autofill: be more open minded about filling in states in <option>s.

components/autofill/core/browser/autofill_field.cc

Index: components/autofill/core/browser/autofill_field.cc
diff --git a/components/autofill/core/browser/autofill_field.cc b/components/autofill/core/browser/autofill_field.cc
index bad3b5b680c23c43149930ab8782e3cc08793dd6..007fb7add8ddacb26c5524aaa20c1ad1b1e0f563 100644
--- a/components/autofill/core/browser/autofill_field.cc
+++ b/components/autofill/core/browser/autofill_field.cc
@@ -66,6 +66,55 @@ bool SetSelectControlValue(const base::string16& value,
   return true;
 }
 
+// Like SetSelectControlValue, but searches within the field values and options
+// for |value|. For example, "NC - North Carolina" would match "north carolina".

[Ilya Sherman, 2014/07/01 00:59:08]

Would "WV - West Virginia" match "virginia"?  It seems like it would, and it
seems like that's problematic...

[Evan Stade, 2014/07/01 01:04:56]

hmm, yes I suppose it would. Perhaps I could make it choose the shortest option
that has a substring match, so it would prefer "VA - Virginia" over "WV - West
Virginia" if both were present. If only "WV - West Virginia" is present, it
would fill it incorrectly, but that seems like a rare case.

[Ilya Sherman, 2014/07/01 01:17:47]

Yeah, maybe.  Alternately, we could just kinda assume that lists will be
alphabetically sorted, and hence things will Just Work(TM).  But it's at least
worth documenting this potential failure case.  (The only other pair with this
property that I can think of is "Washington" and "Washington, D.C.".  Though,
our canonical list of state names lists DC as "district of columbia", so in
practice that pair shouldn't tickle this particular issue.)

+bool SetSelectControlValueSubstringMatch(const base::string16& value,
+                                         FormFieldData* field) {
+  base::string16 value_lowercase = StringToLowerASCII(value);
+  DCHECK_EQ(field->option_values.size(), field->option_contents.size());
+
+  for (size_t i = 0; i < field->option_values.size(); ++i) {
+    if (StringToLowerASCII(field->option_values[i]).find(value_lowercase) !=
+            std::string::npos ||
+        StringToLowerASCII(field->option_contents[i]).find(value_lowercase) !=
+            std::string::npos) {
+      field->value = field->option_values[i];
+      return true;
+    }
+  }
+
+  return false;
+}
+
+// Like SetSelectControlValue, but searches within the field values and options
+// for |value|. First it tokenizes the options, then tries to match against
+// tokens. For example, "NC - North Carolina" would match "nc" but not "ca".
+bool SetSelectControlValueTokenMatch(const base::string16& value,
+                                     FormFieldData* field) {
+  base::string16 value_lowercase = StringToLowerASCII(value);
+  std::vector<base::string16> tokenized;
+  DCHECK_EQ(field->option_values.size(), field->option_contents.size());
+
+  for (size_t i = 0; i < field->option_values.size(); ++i) {
+    SplitStringAlongWhitespace(StringToLowerASCII(field->option_values[i]),
+                               &tokenized);
+    if (std::find(tokenized.begin(), tokenized.end(), value_lowercase) !=
+        tokenized.end()) {
+      field->value = field->option_values[i];
+      return true;
+    }
+
+    SplitStringAlongWhitespace(StringToLowerASCII(field->option_contents[i]),
+                               &tokenized);
+    if (std::find(tokenized.begin(), tokenized.end(), value_lowercase) !=
+        tokenized.end()) {
+      field->value = field->option_values[i];
+      return true;
+    }
+  }
+
+  return false;
+}
 
 // Try to fill a numeric |value| into the given |field|.
 bool FillNumericSelectControl(int value,
@@ -88,11 +137,24 @@ bool FillStateSelectControl(const base::string16& value,
   base::string16 full, abbreviation;
   state_names::GetNameAndAbbreviation(value, &full, &abbreviation);
 
-  // Try the abbreviation first.
-  if (!abbreviation.empty() && SetSelectControlValue(abbreviation, field))
+  // Try an exact match of the abbreviation first.
+  if (!abbreviation.empty() && SetSelectControlValue(abbreviation, field)) {
+    return true;
+  }

[Dan Beam, 2014/07/01 00:34:32]

why so curly?

+
+  // Try an exact match of the full name.
+  if (!full.empty() && SetSelectControlValue(full, field)) {
     return true;
+  }
+
+  // Then try an inexact match of the full name.
+  if (!full.empty() && SetSelectControlValueSubstringMatch(full, field)) {
+    return true;
+  }
 
-  return !full.empty() && SetSelectControlValue(full, field);
+  // Then try an inexact match of the abbreviation name.
+  return !abbreviation.empty() &&
+      SetSelectControlValueTokenMatch(abbreviation, field);
 }
 
 bool FillCountrySelectControl(const base::string16& value,