hh-shipments 0.3.25 → 0.3.26

package/src/extract.js

@@ -10,7 +10,8 @@
     // UPS: 1Z + 16 alphanumerics
     { carrier: 'ups', confidence: 'high', re: /\b(1Z[0-9A-Z]{16})\b/gi },
     // USPS: 20-22 digits starting 92-95 (9400..., 9205..., etc.)
-    { carrier: 'usps', confidence: 'high', re: /\b(9[2-5]\d{18,20})\b/g },
+    // USPS: 20-22 digits starting 91-95 (9400..., 9305..., 9205..., 9101..., etc.)
+    { carrier: 'usps', confidence: 'high', re: /\b(9[1-5]\d{18,20})\b/g },
     // USPS S10 international: 2 letters + 9 digits + 2 letters (ends US = USPS)
     { carrier: 'usps', confidence: 'medium', re: /\b([A-Z]{2}\d{9}US)\b/g },
     // FedEx SmartPost / Ground 22-digit starting 96
@@ -247,9 +248,14 @@
       runPatterns(z.text, found, seen, z.carrier);
       spacedUpsIn(z.text, found, seen, 'high');
       looseUpsIn(z.text, found, seen, 'high');
+      spacedDigitGroupsIn(z.text, found, seen, 'high');
     });
     // 1) Whole-text strict patterns
     runPatterns(text, found, seen, false);
+    // 1b) Spaced digit groups: labels print "9305 5110 4390 0011 2062 42"
+    // with spaces, which strict PATTERNS cannot match. Without this pass a
+    // perfectly-read number is silently dropped.
+    spacedDigitGroupsIn(text, found, seen, 'medium');
     // 2) OCR-tolerant recovery pass
     extractOcrTolerant(text, found, seen);
     var rank = { high: 0, medium: 1, low: 2 };
@@ -294,6 +300,9 @@
         var num = normalize(m[1]);
         if (seen[num]) continue;
         seen[num] = true;
+        // Never accept an IMpb routing prefix (420 + 9 digits) as a FedEx
+        // number — it is postal routing, not a shipment.
+        if (p.carrier === 'fedex' && /^420\d{9}$/.test(num)) continue;
         // Context boost: carrier word near the number, or a hot-zone anchor
         // naming this carrier (a label's "UPS GROUND" line means the number
         // under it really is UPS).
@@ -314,12 +323,61 @@
   function isValidFormat(number, carrier) {
     var n = normalize(number);
     if (carrier === 'ups') return /^1Z[0-9A-Z]{16}$/.test(n);
-    if (carrier === 'usps') return /^9[2-5]\d{18,20}$/.test(n) || /^[A-Z]{2}\d{9}US$/.test(n);
-    if (carrier === 'fedex') return /^(96\d{20}|\d{12}|\d{15}|\d{20})$/.test(n);
+    if (carrier === 'usps') return /^9[1-5]\d{18,20}$/.test(n) || /^[A-Z]{2}\d{9}US$/.test(n);
+    // FedEx: never accept an IMpb routing prefix (420 + 9 digits) — it is
+    // postal routing, not a shipment number.
+    if (carrier === 'fedex') return /^(96\d{20}|\d{12}|\d{15}|\d{20})$/.test(n) && !/^420\d{9}$/.test(n);
     if (carrier === 'dhl') return /^\d{10,11}$/.test(n) || /^[A-Z]{2,4}\d{10,20}$/.test(n);
     return false;
   }
 
+  // Intelligent Mail package barcode: 420 + routing digits + USPS tracking,
+  // e.g. 420078742956 + 9305511043900011206242 (34 digits). The full barcode
+  // is NOT the tracking number — return the trailing tracking number. The
+  // routing portion must never pass as a shipment number.
+  function parseImpb(s) {
+    var num = normalize(s);
+    if (!/^420\d{23,31}$/.test(num)) return null; // 26–34 digits total
+    var m = /(9[1-5]\d{18,20})$/.exec(num);
+    return m ? m[1] : null;
+  }
+
+  // A spaced/dashed digit run is only plausible as a tracking number when
+  // every group has 2+ digits — this rejects ZIP+4 glued to a nearby lone
+  // digit ("07874-2956 1 lb"), phone fragments ("1-800-555-1234"), etc.
+  function validSpacedGroups(raw) {
+    var groups = String(raw).split(/[\s-]+/);
+    if (groups.length < 2) return true; // contiguous — decided elsewhere
+    for (var i = 0; i < groups.length; i++) if (groups[i].length < 2) return false;
+    return true;
+  }
+
+  // Digit groups split by spaces/dashes: "9305 5110 4390 0011 2062 42".
+  // Strict PATTERNS only match contiguous runs, so this pass exists for the
+  // standard label printing format. Only fires when the raw match actually
+  // contains a separator (contiguous runs are already covered by PATTERNS).
+  // Joined candidates must validate against a carrier format; IMpb barcodes
+  // yield their tracking number, never the routing.
+  function spacedDigitGroupsIn(text, found, seen, conf) {
+    var re = /\b\d(?:[\s-]?\d){9,33}\b/g, m;
+    while ((m = re.exec(text)) !== null) {
+      if (!/[\s-]/.test(m[0])) continue;
+      if (!validSpacedGroups(m[0])) continue;
+      var num = normalize(m[0]);
+      if (!num || seen[num]) continue;
+      var impb = parseImpb(num);
+      if (impb) {
+        seen[num] = true;
+        if (!seen[impb]) { seen[impb] = true; found.push({ number: impb, carrier: 'usps', confidence: conf }); }
+        continue;
+      }
+      var carrier = guessCarrier(num);
+      if (carrier === 'unknown') { seen[num] = true; continue; }
+      seen[num] = true;
+      found.push({ number: num, carrier: carrier, confidence: conf });
+    }
+  }
+
   function guessCarrier(num) {
     var carriers = ['ups', 'usps', 'fedex', 'dhl'];
     for (var i = 0; i < carriers.length; i++) if (isValidFormat(num, carriers[i])) return carriers[i];
@@ -349,13 +407,20 @@
     spacedUpsIn(text, found, seen, 'high');
     expandPunctHotZone(text, found, seen);
     looseUpsIn(text, found, seen, 'high');
-    // b) long digit-like runs with OCR confusions -> fix -> re-match digit formats
-    var runRe = /[0-9OILSBoilszb]{10,24}/g, run;
+    // b) long digit-like runs with OCR confusions -> fix -> re-match digit formats.
+    // Space-aware: OCR often reads "9305 51O0 ..." with both mangled chars
+    // AND spaces, so runs may contain separators. Runs up to 34 chars so a
+    // full 34-digit IMpb barcode is tested whole (parseImpb) instead of
+    // being split into chunks that false-positive as DHL/FedEx.
+    var runRe = /[0-9OILSBoilszb](?:[\s-]?[0-9OILSBoilszb]){9,33}/g, run;
     while ((run = runRe.exec(text)) !== null) {
+      if (/[\s-]/.test(run[0]) && !validSpacedGroups(run[0])) continue;
       var fixed = normalize(ocrDigitFix(run[0]));
-      if (/^9[2-5]\d{18,20}$/.test(fixed)) push(fixed, 'usps', 'medium');
+      var impbFixed = parseImpb(fixed);
+      if (impbFixed) { push(impbFixed, 'usps', 'medium'); continue; }
+      if (/^9[1-5]\d{18,20}$/.test(fixed)) push(fixed, 'usps', 'medium');
       else if (/^96\d{20}$/.test(fixed)) push(fixed, 'fedex', 'medium');
-      else if (/^(\d{12}|\d{15}|\d{20})$/.test(fixed)) push(fixed, 'fedex', 'medium');
+      else if (/^(\d{12}|\d{15}|\d{20})$/.test(fixed) && !/^420\d{9}$/.test(fixed)) push(fixed, 'fedex', 'medium');
       else if (/^\d{11}$/.test(fixed)) push(fixed, 'dhl', 'medium');
       else if (/^\d{10}$/.test(fixed)) push(fixed, 'dhl', 'low');
     }
@@ -369,6 +434,7 @@
     });
     tokens.sort(function (a, b) { return b.length - a.length; });
     tokens.slice(0, 3).forEach(function (t) {
+      if (/^420\d{9}$/.test(t)) return; // IMpb routing — never a shipment
       // UPS "1Z" misread as "12"/"IZ" in a joined (space-free) token
       var fixed = repairUpsZ(t);
       if (fixed) { push(fixed, 'ups', 'high'); return; }
@@ -393,7 +459,7 @@
     });
   }
 
-  var api = { extractTrackingNumbers: extractTrackingNumbers, extractContext: extractContext, normalize: normalize, isValidFormat: isValidFormat };
+  var api = { extractTrackingNumbers: extractTrackingNumbers, extractContext: extractContext, normalize: normalize, isValidFormat: isValidFormat, parseImpb: parseImpb };
   if (typeof module !== 'undefined' && module.exports) module.exports = api;
   globalThis.HHExtract = api;
 })();