Merge pull request #1294 from maxmind/horgh/normalize

Replace fewer TLDs when normalizing

Author: ugexe

CHANGELOG.md

@@ -20,11 +20,8 @@ CHANGELOG
 * Duplicate `.com`s are now removed from email domain names when
   `hashAddress` is used. For example, `example.com.com` will become
   `example.com`.
-* Extraneous characters after `.com` are now removed from email domain
-  names when `hashAddress` is used. For example, `example.comfoo` will
-  become `example.com`.
-* Certain `.com` typos are now normalized to `.com` when `hashAddress` is
-  used. For example, `example.cam` will become `example.com`.
+* Certain TLD typos are now normalized when `hashAddress` is used. For
+  example, `example.comcom` will become `example.com`.
 * Additional `gmail.com` domain names with leading digits are now
   normalized when `hashAddress` is used. For example, `100gmail.com` will
   become `gmail.com`.

src/request/email.spec.ts

@@ -180,19 +180,24 @@ describe('Email()', () => {
     },
     {
       email: '[email protected]',
-      md5: md5('foo@example.com'),
+      md5: md5('foo@example.comfoo'),
       domain: 'example.comfoo',
     },
     {
       email: '[email protected]',
-      md5: md5('foo@example.com'),
+      md5: md5('foo@example.cam'),
       domain: 'example.cam',
     },
     {
       email: '[email protected]',
       md5: md5('[email protected]'),
       domain: '10000gmail.com',
     },
+    {
+      email: '[email protected]',
+      md5: md5('[email protected]'),
+      domain: 'example.comcom',
+    },
   ];
 
   test.each(normalizeTests)('%p', (arg) => {

src/request/email.ts

@@ -48,6 +48,54 @@ export default class Email implements EmailProps {
     'putlook.com': 'outlook.com',
   };
 
+  private static readonly typoTLDs: { [key: string]: string } = {
+    comm: 'com',
+    commm: 'com',
+    commmm: 'com',
+    comn: 'com',
+
+    cbm: 'com',
+    ccm: 'com',
+    cdm: 'com',
+    cem: 'com',
+    cfm: 'com',
+    cgm: 'com',
+    chm: 'com',
+    cim: 'com',
+    cjm: 'com',
+    ckm: 'com',
+    clm: 'com',
+    cmm: 'com',
+    cnm: 'com',
+    cpm: 'com',
+    cqm: 'com',
+    crm: 'com',
+    csm: 'com',
+    ctm: 'com',
+    cum: 'com',
+    cvm: 'com',
+    cwm: 'com',
+    cxm: 'com',
+    cym: 'com',
+    czm: 'com',
+
+    col: 'com',
+    con: 'com',
+
+    dom: 'com',
+    don: 'com',
+    som: 'com',
+    son: 'com',
+    vom: 'com',
+    von: 'com',
+    xom: 'com',
+    xon: 'com',
+
+    clam: 'com',
+    colm: 'com',
+    comcom: 'com',
+  };
+
   private static readonly equivalentDomains: { [key: string]: string } = {
     'googlemail.com': 'gmail.com',
     'pm.me': 'protonmail.com',
@@ -318,13 +366,16 @@ export default class Email implements EmailProps {
     domain = punycode.toASCII(domain);
 
     domain = domain.replace(/(?:\.com){2,}$/, '.com');
-    domain = domain.replace(/\.com[^.]+$/, '.com');
-    domain = domain.replace(
-      /(?:\.(?:com|c[a-z]{1,2}m|co[ln]|[dsvx]o[mn]|))$/,
-      '.com'
-    );
     domain = domain.replace(/^\d+(?:gmail?\.com)$/, 'gmail.com');
 
+    const idx = domain.lastIndexOf('.');
+    if (idx !== -1) {
+      const tld = domain.substring(idx + 1);
+      if (Object.prototype.hasOwnProperty.call(Email.typoTLDs, tld)) {
+        domain = domain.substring(0, idx) + '.' + Email.typoTLDs[tld];
+      }
+    }
+
     if (Object.prototype.hasOwnProperty.call(Email.typoDomains, domain)) {
       domain = Email.typoDomains[domain];
     }