Correctly extracting email addresses

core/src/fpdftext/fpdf_text_int.cpp

 // Copyright 2014 PDFium Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 // Original code copyright 2014 Foxit Software Inc. http://www.foxitsoftware.com
 
 #include <cctype>
 #include <cwctype>
 #include <algorithm>
 
@@ skipping 2592 matching lines @@
   if (str.Find(L"www.") != -1) {
     strBeCheck = strBeCheck.Right(str.GetLength() - str.Find(L"www."));
     strBeCheck = L"http://" + strBeCheck;
     return TRUE;
   }
   return FALSE;
 }
 FX_BOOL CPDF_LinkExtract::CheckMailLink(CFX_WideString& str) {
   str.MakeLower();
   int aPos = str.Find(L'@');
+  // Invalid when no '@'.
   if (aPos < 1) {
     return FALSE;
   }
-  if (str.GetAt(aPos - 1) == L'.' || str.GetAt(aPos - 1) == L'_') {
-    return FALSE;
-  }
   int i;

[jun_fang, 2015/12/17 13:30:45]

nit: prefer "for (int i = aPos - 1; i >= 0; i--)"

[Wei Li, 2015/12/17 17:41:41]

Done.

I was not sure what style I should follow when I changed the original code. But
glad to change this one. :)

+  int pPos = aPos;  // Used to track the position of '@' or '.'.
   for (i = aPos - 1; i >= 0; i--) {
     FX_WCHAR ch = str.GetAt(i);
-    if (ch == L'_' || ch == L'.' || (ch >= L'a' && ch <= L'z') ||
+    if (ch == L'_' || ch == L'-' || (ch >= L'a' && ch <= L'z') ||

[jun_fang, 2015/12/17 13:30:45]

Should we handle upper-case letters in email address?

[Wei Li, 2015/12/17 17:41:41]

Done.

I don't know why the original code chose to be case insensitive. Since it seems
no harm, I kept it. But keeping the case is better.

         (ch >= L'0' && ch <= L'9')) {
       continue;
-    } else {
+    }
+    if (ch != L'.' || i == pPos - 1 || i == 0) {
       if (i == aPos - 1) {
+        // There is '.' or invalid char before '@'.
         return FALSE;
       }
-      str = str.Right(str.GetLength() - i - 1);
+      // End extracting for other invalid chars, '.' at the beginning, or
+      // consecutive '.'.
+      int removed_len = i == pPos - 1 ? i + 2 : i + 1;
+      str = str.Right(str.GetLength() - removed_len);
       break;
+    } else {
+      // Found a valid '.'.
+      pPos = i;
     }
   }
-  aPos = str.Find(L'@');
-  if (aPos < 1) {
-    return FALSE;
-  }
-  CFX_WideString strtemp = L"";
-  for (i = 0; i < aPos; i++) {
-    FX_WCHAR wch = str.GetAt(i);
-    if (wch >= L'a' && wch <= L'z') {
-      break;
-    } else {
-      strtemp = str.Right(str.GetLength() - i + 1);
-    }
-  }
-  if (strtemp != L"") {
-    str = strtemp;
-  }
+
+  // Check the host name part.
   aPos = str.Find(L'@');
   if (aPos < 1) {
     return FALSE;
   }
   str.TrimRight(L'.');
-  strtemp = str;
-  int ePos = str.Find(L'.');
-  if (ePos == -1) {
+  CFX_WideString strtemp = str;
+  // At least one '.' in host name, but not at the beginning.
+  int ePos = str.Find(L'.', aPos + 1);
+  if (ePos == -1 || ePos == aPos + 1) {
     return FALSE;
   }
-  while (ePos != -1) {
-    strtemp = strtemp.Right(strtemp.GetLength() - ePos - 1);
-    ePos = strtemp.Find('.');
-  }
-  ePos = strtemp.GetLength();
-  for (i = 0; i < ePos; i++) {
+  // Validate all other chars in host name.
+  int nLen = str.GetLength();
+  pPos = 0;  // Used to track the position of '.'.
+  for (i = aPos + 1; i < nLen; i++) {

[jun_fang, 2015/12/17 13:30:45]

nit: for (int i = aPos + 1; i < nLen; i++).

[Wei Li, 2015/12/17 17:41:41]

Done.

     FX_WCHAR wch = str.GetAt(i);
-    if ((wch >= L'a' && wch <= L'z') || (wch >= L'0' && wch <= L'9')) {
+    if (wch == L'-' || (wch >= L'a' && wch <= L'z') ||

[jun_fang, 2015/12/17 13:30:45]

How about upper-case letters?

[Wei Li, 2015/12/17 17:41:41]

See above.

+        (wch >= L'0' && wch <= L'9')) {
       continue;
+    }
+    if (wch != L'.' || i == pPos + 1) {
+      // Host name should end before invalid char.
+      int host_end = i == pPos + 1 ? i - 2 : i - 1;
+      if (pPos > 0 && host_end - aPos >= 3) {
+        // Trim the ending invalid chars if there is at least one '.' and name.
+        str = str.Left(host_end + 1);
+        break;
+      }
+      return FALSE;
     } else {
-      str = str.Left(str.GetLength() - ePos + i + 1);
-      ePos = ePos - i - 1;
-      break;
+      pPos = i;
     }
   }
-  int nLen = str.GetLength();
-  for (i = aPos + 1; i < nLen - ePos; i++) {
-    FX_WCHAR wch = str.GetAt(i);
-    if (wch == L'-' || wch == L'.' || (wch >= L'a' && wch <= L'z') ||
-        (wch >= L'0' && wch <= L'9')) {
-      continue;
-    } else {
-      return FALSE;
-    }
-  }
+
   if (str.Find(L"mailto:") == -1) {
     str = L"mailto:" + str;
   }
   return TRUE;
 }
 
 void CPDF_LinkExtract::AppendToLinkList(int start,
                                         int count,
                                         const CFX_WideString& strUrl) {
   CPDF_LinkExt* linkInfo = new CPDF_LinkExt;
@@ skipping 32 matching lines @@
   if (!m_bIsParsed || index < 0 || index >= m_LinkList.GetSize()) {
     return;
   }
   CPDF_LinkExt* link = NULL;
   link = m_LinkList.GetAt(index);
   if (!link) {
     return;
   }
   m_pTextPage->GetRectArray(link->m_Start, link->m_Count, rects);
 }