Home | History | Annotate | Download | only in omnibox
      1 // Copyright 2014 The Chromium Authors. All rights reserved.
      2 // Use of this source code is governed by a BSD-style license that can be
      3 // found in the LICENSE file.
      4 
      5 #ifndef COMPONENTS_OMNIBOX_AUTOCOMPLETE_PROVIDER_H_
      6 #define COMPONENTS_OMNIBOX_AUTOCOMPLETE_PROVIDER_H_
      7 
      8 #include "base/basictypes.h"
      9 #include "base/memory/ref_counted.h"
     10 #include "base/strings/string16.h"
     11 #include "components/metrics/proto/omnibox_event.pb.h"
     12 #include "components/omnibox/autocomplete_match.h"
     13 
     14 class AutocompleteInput;
     15 
     16 typedef std::vector<metrics::OmniboxEventProto_ProviderInfo> ProvidersInfo;
     17 
     18 // The AutocompleteProviders each return different kinds of matches,
     19 // such as history or search matches.  These matches are given
     20 // "relevance" scores.  Higher scores are better matches than lower
     21 // scores.  The relevance scores and classes providing the respective
     22 // matches are as listed below.
     23 //
     24 // IMPORTANT CAVEAT: The tables below are NOT COMPLETE.  Developers
     25 // often forget to keep these tables in sync with the code when they
     26 // change scoring algorithms or add new providers.  For example,
     27 // neither the HistoryQuickProvider (which is a provider that appears
     28 // often) nor the ShortcutsProvider are listed here.  For the best
     29 // idea of how scoring works and what providers are affecting which
     30 // queries, play with chrome://omnibox/ for a while.  While the tables
     31 // below may have some utility, nothing compares with first-hand
     32 // investigation and experience.
     33 //
     34 // UNKNOWN input type:
     35 // --------------------------------------------------------------------|-----
     36 // Keyword (non-substituting or in keyword UI mode, exact match)       | 1500
     37 // HistoryURL (good exact or inline autocomplete matches, some inexact)| 1410++
     38 // HistoryURL (intranet url never visited match, some inexact matches) | 1400++
     39 // Search Primary Provider (past query in history within 2 days)       | 1399**
     40 // Search Primary Provider (what you typed)                            | 1300
     41 // HistoryURL (what you typed, some inexact matches)                   | 1200++
     42 // Keyword (substituting, exact match)                                 | 1100
     43 // Search Primary Provider (past query in history older than 2 days)   | 1050--
     44 // HistoryURL (some inexact matches)                                   |  900++
     45 // BookmarkProvider (prefix match in bookmark title)                   |  900+-
     46 // Built-in                                                            |  860++
     47 // Search Primary Provider (navigational suggestion)                   |  800++
     48 // Search Primary Provider (suggestion)                                |  600++
     49 // Keyword (inexact match)                                             |  450
     50 // Search Secondary Provider (what you typed)                          |  250
     51 // Search Secondary Provider (past query in history)                   |  200--
     52 // Search Secondary Provider (navigational suggestion)                 |  150++
     53 // Search Secondary Provider (suggestion)                              |  100++
     54 //
     55 // URL input type:
     56 // --------------------------------------------------------------------|-----
     57 // Keyword (non-substituting or in keyword UI mode, exact match)       | 1500
     58 // HistoryURL (good exact or inline autocomplete matches, some inexact)| 1410++
     59 // HistoryURL (intranet url never visited match, some inexact matches) | 1400++
     60 // HistoryURL (what you typed, some inexact matches)                   | 1200++
     61 // Keyword (substituting, exact match)                                 | 1100
     62 // HistoryURL (some inexact matches)                                   |  900++
     63 // Built-in                                                            |  860++
     64 // Search Primary Provider (what you typed)                            |  850
     65 // Search Primary Provider (navigational suggestion)                   |  800++
     66 // Search Primary Provider (past query in history)                     |  750--
     67 // Keyword (inexact match)                                             |  700
     68 // Search Primary Provider (suggestion)                                |  300++
     69 // Search Secondary Provider (what you typed)                          |  250
     70 // Search Secondary Provider (past query in history)                   |  200--
     71 // Search Secondary Provider (navigational suggestion)                 |  150++
     72 // Search Secondary Provider (suggestion)                              |  100++
     73 //
     74 // QUERY input type:
     75 // --------------------------------------------------------------------|-----
     76 // Search Primary or Secondary (past query in history within 2 days)   | 1599**
     77 // Keyword (non-substituting or in keyword UI mode, exact match)       | 1500
     78 // Keyword (substituting, exact match)                                 | 1450
     79 // Search Primary Provider (past query in history within 2 days)       | 1399**
     80 // Search Primary Provider (what you typed)                            | 1300
     81 // Search Primary Provider (past query in history older than 2 days)   | 1050--
     82 // HistoryURL (inexact match)                                          |  900++
     83 // BookmarkProvider (prefix match in bookmark title)                   |  900+-
     84 // Search Primary Provider (navigational suggestion)                   |  800++
     85 // Search Primary Provider (suggestion)                                |  600++
     86 // Keyword (inexact match)                                             |  450
     87 // Search Secondary Provider (what you typed)                          |  250
     88 // Search Secondary Provider (past query in history)                   |  200--
     89 // Search Secondary Provider (navigational suggestion)                 |  150++
     90 // Search Secondary Provider (suggestion)                              |  100++
     91 //
     92 // FORCED_QUERY input type:
     93 // --------------------------------------------------------------------|-----
     94 // Search Primary Provider (past query in history within 2 days)       | 1399**
     95 // Search Primary Provider (what you typed)                            | 1300
     96 // Search Primary Provider (past query in history older than 2 days)   | 1050--
     97 // Search Primary Provider (navigational suggestion)                   |  800++
     98 // Search Primary Provider (suggestion)                                |  600++
     99 //
    100 // (A search keyword is a keyword with a replacement string; a bookmark keyword
    101 // is a keyword with no replacement string, that is, a shortcut for a URL.)
    102 //
    103 // There are two possible providers for search suggestions. If the user has
    104 // typed a keyword, then the primary provider is the keyword provider and the
    105 // secondary provider is the default provider. If the user has not typed a
    106 // keyword, then the primary provider corresponds to the default provider.
    107 //
    108 // Search providers may supply relevance values along with their results to be
    109 // used in place of client-side calculated values.
    110 //
    111 // The value column gives the ranking returned from the various providers.
    112 // ++: a series of matches with relevance from n up to (n + max_matches).
    113 // --: relevance score falls off over time (discounted 50 points @ 15 minutes,
    114 //     450 points @ two weeks)
    115 // **: relevance score falls off over two days (discounted 99 points after two
    116 //     days).
    117 // +-: A base score that the provider will adjust upward or downward based on
    118 //     provider-specific metrics.
    119 //
    120 // A single result provider for the autocomplete system.  Given user input, the
    121 // provider decides what (if any) matches to return, their relevance, and their
    122 // classifications.
    123 class AutocompleteProvider
    124     : public base::RefCountedThreadSafe<AutocompleteProvider> {
    125  public:
    126   // Different AutocompleteProvider implementations.
    127   enum Type {
    128     TYPE_BOOKMARK         = 1 << 0,
    129     TYPE_BUILTIN          = 1 << 1,
    130     TYPE_HISTORY_QUICK    = 1 << 2,
    131     TYPE_HISTORY_URL      = 1 << 3,
    132     TYPE_KEYWORD          = 1 << 4,
    133     TYPE_SEARCH           = 1 << 5,
    134     TYPE_SHORTCUTS        = 1 << 6,
    135     TYPE_ZERO_SUGGEST     = 1 << 7,
    136   };
    137 
    138   explicit AutocompleteProvider(Type type);
    139 
    140   // Returns a string describing a particular AutocompleteProvider type.
    141   static const char* TypeToString(Type type);
    142 
    143   // Called to start an autocomplete query.  The provider is responsible for
    144   // tracking its matches for this query and whether it is done processing the
    145   // query.  When new matches are available or the provider finishes, it
    146   // calls the controller's OnProviderUpdate() method.  The controller can then
    147   // get the new matches using the provider's accessors.
    148   // Exception: Matches available immediately after starting the query (that
    149   // is, synchronously) do not cause any notifications to be sent.  The
    150   // controller is expected to check for these without prompting (since
    151   // otherwise, starting each provider running would result in a flurry of
    152   // notifications).
    153   //
    154   // Once Stop() has been called, no more notifications should be sent.
    155   //
    156   // |minimal_changes| is an optimization that lets the provider do less work
    157   // when the |input|'s text hasn't changed.  See the body of
    158   // OmniboxPopupModel::StartAutocomplete().
    159   virtual void Start(const AutocompleteInput& input, bool minimal_changes) = 0;
    160 
    161   // Called when a provider must not make any more callbacks for the current
    162   // query. This will be called regardless of whether the provider is already
    163   // done.  If the provider caches any results, it should clear the cache based
    164   // on the value of |clear_cached_results|.
    165   virtual void Stop(bool clear_cached_results);
    166 
    167   // Returns the enum equivalent to the name of this provider.
    168   // TODO(derat): Make metrics use AutocompleteProvider::Type directly, or at
    169   // least move this method to the metrics directory.
    170   metrics::OmniboxEventProto_ProviderType AsOmniboxEventProviderType() const;
    171 
    172   // Called to delete a match and the backing data that produced it.  This
    173   // match should not appear again in this or future queries.  This can only be
    174   // called for matches the provider marks as deletable.  This should only be
    175   // called when no query is running.
    176   // NOTE: Do NOT call OnProviderUpdate() in this method, it is the
    177   // responsibility of the caller to do so after calling us.
    178   virtual void DeleteMatch(const AutocompleteMatch& match);
    179 
    180   // Called when an omnibox event log entry is generated.  This gives
    181   // a provider the opportunity to add diagnostic information to the
    182   // logs.  A provider is expected to append a single entry of whatever
    183   // information it wants to |provider_info|.
    184   virtual void AddProviderInfo(ProvidersInfo* provider_info) const;
    185 
    186   // Called when a new omnibox session starts or the current session ends.
    187   // This gives the opportunity to reset the internal state, if any, associated
    188   // with the previous session.
    189   virtual void ResetSession();
    190 
    191   // Returns the set of matches for the current query.
    192   const ACMatches& matches() const { return matches_; }
    193 
    194   // Returns whether the provider is done processing the query.
    195   bool done() const { return done_; }
    196 
    197   // Returns this provider's type.
    198   Type type() const { return type_; }
    199 
    200   // Returns a string describing this provider's type.
    201   const char* GetName() const;
    202 
    203   // A suggested upper bound for how many matches a provider should return.
    204   // TODO(pkasting): http://b/1111299 , http://b/933133 This should go away once
    205   // we have good relevance heuristics; the controller should handle all
    206   // culling.
    207   static const size_t kMaxMatches;
    208 
    209  protected:
    210   friend class base::RefCountedThreadSafe<AutocompleteProvider>;
    211   FRIEND_TEST_ALL_PREFIXES(BookmarkProviderTest, InlineAutocompletion);
    212 
    213   typedef std::pair<bool, base::string16> FixupReturn;
    214 
    215   virtual ~AutocompleteProvider();
    216 
    217   // Fixes up user URL input to make it more possible to match against.  Among
    218   // many other things, this takes care of the following:
    219   // * Prepending file:// to file URLs
    220   // * Converting drive letters in file URLs to uppercase
    221   // * Converting case-insensitive parts of URLs (like the scheme and domain)
    222   //   to lowercase
    223   // * Convert spaces to %20s
    224   // Note that we don't do this in AutocompleteInput's constructor, because if
    225   // e.g. we convert a Unicode hostname to punycode, other providers will show
    226   // output that surprises the user ("Search Google for xn--6ca.com").
    227   // Returns a bool indicating whether fixup succeeded, as well as the fixed-up
    228   // input text.  The returned string will be the same as the input string if
    229   // fixup failed; this lets callers who don't care about failure simply use the
    230   // string unconditionally.
    231   static FixupReturn FixupUserInput(const AutocompleteInput& input);
    232 
    233   // Trims "http:" and up to two subsequent slashes from |url|.  Returns the
    234   // number of characters that were trimmed.
    235   // NOTE: For a view-source: URL, this will trim from after "view-source:" and
    236   // return 0.
    237   static size_t TrimHttpPrefix(base::string16* url);
    238 
    239   ACMatches matches_;
    240   bool done_;
    241 
    242   Type type_;
    243 
    244  private:
    245   DISALLOW_COPY_AND_ASSIGN(AutocompleteProvider);
    246 };
    247 
    248 #endif  // COMPONENTS_OMNIBOX_AUTOCOMPLETE_PROVIDER_H_
    249