hh-shipments 0.3.4 → 0.3.5

package/package.json

@@ -18,7 +18,7 @@
     "test": "node test/run-tests.js && node test/run-carriers-tests.js"
   },
   "type": "commonjs",
-  "version": "0.3.4",
+  "version": "0.3.5",
   "bin": {
     "hh-shipments": "server.js"
   },

package/public/extract.js

@@ -111,6 +111,9 @@
   }
 
   function extractOcrTolerant(text, found, seen) {
+    // Strict UPS structure: 1Z + 6-char shipper number + 2-digit service + 8-digit serial.
+    // Used to validate "1Z"-as-"12" OCR repairs so we don't invent false positives.
+    var UPS_STRICT = /^1Z[0-9A-Z]{6}[0-9]{2}[0-9]{8}$/;
     function push(num, carrier, confidence, extra) {
       num = normalize(num);
       if (!num || seen[num]) return;
@@ -119,11 +122,27 @@
       if (extra) for (var k in extra) c[k] = extra[k];
       found.push(c);
     }
-    // a) UPS number split by spaces/dashes: "1Z 9999 99 ..."
-    var m, spacedUps = /\b1Z[\s-]?(?:[0-9A-Z][\s-]?){16}\b/gi;
+    function repairUpsZ(ups) {
+      // Tesseract often reads the "1Z" anchor as "12" (or "IZ"). Repair it
+      // only when the rest has strict UPS structure AND the shipper segment
+      // contains a letter — real UPS shipper numbers do, random digit runs
+      // don't. Never invent numbers from thin air.
+      var cand = null;
+      if (/^12[0-9A-Z]{16}$/.test(ups)) cand = '1Z' + ups.slice(2);
+      else if (/^IZ[0-9A-Z]{16}$/i.test(ups)) cand = '1Z' + ups.slice(2);
+      if (!cand || !UPS_STRICT.test(cand)) return null;
+      if (!/[A-Z]/.test(cand.slice(2, 8))) return null;
+      return cand;
+    }
+    // a) UPS number split by spaces/dashes: "1Z 9999 99 ..." (or "12 ..."/"IZ ..." misreads)
+    var m, spacedUps = /\b(?:1Z|12|IZ)[\s-]?(?:[0-9A-Z][\s-]?){16}\b/gi;
     while ((m = spacedUps.exec(text)) !== null) {
       var ups = normalize(m[0]);
       if (/^1Z[0-9A-Z]{16}$/.test(ups)) push(ups, 'ups', 'medium');
+      else {
+        var fixed = repairUpsZ(ups);
+        if (fixed) push(fixed, 'ups', 'medium');
+      }
     }
     // b) long digit-like runs with OCR confusions -> fix -> re-match digit formats
     var runRe = /[0-9OILSBoilszb]{10,24}/g, run;
@@ -145,6 +164,13 @@
     });
     tokens.sort(function (a, b) { return b.length - a.length; });
     tokens.slice(0, 3).forEach(function (t) {
+      // UPS "1Z" misread as "12"/"IZ" in a joined (space-free) token
+      var fixed = repairUpsZ(t);
+      if (fixed) { push(fixed, 'ups', 'medium'); return; }
+      // OCR-fixed variant may match a carrier format (e.g. I->1 at the start)
+      var digitFixed = ocrDigitFix(t);
+      var gc = guessCarrier(digitFixed);
+      if (gc !== 'unknown') { push(digitFixed, gc, 'low', { possible: true }); return; }
       push(t, guessCarrier(t), 'low', { possible: true });
     });
   }

package/src/extract.js

@@ -111,6 +111,9 @@
   }
 
   function extractOcrTolerant(text, found, seen) {
+    // Strict UPS structure: 1Z + 6-char shipper number + 2-digit service + 8-digit serial.
+    // Used to validate "1Z"-as-"12" OCR repairs so we don't invent false positives.
+    var UPS_STRICT = /^1Z[0-9A-Z]{6}[0-9]{2}[0-9]{8}$/;
     function push(num, carrier, confidence, extra) {
       num = normalize(num);
       if (!num || seen[num]) return;
@@ -119,11 +122,27 @@
       if (extra) for (var k in extra) c[k] = extra[k];
       found.push(c);
     }
-    // a) UPS number split by spaces/dashes: "1Z 9999 99 ..."
-    var m, spacedUps = /\b1Z[\s-]?(?:[0-9A-Z][\s-]?){16}\b/gi;
+    function repairUpsZ(ups) {
+      // Tesseract often reads the "1Z" anchor as "12" (or "IZ"). Repair it
+      // only when the rest has strict UPS structure AND the shipper segment
+      // contains a letter — real UPS shipper numbers do, random digit runs
+      // don't. Never invent numbers from thin air.
+      var cand = null;
+      if (/^12[0-9A-Z]{16}$/.test(ups)) cand = '1Z' + ups.slice(2);
+      else if (/^IZ[0-9A-Z]{16}$/i.test(ups)) cand = '1Z' + ups.slice(2);
+      if (!cand || !UPS_STRICT.test(cand)) return null;
+      if (!/[A-Z]/.test(cand.slice(2, 8))) return null;
+      return cand;
+    }
+    // a) UPS number split by spaces/dashes: "1Z 9999 99 ..." (or "12 ..."/"IZ ..." misreads)
+    var m, spacedUps = /\b(?:1Z|12|IZ)[\s-]?(?:[0-9A-Z][\s-]?){16}\b/gi;
     while ((m = spacedUps.exec(text)) !== null) {
       var ups = normalize(m[0]);
       if (/^1Z[0-9A-Z]{16}$/.test(ups)) push(ups, 'ups', 'medium');
+      else {
+        var fixed = repairUpsZ(ups);
+        if (fixed) push(fixed, 'ups', 'medium');
+      }
     }
     // b) long digit-like runs with OCR confusions -> fix -> re-match digit formats
     var runRe = /[0-9OILSBoilszb]{10,24}/g, run;
@@ -145,6 +164,13 @@
     });
     tokens.sort(function (a, b) { return b.length - a.length; });
     tokens.slice(0, 3).forEach(function (t) {
+      // UPS "1Z" misread as "12"/"IZ" in a joined (space-free) token
+      var fixed = repairUpsZ(t);
+      if (fixed) { push(fixed, 'ups', 'medium'); return; }
+      // OCR-fixed variant may match a carrier format (e.g. I->1 at the start)
+      var digitFixed = ocrDigitFix(t);
+      var gc = guessCarrier(digitFixed);
+      if (gc !== 'unknown') { push(digitFixed, gc, 'low', { possible: true }); return; }
       push(t, guessCarrier(t), 'low', { possible: true });
     });
   }