Tighten arXiv extraction boundaries

* prevent partial matches inside larger tokens
* reject hyphen- and slash-adjacent false positives
* add regression tests for wrapped and malformed arXiv cases

Author: Thomas-Rauter

R/extract_scholid.R

@@ -128,7 +128,7 @@ extract_orcid <- function(text) {
 #'
 #' @noRd
 extract_isbn <- function(text) {
-    pat <- "(?<![[:alnum:]_])([0-9Xx][0-9Xx\\- ]{8,16}[0-9Xx])(?![[:alnum:]_])"
+    pat <- "(?<![[:alnum:]_])([0-9Xx][0-9Xx\\- ]{8,16}[0-9Xx])(?![[:alnum:]_\\-/])"
     out <- .extract_with_pattern(
         text = text,
         pat  = pat
@@ -202,11 +202,38 @@ extract_issn <- function(text) {
 #'
 #' @noRd
 extract_arxiv <- function(text) {
-    pat <- "(\\d{4}\\.\\d{4,5}(v\\d+)?|[a-z\\-]+/\\d{7}(v\\d+)?)"
-    .extract_with_pattern(
+    pat <- paste0(
+        "(?<![[:alnum:]_\\./-])",
+        "(",
+        "\\d{4}\\.\\d{4,5}(v\\d+)?",
+        "|",
+        "[a-z\\-]+/\\d{7}(v\\d+)?",
+        ")",
+        "(?![[:alnum:]_\\-/])"
+    )
+
+    out <- .extract_with_pattern(
         text = text,
         pat  = pat
     )
+
+    lapply(out, function(hits) {
+        if (!length(hits)) {
+            return(character(0))
+        }
+
+        cleaned <- vapply(
+            hits,
+            .clean_extracted_arxiv,
+            character(1),
+            USE.NAMES = FALSE
+        )
+
+        cleaned <- cleaned[nzchar(cleaned)]
+        cleaned <- cleaned[!is.na(cleaned)]
+        cleaned <- cleaned[is_arxiv(cleaned)]
+        cleaned
+    })
 }
 
 
@@ -376,6 +403,28 @@ extract_pmcid <- function(text) {
 }
 
 
+#' Clean an extracted arXiv candidate
+#'
+#' @description
+#' Removes trailing punctuation and surrounding whitespace from an extracted
+#' arXiv candidate.
+#'
+#' @param x A single extracted arXiv candidate.
+#'
+#' @return A cleaned arXiv candidate string, or `""` if empty.
+#'
+#' @noRd
+.clean_extracted_arxiv <- function(x) {
+    if (is.na(x) || !nzchar(x)) {
+        return("")
+    }
+
+    x <- sub("[[:space:][:punct:]]+$", "", x, perl = TRUE)
+    x <- trimws(x)
+    x
+}
+
+
 # Level 3 functions (functions called by level 2 functions) definitions --------
 
 

tests/testthat/test-extract_scholid.R

@@ -281,6 +281,94 @@ testthat::test_that(
     }
 )
 
+testthat::test_that(
+    "extract_arxiv rejects partial matches inside larger tokens",
+    {
+        txt <- c(
+            "bad modern 21011.12345",
+            "bad version 2101.12345v",
+            "bad version hep-th/9901001v",
+            "phone-like 2101.12345-90",
+            "slash tail 2101.12345/extra",
+            "arXiv:2101.12345v2",
+            "legacy hep-th/9901001v3"
+        )
+
+        got <- extract_scholid(
+            txt,
+            "arxiv"
+        )
+
+        testthat::expect_identical(
+            got[[1]],
+            character(0)
+        )
+        testthat::expect_identical(
+            got[[2]],
+            character(0)
+        )
+        testthat::expect_identical(
+            got[[3]],
+            character(0)
+        )
+        testthat::expect_identical(
+            got[[4]],
+            character(0)
+        )
+        testthat::expect_identical(
+            got[[5]],
+            character(0)
+        )
+        testthat::expect_identical(
+            got[[6]],
+            "2101.12345v2"
+        )
+        testthat::expect_identical(
+            got[[7]],
+            "hep-th/9901001v3"
+        )
+    }
+)
+
+testthat::test_that(
+    "extract_arxiv handles wrappers and multiple valid matches",
+    {
+        txt <- c(
+            "Quoted '2101.12345'.",
+            "Wrapped (2101.12345v2).",
+            "Two IDs: 2101.12345 and hep-th/9901001v2.",
+            "[hep-th/9901001]",
+            "{math/0303001}."
+        )
+
+        got <- extract_scholid(
+            txt,
+            "arxiv"
+        )
+
+        testthat::expect_identical(
+            got[[1]],
+            "2101.12345"
+        )
+        testthat::expect_identical(
+            got[[2]],
+            "2101.12345v2"
+        )
+        testthat::expect_identical(
+            got[[3]],
+            c("2101.12345", "hep-th/9901001v2")
+        )
+        testthat::expect_identical(
+            got[[4]],
+            "hep-th/9901001"
+        )
+        testthat::expect_identical(
+            got[[5]],
+            "math/0303001"
+        )
+    }
+)
+
 testthat::test_that(
     "extract_arxiv finds modern and legacy arXiv identifiers",
     {