Tighten ISSN extraction and normalization rules

* prevent ISSN extraction inside longer hyphenated tokens
* reject invalid ISSNs during normalization
* update tests to use checksum-valid ISSN examples
* add regression coverage for ISSN boundaries and wrapped forms

Author: Thomas-Rauter

R/extract_scholid.R

@@ -165,11 +165,29 @@ extract_isbn <- function(text) {
 #'
 #' @noRd
 extract_issn <- function(text) {
-    pat <- "(\\d{4}-\\d{3}[0-9X])"
-    .extract_with_pattern(
+    pat <- "(?<![[:alnum:]_\\-])(\\d{4}-\\d{3}[0-9Xx])(?![[:alnum:]_\\-])"
+    out <- .extract_with_pattern(
         text = text,
         pat  = pat
     )
+
+    lapply(out, function(hits) {
+        if (!length(hits)) {
+            return(character(0))
+        }
+
+        cleaned <- vapply(
+            hits,
+            .clean_extracted_issn,
+            character(1),
+            USE.NAMES = FALSE
+        )
+
+        cleaned <- cleaned[nzchar(cleaned)]
+        cleaned <- cleaned[!is.na(cleaned)]
+        cleaned <- cleaned[is_issn(cleaned)]
+        cleaned
+    })
 }
 
 
@@ -336,6 +354,28 @@ extract_pmcid <- function(text) {
 }
 
 
+#' Clean an extracted ISSN candidate
+#'
+#' @description
+#' Removes trailing punctuation and surrounding whitespace from an extracted
+#' ISSN candidate.
+#'
+#' @param x A single extracted ISSN candidate.
+#'
+#' @return A cleaned ISSN candidate string, or `""` if empty.
+#'
+#' @noRd
+.clean_extracted_issn <- 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 --------
 
 

R/normalize_scholid.R

@@ -214,6 +214,9 @@ normalize_issn <- function(x) {
         paste0(substr(y, 1, 4), "-", substr(y, 5, 8))
     )
 
+    # Keep only checksum-valid ISSNs
+    y[!is.na(y) & !is_issn(y)] <- NA_character_
+
     out[ok] <- y
     out
 }

tests/testthat/test-extract_scholid.R

@@ -264,7 +264,7 @@ testthat::test_that(
     "extract_issn finds ISSN with X and ignores non-matches",
     {
         txt <- c(
-            "ISSN 1234-567X",
+            "ISSN 2434-561X",
             "no issn here"
         )
 
@@ -273,7 +273,7 @@ testthat::test_that(
             "issn"
         )
 
-        testthat::expect_true(any(got[[1]] == "1234-567X"))
+        testthat::expect_true(any(got[[1]] == "2434-561X"))
         testthat::expect_identical(
             got[[2]],
             character(0)
@@ -351,7 +351,7 @@ testthat::test_that(
             "arXiv:2101.00001v2",
             "PMCID: PMC12345",
             "PMID 1234567",
-            "ISSN 1234-567X",
+            "ISSN 2434-561X",
             "ISBN 0-306-40615-2"
         )
 
@@ -675,3 +675,79 @@ testthat::test_that(
         )
     }
 )
+
+
+testthat::test_that(
+    "extract_issn respects token boundaries and wrappers",
+    {
+        txt <- c(
+            "ISSN 2434-561X",
+            "Quoted '2434-561X'.",
+            "Wrapped (2434-561X).",
+            "Noise: abc 2434-561X xyz",
+            "Phone-like 2434-561X-90",
+            "Double token 2434-561X-0378-5955",
+            "Prefix hyphen abc-2434-561X",
+            "Surrounded by spaces 2434-561X okay"
+        )
+
+        got <- extract_scholid(
+            txt,
+            "issn"
+        )
+
+        testthat::expect_identical(
+            got[[1]],
+            "2434-561X"
+        )
+        testthat::expect_identical(
+            got[[2]],
+            "2434-561X"
+        )
+        testthat::expect_identical(
+            got[[3]],
+            "2434-561X"
+        )
+        testthat::expect_identical(
+            got[[4]],
+            "2434-561X"
+        )
+        testthat::expect_identical(
+            got[[5]],
+            character(0)
+        )
+        testthat::expect_identical(
+            got[[6]],
+            character(0)
+        )
+        testthat::expect_identical(
+            got[[7]],
+            character(0)
+        )
+        testthat::expect_identical(
+            got[[8]],
+            "2434-561X"
+        )
+    }
+)
+
+testthat::test_that(
+    "detect_scholid_type detects wrapped issn values",
+    {
+        x <- c(
+            "2434-561X",
+            "0378-5955",
+            "ISSN 2434-561X",
+            "issn: 0378-5955",
+            "2434-561x",
+            "not an issn"
+        )
+
+        got <- detect_scholid_type(x)
+
+        testthat::expect_identical(
+            got,
+            c("issn", "issn", "issn", "issn", "issn", NA_character_)
+        )
+    }
+)

tests/testthat/test-normalize_scholid.R

@@ -235,3 +235,29 @@ testthat::test_that(
         )
     }
 )
+
+
+testthat::test_that(
+    "issn normalization rejects invalid values and canonicalizes valid ones",
+    {
+        testthat::expect_false(is_issn("9999-9999"))
+        testthat::expect_false(is_issn("2434-561X-90"))
+
+        testthat::expect_identical(
+            normalize_scholid("9999-9999", "issn"),
+            NA_character_
+        )
+        testthat::expect_identical(
+            normalize_scholid("2434-561X-90", "issn"),
+            NA_character_
+        )
+        testthat::expect_identical(
+            normalize_scholid("2434561X", "issn"),
+            "2434-561X"
+        )
+        testthat::expect_identical(
+            normalize_scholid("2434-561x", "issn"),
+            "2434-561X"
+        )
+    }
+)