hh-shipments 0.3.8 → 0.3.9

package/package.json

@@ -18,7 +18,7 @@
     "test": "node test/run-tests.js && node test/run-carriers-tests.js"
   },
   "type": "commonjs",
-  "version": "0.3.8",
+  "version": "0.3.9",
   "bin": {
     "hh-shipments": "server.js"
   },

package/src/extract.js

@@ -76,7 +76,13 @@
     var m, re = /\b(?:1Z|12|IZ)[\s-]?(?:[0-9A-Z][\s-]?){16}\b/gi;
     while ((m = re.exec(text)) !== null) {
       var ups = normalize(m[0]);
-      var num = /^1Z[0-9A-Z]{16}$/.test(ups) ? ups : repairUpsZ(ups);
+      var num = null;
+      if (/^1Z[0-9A-Z]{6}[0-9]{10}$/.test(ups)) num = ups; // fully strict
+      else if (/^1Z[0-9A-Z]{16}$/.test(ups)) {
+        // letters sitting in the digit zone are OCR confusions — repair them
+        num = repairUpsTail(ups.slice(2)) || ups;
+      }
+      else num = repairUpsZ(ups) || repairUpsTail(ups.slice(2));
       if (num && !seen[num]) {
         seen[num] = true;
         found.push({ number: num, carrier: 'ups', confidence: conf, ocr_fixed: num !== ups });
@@ -84,6 +90,39 @@
     }
   }
 
+  // Tesseract also mangles punctuation on labels: ")" / "]" / "}" for J,
+  // "(" / "[" / "{" for C, "|" / "!" for 1 (e.g. "1Z 1X) 539 ..." for 1XJ).
+  // The UPS structure is strict (1Z + 6-char shipper + 2-digit service +
+  // 8-digit serial), so a mangled tail can be repaired and accepted only
+  // when the repaired result validates strictly — never invented.
+  var OCR_PUNCT_FIX = { ')': 'J', ']': 'J', '}': 'J', '(': 'C', '[': 'C', '{': 'C', '|': '1', '!': '1' };
+  var OCR_TAIL_DIGIT_FIX = { 'O': '0', 'Q': '0', 'I': '1', 'l': '1', '|': '1', '!': '1', 'Z': '2', 'S': '5', 'B': '8' };
+  function repairUpsTail(tail) {
+    // tail: 16 chars after the 1Z anchor (6-char shipper + 2-digit service +
+    // 8-digit serial), possibly with OCR confusions.
+    var ship = tail.slice(0, 6), digits = tail.slice(6, 16);
+    var fixShip = ship.replace(/[)\]}(\[{|!]/g, function (c) { return OCR_PUNCT_FIX[c]; });
+    var fixDigits = digits.replace(/[OQIil|!ZSB]/g, function (c) { return OCR_TAIL_DIGIT_FIX[c]; });
+    var cand = '1Z' + fixShip + fixDigits;
+    if (!UPS_STRICT.test(cand)) return null;
+    if (!/[A-Z]/.test(cand.slice(2, 8))) return null; // real shipper numbers contain a letter
+    return cand;
+  }
+  function looseUpsIn(text, found, seen, conf) {
+    // Loose: anchor + 16 chars from an OCR-confusion-tolerant alphabet,
+    // spaces/dashes allowed. repairUpsTail decides; strictness lives there.
+    var m, re = /\b(?:1Z|12|IZ)[\s-]?((?:[0-9A-Z)\]}(\[{|!][\s-]?){16})\b/gi;
+    while ((m = re.exec(text)) !== null) {
+      var tail = m[1].replace(/[\s-]/g, '').toUpperCase();
+      if (tail.length !== 16) continue;
+      var num = repairUpsTail(tail);
+      if (num && !seen[num]) {
+        seen[num] = true;
+        found.push({ number: num, carrier: 'ups', confidence: conf, ocr_fixed: true });
+      }
+    }
+  }
+
   function extractTrackingNumbers(text) {
     var found = [];
     var seen = {};
@@ -95,6 +134,7 @@
     anchorZones(text).forEach(function (z) {
       runPatterns(z.text, found, seen, z.carrier);
       spacedUpsIn(z.text, found, seen, 'high');
+      looseUpsIn(z.text, found, seen, 'high');
     });
     // 1) Whole-text strict patterns
     runPatterns(text, found, seen, false);
@@ -194,6 +234,7 @@
     // a) UPS number split by spaces/dashes (whole-text fallback; the anchor
     // pass already tried the hot zones at high confidence)
     spacedUpsIn(text, found, seen, 'medium');
+    looseUpsIn(text, found, seen, 'medium');
     // b) long digit-like runs with OCR confusions -> fix -> re-match digit formats
     var runRe = /[0-9OILSBoilszb]{10,24}/g, run;
     while ((run = runRe.exec(text)) !== null) {