
||||
# replace the placeholders by their translations
|
||||
has_placeholder <- grepl("\001", terms, fixed = TRUE)
|
||||
if (any(has_placeholder)) {
|
||||
found <- gregexpr("\001[^\001\002]{3}\002", terms[has_placeholder], perl = TRUE)
|
||||
regmatches(terms[has_placeholder], found) <- lapply(regmatches(terms[has_placeholder], found), function(p) trns$literal[match(p, trns$placeholder)])
|
||||
}
|
||||
from_unique_translated[!is_combination] <- terms
|
||||
}
|
||||
|
||||
# force UTF-8 for diacritics
|
||||
from_unique_translated <- enc2utf8(from_unique_translated)
|
||||
|
||||
# a kind of left join to get all results back
|
||||
out <- from_unique_translated[match(from.bak, from_unique)]
|
||||
|
||||
if (!identical(from.bak, out) && get_AMR_locale() == lang && is.null(getOption("AMR_locale", default = NULL)) && message_not_thrown_before("translation", entire_session = TRUE) && interactive()) {
|
||||
message(font_blue(word_wrap(
|
||||
"Assuming the ", LANGUAGES_SUPPORTED_NAMES[[lang]]$exonym, " language (",
|
||||
LANGUAGES_SUPPORTED_NAMES[[lang]]$endonym, ") for the AMR package. See `set_AMR_locale()` to change this or to silence this once-per-session note.",
|
||||
as_note = TRUE
|
||||
)))
|
||||
}
|
||||
|
||||
out
|
||||
}
|
||||
|
||||
# the translation table for a language and the settings of translate_into_language(), built once per session as TRANSLATIONS is fixed
|
||||
get_translation_table <- function(lang, only_unknown = FALSE, only_affect_ab_names = FALSE, only_affect_mo_names = FALSE) {
|
||||
key <- paste(lang, only_unknown, only_affect_ab_names, only_affect_mo_names, sep = "|")
|
||||
if (!key %in% names(AMR_env$translation_tables)) {
|
||||
AMR_env$translation_tables[key] <- list(build_translation_table(lang, only_unknown, only_affect_ab_names, only_affect_mo_names))
|
||||
}
|
||||
AMR_env$translation_tables[[key]]
|
||||
}
|
||||
|
||||
build_translation_table <- function(lang, only_unknown, only_affect_ab_names, only_affect_mo_names) {
|
||||
df_trans <- TRANSLATIONS # internal data file
|
||||
# only keep lines where translation is available for this language
|
||||
df_trans <- df_trans[which(!is.na(df_trans[, lang, drop = TRUE])), , drop = FALSE]
|
||||
# and where the original string is not equal to the string in the target language
|
||||
@@ -217,7 +294,7 @@ translate_into_language <- function(from,
|
||||
df_trans <- subset(df_trans, affect_mo_name == TRUE)
|
||||
}
|
||||
if (NROW(df_trans) == 0) {
|
||||
return(from)
|
||||
return(NULL)
|
||||
}
|
||||
|
||||
# default: case sensitive if value if 'case_sensitive' is missing:
|
||||
@@ -225,18 +302,8 @@ translate_into_language <- function(from,
|
||||
# default: not using regular expressions if 'regular_expr' is missing:
|
||||
df_trans$regular_expr[is.na(df_trans$regular_expr)] <- FALSE
|
||||
|
||||
# check if text to look for is in one of the patterns
|
||||
any_form_in_patterns <- tryCatch(
|
||||
any(from_unique %like% paste0("(", paste(gsub(" +\\(.*", "", df_trans$pattern), collapse = "|"), ")")),
|
||||
error = function(e) {
|
||||
warning_("Translation not possible. Please create an issue at {.url https://github.com/msberends/AMR/issues}. Many thanks!")
|
||||
return(FALSE)
|
||||
}
|
||||
)
|
||||
|
||||
if (NROW(df_trans) == 0 | !any_form_in_patterns) {
|
||||
return(from)
|
||||
}
|
||||
# to check if text to look for is in one of the patterns
|
||||
any_pattern <- paste0("(", paste(gsub(" +\\(.*", "", df_trans$pattern), collapse = "|"), ")")
|
||||
|
||||
if (only_affect_ab_names == TRUE) {
|
||||
df_trans$pattern[df_trans$regular_expr == TRUE] <- paste0(df_trans$pattern[df_trans$regular_expr == TRUE], "$")
|
||||
@@ -257,52 +324,16 @@ translate_into_language <- function(from,
|
||||
replacements,
|
||||
vapply(FUN.VALUE = character(1), seq_along(replacements), function(i) sub(literals[i], placeholders[i], replacements[i], fixed = TRUE))
|
||||
)
|
||||
translate_exec <- function(term) {
|
||||
for (i in seq_len(nrow(trns))) {
|
||||
term <- gsub(
|
||||
pattern = trns$pattern[i],
|
||||
replacement = replacements_placeholder[i],
|
||||
x = term,
|
||||
ignore.case = !trns$case_sensitive[i] & trns$regular_expr[i],
|
||||
fixed = !trns$regular_expr[i],
|
||||
perl = trns$regular_expr[i]
|
||||
)
|
||||
}
|
||||
for (i in which(vapply(FUN.VALUE = logical(1), placeholders, grepl, x = term, fixed = TRUE))) {
|
||||
term <- gsub(placeholders[i], literals[i], term, fixed = TRUE)
|
||||
}
|
||||
term
|
||||
}
|
||||
|
||||
from_unique_translated <- from_unique
|
||||
# combinations like "Ticarcillin/clavulanic acid" are translated per part, unless the combination itself is in the translations
|
||||
is_combination <- grepl("/", from_unique, fixed = TRUE) & !from_unique %in% df_trans$pattern
|
||||
from_unique_translated[is_combination] <- vapply(
|
||||
FUN.VALUE = character(1),
|
||||
USE.NAMES = FALSE,
|
||||
strsplit(from_unique[is_combination], "/", fixed = TRUE),
|
||||
function(parts) paste(translate_into_language(parts, language = lang, only_unknown = only_unknown, only_affect_ab_names = only_affect_ab_names, only_affect_mo_names = only_affect_mo_names), collapse = "/")
|
||||
list(
|
||||
# all patterns, including combinations, to check whether a combination is translated as a whole
|
||||
pattern_all = df_trans$pattern,
|
||||
any_pattern = tolower(any_pattern),
|
||||
any_pattern_fixed = !is_valid_regex(any_pattern),
|
||||
pattern = trns$pattern,
|
||||
replacement = replacements_placeholder,
|
||||
ignore_case = !trns$case_sensitive & trns$regular_expr,
|
||||
fixed = !trns$regular_expr,
|
||||
placeholder = placeholders,
|
||||
literal = literals
|
||||
)
|
||||
from_unique_translated[!is_combination] <- vapply(
|
||||
FUN.VALUE = character(1),
|
||||
USE.NAMES = FALSE,
|
||||
from_unique[!is_combination],
|
||||
translate_exec
|
||||
)
|
||||
|
||||
# force UTF-8 for diacritics
|
||||
from_unique_translated <- enc2utf8(from_unique_translated)
|
||||
|
||||
# a kind of left join to get all results back
|
||||
out <- from_unique_translated[match(from.bak, from_unique)]
|
||||
|
||||
if (!identical(from.bak, out) && get_AMR_locale() == lang && is.null(getOption("AMR_locale", default = NULL)) && message_not_thrown_before("translation", entire_session = TRUE) && interactive()) {
|
||||
message(font_blue(word_wrap(
|
||||
"Assuming the ", LANGUAGES_SUPPORTED_NAMES[[lang]]$exonym, " language (",
|
||||
LANGUAGES_SUPPORTED_NAMES[[lang]]$endonym, ") for the AMR package. See `set_AMR_locale()` to change this or to silence this once-per-session note.",
|
||||
as_note = TRUE
|
||||
)))
|
||||
}
|
||||
|
||||
out
|
||||
}
|
||||
Reference in new issue
Block a user