Omnibox: Make Keyword Provide More Generous with Matching

components/omnibox/browser/omnibox_field_trial.h

Index: components/omnibox/browser/omnibox_field_trial.h
diff --git a/components/omnibox/browser/omnibox_field_trial.h b/components/omnibox/browser/omnibox_field_trial.h
index b6e354993d5d8c59383ce63ebb299987de581cb9..8e94ea2b2739ab4eea291382bd8517d652237053 100644
--- a/components/omnibox/browser/omnibox_field_trial.h
+++ b/components/omnibox/browser/omnibox_field_trial.h
@@ -320,6 +320,29 @@ class OmniboxFieldTrial {
   static bool PreventUWYTDefaultForNonURLInputs();
 
   // ---------------------------------------------------------
+  // For the aggressive keyword matching experiment that's part of the bundled
+  // omnibox field trial.
+
+  // Return true if KeywordProvider (via TemplateURLService), for auto-

[Peter Kasting, 2015/11/09 23:11:14]

Nit: Return -> Returns

[Mark P, 2015/11/11 07:41:00]

Done.

+  // generated keywords, should consider the registry (e.g., co.uk) an
+  // important part of the keyword name for matching purposes.  Returns true

[Peter Kasting, 2015/11/09 23:11:15]

Nit: Awkward, how about:

Returns whether the KeywordProvider should require users to type any effective
TLD portion ("co.uk") of autogenerated keywords to match against them.

Then name the function KeywordRequiresTld().

[Mark P, 2015/11/11 07:41:00]

This isn't correct.  It can match against them even if this is false.  It's more
about scoring and whether this is considered meaningful.  Revises somewhat
anyway.
I'm sticking with registry for consistency with the underlying net functions.

+  // if the experiment isn't active.
+  static bool KeywordRequiresRegistry();
+
+  // Return true if KeywordProvider (via TemplateURLService), for auto-

[Peter Kasting, 2015/11/09 23:11:14]

Nit: Return -> Returns

[Mark P, 2015/11/11 07:41:00]

Done.

+  // generated keywords, should return a keyword if the user starts typing
+  // the domain name portion, skipping any host before the domain name.
+  // In other words, returns whether the prefix before the domain name should
+  // be considered important for matching purposes.  Returns true if the

[Peter Kasting, 2015/11/09 23:11:14]

Nit: Awkward, how about:

For autogenerated keywords containing more than just a TLD+1, returns whether
the KeywordProvider should require users to type the full hostname to match
against them, rather than just the TLD+1 portion.

Then name the function KeywordRequiresFullHostname().

[Mark P, 2015/11/11 07:41:00]

Revised somewhat in the direction you suggested.
That's not true.  Prefixes are okay.  Instead revised to 
KeywordRequiresPrefixMatch.

+  // experiment isn't active.
+  static bool KeywordRequiresPrefixBeforeDomain();
+
+  // For the relevance score that KeywordProvider should assign to keywords

[Peter Kasting, 2015/11/09 23:11:15]

Nit: For -> Returns

[Mark P, 2015/11/11 07:41:00]

Done.

+  // with nearly-complete matches, i.e., the user has typed all of the
+  // important part of the keyword.  Returns -1 if the experiment isn't active.
+  static int KeywordScoreForNearlyCompleteMatch();
+
+  // ---------------------------------------------------------
   // Exposed publicly for the sake of unittests.
   static const char kBundledExperimentFieldTrialName[];
   // Rule names used by the bundled experiment.
@@ -341,6 +364,9 @@ class OmniboxFieldTrial {
   static const char kHQPNumTitleWordsRule[];
   static const char kHQPAlsoDoHUPLikeScoringRule[];
   static const char kPreventUWYTDefaultForNonURLInputsRule[];
+  static const char kKeywordRequiresRegistryRule[];
+  static const char kKeywordRequiresPrefixBeforeDomainRule[];
+  static const char kKeywordScoreForNearlyCompleteMatchRule[];
 
   // Parameter names used by the HUP new scoring experiments.
   static const char kHUPNewScoringEnabledParam[];