diff options
author | mrossetti@chromium.org <mrossetti@chromium.org@0039d316-1c4b-4281-b951-d872f2087c98> | 2010-08-04 16:07:34 +0000 |
---|---|---|
committer | mrossetti@chromium.org <mrossetti@chromium.org@0039d316-1c4b-4281-b951-d872f2087c98> | 2010-08-04 16:07:34 +0000 |
commit | 7c6821db64d3eb49fe986212d621d11a066912ce (patch) | |
tree | faca56e0577c5424857b366a8b43991904c20d5b /chrome | |
parent | b3ccac85bee8a3d89c7b5ca3a64ee311f3f565c3 (diff) | |
download | chromium_src-7c6821db64d3eb49fe986212d621d11a066912ce.zip chromium_src-7c6821db64d3eb49fe986212d621d11a066912ce.tar.gz chromium_src-7c6821db64d3eb49fe986212d621d11a066912ce.tar.bz2 |
Add Framework for History Quick Provider
Incorporate 'quick' history provider index as an autocomplete provider in preparation for the replacement of the current history_url_provider. This step refactors the history provider code a bit, introduces the replacement class (history-quick_provider), and adds a unittest template. Note that the 'quick' provider will only support the fast, synchronous autocomplete pass. Once the quick provider has been completed the current history_url_provider will be replaced. In the meantime, the latter will remain in operation in order to provide the slower history results.
BUG=None
TEST=None
Review URL: http://codereview.chromium.org/3005050
git-svn-id: svn://svn.chromium.org/chrome/trunk/src@54908 0039d316-1c4b-4281-b951-d872f2087c98
Diffstat (limited to 'chrome')
-rw-r--r-- | chrome/browser/autocomplete/autocomplete.cc | 9 | ||||
-rw-r--r-- | chrome/browser/autocomplete/history_quick_provider.cc | 10 | ||||
-rw-r--r-- | chrome/browser/autocomplete/history_quick_provider.h | 39 | ||||
-rw-r--r-- | chrome/browser/autocomplete/history_quick_provider_unittest.cc | 35 | ||||
-rw-r--r-- | chrome/browser/autocomplete/history_url_provider.cc | 210 | ||||
-rw-r--r-- | chrome/browser/autocomplete/history_url_provider.h | 109 | ||||
-rw-r--r-- | chrome/browser/history/history.cc | 10 | ||||
-rw-r--r-- | chrome/browser/history/history.h | 3 | ||||
-rw-r--r-- | chrome/browser/history/history_backend.cc | 2 | ||||
-rw-r--r-- | chrome/browser/history/history_types.cc | 7 | ||||
-rw-r--r-- | chrome/browser/history/history_types.h | 79 | ||||
-rw-r--r-- | chrome/browser/history/in_memory_history_backend.cc | 11 | ||||
-rw-r--r-- | chrome/browser/history/in_memory_history_backend.h | 3 | ||||
-rw-r--r-- | chrome/browser/history/in_memory_url_index.cc | 10 | ||||
-rw-r--r-- | chrome/browser/history/in_memory_url_index.h | 9 | ||||
-rw-r--r-- | chrome/chrome_browser.gypi | 2 | ||||
-rw-r--r-- | chrome/chrome_tests.gypi | 1 |
17 files changed, 345 insertions, 204 deletions
diff --git a/chrome/browser/autocomplete/autocomplete.cc b/chrome/browser/autocomplete/autocomplete.cc index 0693ca2..66d7459 100644 --- a/chrome/browser/autocomplete/autocomplete.cc +++ b/chrome/browser/autocomplete/autocomplete.cc @@ -8,9 +8,11 @@ #include "app/l10n_util.h" #include "base/basictypes.h" +#include "base/command_line.h" #include "base/i18n/number_formatting.h" #include "base/string_number_conversions.h" #include "base/string_util.h" +#include "chrome/browser/autocomplete/history_quick_provider.h" #include "chrome/browser/autocomplete/history_url_provider.h" #include "chrome/browser/autocomplete/history_contents_provider.h" #include "chrome/browser/autocomplete/keyword_provider.h" @@ -21,6 +23,7 @@ #include "chrome/browser/net/url_fixer_upper.h" #include "chrome/browser/pref_service.h" #include "chrome/browser/profile.h" +#include "chrome/common/chrome_switches.h" #include "chrome/common/notification_service.h" #include "chrome/common/pref_names.h" #include "chrome/common/url_constants.h" @@ -756,7 +759,11 @@ AutocompleteController::AutocompleteController(Profile* profile) have_committed_during_this_query_(false), done_(true) { providers_.push_back(new SearchProvider(this, profile)); - providers_.push_back(new HistoryURLProvider(this, profile)); + if (CommandLine::ForCurrentProcess()->HasSwitch( + switches::kEnableInMemoryURLIndex)) + providers_.push_back(new HistoryQuickProvider(this, profile)); + else + providers_.push_back(new HistoryURLProvider(this, profile)); providers_.push_back(new KeywordProvider(this, profile)); history_contents_provider_ = new HistoryContentsProvider(this, profile); providers_.push_back(history_contents_provider_); diff --git a/chrome/browser/autocomplete/history_quick_provider.cc b/chrome/browser/autocomplete/history_quick_provider.cc new file mode 100644 index 0000000..0d4fdda --- /dev/null +++ b/chrome/browser/autocomplete/history_quick_provider.cc @@ -0,0 +1,10 @@ +// Copyright (c) 2010 The Chromium Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. + +#include "chrome/browser/autocomplete/history_quick_provider.h" + +void HistoryQuickProvider::Start(const AutocompleteInput& input, + bool minimal_changes) { + // TODO(mrossetti): Implement. +} diff --git a/chrome/browser/autocomplete/history_quick_provider.h b/chrome/browser/autocomplete/history_quick_provider.h new file mode 100644 index 0000000..c0dcc15 --- /dev/null +++ b/chrome/browser/autocomplete/history_quick_provider.h @@ -0,0 +1,39 @@ +// Copyright (c) 2010 The Chromium Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. + +#ifndef CHROME_BROWSER_AUTOCOMPLETE_HISTORY_QUICK_PROVIDER_H_ +#define CHROME_BROWSER_AUTOCOMPLETE_HISTORY_QUICK_PROVIDER_H_ + +#include "chrome/browser/autocomplete/autocomplete.h" + +// This class is an autocomplete provider (a pseudo-internal component of +// the history system) which quickly (and synchronously) provides matching +// results from recently or frequently visited sites in the profile's +// history. +// +// TODO(mrossetti): Review to see if the following applies since we're not +// using the database during the autocomplete pass. +// +// Note: This object can get leaked on shutdown if there are pending +// requests on the database (which hold a reference to us). Normally, these +// messages get flushed for each thread. We do a round trip from main, to +// history, back to main while holding a reference. If the main thread +// completes before the history thread, the message to delegate back to the +// main thread will not run and the reference will leak. Therefore, don't do +// anything on destruction. +class HistoryQuickProvider : public AutocompleteProvider { + public: + HistoryQuickProvider(ACProviderListener* listener, Profile* profile) + : AutocompleteProvider(listener, profile, "HistoryQuickProvider") {} + + // no destructor (see note above) + + // AutocompleteProvider + void Start(const AutocompleteInput& input, bool minimal_changes); + + private: + ~HistoryQuickProvider() {} +}; + +#endif // CHROME_BROWSER_AUTOCOMPLETE_HISTORY_QUICK_PROVIDER_H_ diff --git a/chrome/browser/autocomplete/history_quick_provider_unittest.cc b/chrome/browser/autocomplete/history_quick_provider_unittest.cc new file mode 100644 index 0000000..24ed677 --- /dev/null +++ b/chrome/browser/autocomplete/history_quick_provider_unittest.cc @@ -0,0 +1,35 @@ +// Copyright (c) 2010 The Chromium Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. + +#include "chrome/browser/autocomplete/history_quick_provider.h" + +#include "base/scoped_ptr.h" +#include "chrome/test/testing_profile.h" +#include "testing/gtest/include/gtest/gtest.h" + +class HistoryQuickProviderTest : public testing::Test, + public ACProviderListener { + public: + // ACProviderListener + virtual void OnProviderUpdate(bool updated_matches); + + protected: + void SetUp() { + profile_.reset(new TestingProfile()); + provider_ = new HistoryQuickProvider(this, profile_.get()); + } + void TearDown() { + provider_ = NULL; + } + + scoped_refptr<HistoryQuickProvider> provider_; + scoped_ptr<TestingProfile> profile_; +}; + +void HistoryQuickProviderTest::OnProviderUpdate(bool updated_matches) { +} + +TEST_F(HistoryQuickProviderTest, Construction) { + EXPECT_TRUE(provider_.get()); +} diff --git a/chrome/browser/autocomplete/history_url_provider.cc b/chrome/browser/autocomplete/history_url_provider.cc index bac2d1a..0f339f4 100644 --- a/chrome/browser/autocomplete/history_url_provider.cc +++ b/chrome/browser/autocomplete/history_url_provider.cc @@ -7,7 +7,6 @@ #include <algorithm> #include "base/basictypes.h" -#include "base/command_line.h" #include "base/histogram.h" #include "base/message_loop.h" #include "base/string_util.h" @@ -18,7 +17,6 @@ #include "chrome/browser/net/url_fixer_upper.h" #include "chrome/browser/pref_service.h" #include "chrome/browser/profile.h" -#include "chrome/common/chrome_switches.h" #include "chrome/common/pref_names.h" #include "chrome/common/url_constants.h" #include "googleurl/src/gurl.h" @@ -29,6 +27,80 @@ using base::Time; using base::TimeDelta; using base::TimeTicks; +// TODO(mrossetti): Move these to a more appropriate place. +using history::Prefix; +using history::Prefixes; +using history::HistoryMatch; +using history::HistoryMatches; + +namespace history { + +// Returns true if |url| is just a host (e.g. "http://www.google.com/") and +// not some other subpage (e.g. "http://www.google.com/foo.html"). +bool IsHostOnly(const GURL& url) { + DCHECK(url.is_valid()); + return (!url.has_path() || (url.path() == "/")) && !url.has_query() && + !url.has_ref(); +} + +// Acts like the > operator for URLInfo classes. +bool CompareHistoryMatch(const HistoryMatch& a, const HistoryMatch& b) { + // A URL that has been typed at all is better than one that has never been + // typed. (Note "!"s on each side) + if (!a.url_info.typed_count() != !b.url_info.typed_count()) + return a.url_info.typed_count() > b.url_info.typed_count(); + + // Innermost matches (matches after any scheme or "www.") are better than + // non-innermost matches. + if (a.innermost_match != b.innermost_match) + return a.innermost_match; + + // URLs that have been typed more often are better. + if (a.url_info.typed_count() != b.url_info.typed_count()) + return a.url_info.typed_count() > b.url_info.typed_count(); + + // For URLs that have each been typed once, a host (alone) is better than a + // page inside. + if (a.url_info.typed_count() == 1) { + const bool a_is_host_only = history::IsHostOnly(a.url_info.url()); + if (a_is_host_only != history::IsHostOnly(b.url_info.url())) + return a_is_host_only; + } + + // URLs that have been visited more often are better. + if (a.url_info.visit_count() != b.url_info.visit_count()) + return a.url_info.visit_count() > b.url_info.visit_count(); + + // URLs that have been visited more recently are better. + return a.url_info.last_visit() > b.url_info.last_visit(); +} + +// Given the user's |input| and a |match| created from it, reduce the +// match's URL to just a host. If this host still matches the user input, +// return it. Returns the empty string on failure. +GURL ConvertToHostOnly(const HistoryMatch& match, const std::wstring& input) { + // See if we should try to do host-only suggestions for this URL. Nonstandard + // schemes means there's no authority section, so suggesting the host name + // is useless. File URLs are standard, but host suggestion is not useful for + // them either. + const GURL& url = match.url_info.url(); + if (!url.is_valid() || !url.IsStandard() || url.SchemeIsFile()) + return GURL(); + + // Transform to a host-only match. Bail if the host no longer matches the + // user input (e.g. because the user typed more than just a host). + GURL host = url.GetWithEmptyPath(); + if ((host.spec().length() < (match.input_location + input.length()))) + return GURL(); // User typing is longer than this host suggestion. + + const std::wstring spec = UTF8ToWide(host.spec()); + if (spec.compare(match.input_location, input.length(), input)) + return GURL(); // User typing is no longer a prefix. + + return host; +} + +} // namespace history HistoryURLProviderParams::HistoryURLProviderParams( const AutocompleteInput& input, @@ -151,30 +223,25 @@ void HistoryURLProvider::DoAutocomplete(history::HistoryBackend* backend, URLRowVector url_matches; HistoryMatches history_matches; - if (CommandLine::ForCurrentProcess()->HasSwitch( - switches::kEnableInMemoryURLIndex)) { - // TODO(rohitrao): Fetch results from the index. - } else { - for (Prefixes::const_iterator i(prefixes_.begin()); i != prefixes_.end(); - ++i) { - if (params->cancel) - return; // Canceled in the middle of a query, give up. - // We only need kMaxMatches results in the end, but before we get there we - // need to promote lower-quality matches that are prefixes of - // higher-quality matches, and remove lower-quality redirects. So we ask - // for more results than we need, of every prefix type, in hopes this will - // give us far more than enough to work with. CullRedirects() will then - // reduce the list to the best kMaxMatches results. - db->AutocompleteForPrefix(WideToUTF16(i->prefix + params->input.text()), - kMaxMatches * 2, &url_matches); - for (URLRowVector::const_iterator j(url_matches.begin()); - j != url_matches.end(); ++j) { - const Prefix* best_prefix = BestPrefix(j->url(), std::wstring()); - DCHECK(best_prefix != NULL); - history_matches.push_back(HistoryMatch(*j, i->prefix.length(), - !i->num_components, - i->num_components >= best_prefix->num_components)); - } + for (Prefixes::const_iterator i(prefixes_.begin()); i != prefixes_.end(); + ++i) { + if (params->cancel) + return; // Canceled in the middle of a query, give up. + // We only need kMaxMatches results in the end, but before we get there we + // need to promote lower-quality matches that are prefixes of + // higher-quality matches, and remove lower-quality redirects. So we ask + // for more results than we need, of every prefix type, in hopes this will + // give us far more than enough to work with. CullRedirects() will then + // reduce the list to the best kMaxMatches results. + db->AutocompleteForPrefix(WideToUTF16(i->prefix + params->input.text()), + kMaxMatches * 2, &url_matches); + for (URLRowVector::const_iterator j(url_matches.begin()); + j != url_matches.end(); ++j) { + const Prefix* best_prefix = BestPrefix(j->url(), std::wstring()); + DCHECK(best_prefix != NULL); + history_matches.push_back(HistoryMatch(*j, i->prefix.length(), + !i->num_components, + i->num_components >= best_prefix->num_components)); } } @@ -360,7 +427,7 @@ bool HistoryURLProvider::PromoteMatchForInlineAutocomplete( // hand, we wouldn't want to immediately start autocompleting it. if (!match.url_info.typed_count() || ((match.url_info.typed_count() == 1) && - !IsHostOnly(match.url_info.url()))) + !history::IsHostOnly(match.url_info.url()))) return false; params->matches.push_back(HistoryMatchToACMatch(params, match, @@ -454,47 +521,7 @@ size_t HistoryURLProvider::TrimHttpPrefix(std::wstring* url) { } // static -bool HistoryURLProvider::IsHostOnly(const GURL& url) { - DCHECK(url.is_valid()); - return (!url.has_path() || (url.path() == "/")) && !url.has_query() && - !url.has_ref(); -} - -// static -bool HistoryURLProvider::CompareHistoryMatch(const HistoryMatch& a, - const HistoryMatch& b) { - // A URL that has been typed at all is better than one that has never been - // typed. (Note "!"s on each side) - if (!a.url_info.typed_count() != !b.url_info.typed_count()) - return a.url_info.typed_count() > b.url_info.typed_count(); - - // Innermost matches (matches after any scheme or "www.") are better than - // non-innermost matches. - if (a.innermost_match != b.innermost_match) - return a.innermost_match; - - // URLs that have been typed more often are better. - if (a.url_info.typed_count() != b.url_info.typed_count()) - return a.url_info.typed_count() > b.url_info.typed_count(); - - // For URLs that have each been typed once, a host (alone) is better than a - // page inside. - if (a.url_info.typed_count() == 1) { - const bool a_is_host_only = IsHostOnly(a.url_info.url()); - if (a_is_host_only != IsHostOnly(b.url_info.url())) - return a_is_host_only; - } - - // URLs that have been visited more often are better. - if (a.url_info.visit_count() != b.url_info.visit_count()) - return a.url_info.visit_count() > b.url_info.visit_count(); - - // URLs that have been visited more recently are better. - return a.url_info.last_visit() > b.url_info.last_visit(); -} - -// static -HistoryURLProvider::Prefixes HistoryURLProvider::GetPrefixes() { +history::Prefixes HistoryURLProvider::GetPrefixes() { // We'll complete text following these prefixes. // NOTE: There's no requirement that these be in any particular order. Prefixes prefixes; @@ -526,30 +553,6 @@ int HistoryURLProvider::CalculateRelevance(AutocompleteInput::Type input_type, } // static -GURL HistoryURLProvider::ConvertToHostOnly(const HistoryMatch& match, - const std::wstring& input) { - // See if we should try to do host-only suggestions for this URL. Nonstandard - // schemes means there's no authority section, so suggesting the host name - // is useless. File URLs are standard, but host suggestion is not useful for - // them either. - const GURL& url = match.url_info.url(); - if (!url.is_valid() || !url.IsStandard() || url.SchemeIsFile()) - return GURL(); - - // Transform to a host-only match. Bail if the host no longer matches the - // user input (e.g. because the user typed more than just a host). - GURL host = url.GetWithEmptyPath(); - if ((host.spec().length() < (match.input_location + input.length()))) - return GURL(); // User typing is longer than this host suggestion. - - const std::wstring spec = UTF8ToWide(host.spec()); - if (spec.compare(match.input_location, input.length(), input)) - return GURL(); // User typing is no longer a prefix. - - return host; -} - -// static void HistoryURLProvider::PromoteOrCreateShorterSuggestion( history::URLDatabase* db, const HistoryURLProviderParams& params, @@ -563,7 +566,7 @@ void HistoryURLProvider::PromoteOrCreateShorterSuggestion( // itself be added as a match. We can add the base iff it's not "effectively // the same" as any "what you typed" match. const HistoryMatch& match = matches->front(); - GURL search_base = ConvertToHostOnly(match, params.input.text()); + GURL search_base = history::ConvertToHostOnly(match, params.input.text()); bool can_add_search_base_to_matches = !have_what_you_typed_match; if (search_base.is_empty()) { // Search from what the user typed when we couldn't reduce the best match @@ -576,6 +579,9 @@ void HistoryURLProvider::PromoteOrCreateShorterSuggestion( std::string new_match = match.url_info.url().possibly_invalid_spec(). substr(0, match.input_location + params.input.text().length()); search_base = GURL(new_match); + // TODO(mrossetti): There is a degenerate case where the following may + // cause a failure: http://www/~someword/fubar.html. Diagnose. + // See: http://crbug.com/50101 if (search_base.is_empty()) return; // Can't construct a valid URL from which to start a search. } else if (!can_add_search_base_to_matches) { @@ -722,7 +728,7 @@ void HistoryURLProvider::RunAutocompletePasses( } } -const HistoryURLProvider::Prefix* HistoryURLProvider::BestPrefix( +const history::Prefix* HistoryURLProvider::BestPrefix( const GURL& url, const std::wstring& prefix_suffix) const { const Prefix* best_prefix = NULL; @@ -742,7 +748,7 @@ const HistoryURLProvider::Prefix* HistoryURLProvider::BestPrefix( void HistoryURLProvider::SortMatches(HistoryMatches* matches) const { // Sort by quality, best first. - std::sort(matches->begin(), matches->end(), &CompareHistoryMatch); + std::sort(matches->begin(), matches->end(), &history::CompareHistoryMatch); // Remove duplicate matches (caused by the search string appearing in one of // the prefixes as well as after it). Consider the following scenario: @@ -775,15 +781,13 @@ void HistoryURLProvider::SortMatches(HistoryMatches* matches) const { } void HistoryURLProvider::CullPoorMatches(HistoryMatches* matches) const { - static const int kLowQualityMatchTypedLimit = 1; - static const int kLowQualityMatchVisitLimit = 3; - static const int kLowQualityMatchAgeLimitInDays = 3; - Time recent_threshold = - Time::Now() - TimeDelta::FromDays(kLowQualityMatchAgeLimitInDays); + Time recent_threshold = history::AutocompleteAgeThreshold(); for (HistoryMatches::iterator i(matches->begin()); i != matches->end();) { - const history::URLRow& url_info = i->url_info; - if ((url_info.typed_count() <= kLowQualityMatchTypedLimit) && - (url_info.visit_count() <= kLowQualityMatchVisitLimit) && + const history::URLRow& url_info(i->url_info); + if ((url_info.typed_count() <= + history::kLowQualityMatchTypedLimit) && + (url_info.visit_count() <= + history::kLowQualityMatchVisitLimit) && (url_info.last_visit() < recent_threshold)) { i = matches->erase(i); } else { diff --git a/chrome/browser/autocomplete/history_url_provider.h b/chrome/browser/autocomplete/history_url_provider.h index 56633b4..bfd13b9 100644 --- a/chrome/browser/autocomplete/history_url_provider.h +++ b/chrome/browser/autocomplete/history_url_provider.h @@ -17,10 +17,9 @@ class MessageLoop; class Profile; namespace history { -class HistoryBackend; + class HistoryBackend; } // namespace history - // How history autocomplete works // ============================== // @@ -180,69 +179,6 @@ class HistoryURLProvider : public AutocompleteProvider { private: ~HistoryURLProvider() {} - struct Prefix { - Prefix(std::wstring prefix, int num_components) - : prefix(prefix), - num_components(num_components) { } - - std::wstring prefix; - - // The number of "components" in the prefix. The scheme is a component, - // and the initial "www." or "ftp." is a component. So "http://foo.com" - // and "www.bar.com" each have one component, "ftp://ftp.ftp.com" has two, - // and "mysite.com" has none. This is used to tell whether the user's - // input is an innermost match or not. See comments in HistoryMatch. - int num_components; - }; - typedef std::vector<Prefix> Prefixes; - - // Used for intermediate history result operations. - struct HistoryMatch { - // Required for STL, we don't use this directly. - HistoryMatch() - : url_info(), - input_location(std::wstring::npos), - match_in_scheme(false), - innermost_match(true) { - } - - HistoryMatch(const history::URLRow& url_info, - size_t input_location, - bool match_in_scheme, - bool innermost_match) - : url_info(url_info), - input_location(input_location), - match_in_scheme(match_in_scheme), - innermost_match(innermost_match) { - } - - bool operator==(const GURL& url) const { - return url_info.url() == url; - } - - history::URLRow url_info; - - // The offset of the user's input within the URL. - size_t input_location; - - // Whether this is a match in the scheme. This determines whether we'll go - // ahead and show a scheme on the URL even if the user didn't type one. - // If our best match was in the scheme, not showing the scheme is both - // confusing and, for inline autocomplete of the fill_into_edit, dangerous. - // (If the user types "h" and we match "http://foo/", we need to inline - // autocomplete that, not "foo/", which won't show anything at all, and - // will mislead the user into thinking the What You Typed match is what's - // selected.) - bool match_in_scheme; - - // A match after any scheme/"www.", if the user input could match at both - // locations. If the user types "w", an innermost match ("website.com") is - // better than a non-innermost match ("www.google.com"). If the user types - // "x", no scheme in our prefix list (or "www.") begins with x, so all - // matches are, vacuously, "innermost matches". - bool innermost_match; - }; - typedef std::deque<HistoryMatch> HistoryMatches; enum MatchType { NORMAL, @@ -268,16 +204,8 @@ class HistoryURLProvider : public AutocompleteProvider { // return 0. static size_t TrimHttpPrefix(std::wstring* url); - // Returns true if |url| is just a host (e.g. "http://www.google.com/") and - // not some other subpage (e.g. "http://www.google.com/foo.html"). - static bool IsHostOnly(const GURL& url); - - // Acts like the > operator for URLInfo classes. - static bool CompareHistoryMatch(const HistoryMatch& a, - const HistoryMatch& b); - // Returns the set of prefixes to use for prefixes_. - static Prefixes GetPrefixes(); + static history::Prefixes GetPrefixes(); // Determines the relevance for some input, given its type and which match it // is. If |match_type| is NORMAL, |match_number| is a number @@ -290,7 +218,7 @@ class HistoryURLProvider : public AutocompleteProvider { // Given the user's |input| and a |match| created from it, reduce the // match's URL to just a host. If this host still matches the user input, // return it. Returns the empty string on failure. - static GURL ConvertToHostOnly(const HistoryMatch& match, + static GURL ConvertToHostOnly(const history::HistoryMatch& match, const std::wstring& input); // See if a shorter version of the best match should be created, and if so @@ -305,7 +233,7 @@ class HistoryURLProvider : public AutocompleteProvider { const HistoryURLProviderParams& params, bool have_what_you_typed_match, const AutocompleteMatch& what_you_typed_match, - HistoryMatches* matches); + history::HistoryMatches* matches); // Ensures that |matches| contains an entry for |info|, which may mean adding // a new such entry (using |input_location| and |match_in_scheme|). @@ -317,7 +245,7 @@ class HistoryURLProvider : public AutocompleteProvider { static void EnsureMatchPresent(const history::URLRow& info, std::wstring::size_type input_location, bool match_in_scheme, - HistoryMatches* matches, + history::HistoryMatches* matches, bool promote); // Helper function that actually launches the two autocomplete passes. @@ -330,8 +258,8 @@ class HistoryURLProvider : public AutocompleteProvider { // |prefix_suffix| (which may be empty) is appended to every attempted // prefix. This is useful when you need to figure out the innermost match // for some user input in a URL. - const Prefix* BestPrefix(const GURL& text, - const std::wstring& prefix_suffix) const; + const history::Prefix* BestPrefix(const GURL& text, + const std::wstring& prefix_suffix) const; // Returns a match corresponding to exactly what the user has typed. AutocompleteMatch SuggestExactInput(const AutocompleteInput& input, @@ -346,25 +274,25 @@ class HistoryURLProvider : public AutocompleteProvider { bool FixupExactSuggestion(history::URLDatabase* db, const AutocompleteInput& input, AutocompleteMatch* match, - HistoryMatches* matches) const; + history::HistoryMatches* matches) const; // Determines if |match| is suitable for inline autocomplete, and promotes it // if so. bool PromoteMatchForInlineAutocomplete(HistoryURLProviderParams* params, - const HistoryMatch& match); + const history::HistoryMatch& match); // Sorts the given list of matches. - void SortMatches(HistoryMatches* matches) const; + void SortMatches(history::HistoryMatches* matches) const; // Removes results that have been rarely typed or visited, and not any time // recently. The exact parameters for this heuristic can be found in the // function body. - void CullPoorMatches(HistoryMatches* matches) const; + void CullPoorMatches(history::HistoryMatches* matches) const; // Removes results that redirect to each other, leaving at most |max_results| // results. void CullRedirects(history::HistoryBackend* backend, - HistoryMatches* matches, + history::HistoryMatches* matches, size_t max_results) const; // Helper function for CullRedirects, this removes all but the first @@ -376,18 +304,19 @@ class HistoryURLProvider : public AutocompleteProvider { // is removed, the next item will be shifted, and this allows the caller to // pick up on the next one when this happens. size_t RemoveSubsequentMatchesOf( - HistoryMatches* matches, + history::HistoryMatches* matches, size_t source_index, const std::vector<GURL>& remove) const; // Converts a line from the database into an autocomplete match for display. - AutocompleteMatch HistoryMatchToACMatch(HistoryURLProviderParams* params, - const HistoryMatch& history_match, - MatchType match_type, - size_t match_number); + AutocompleteMatch HistoryMatchToACMatch( + HistoryURLProviderParams* params, + const history::HistoryMatch& history_match, + MatchType match_type, + size_t match_number); // Prefixes to try appending to user input when looking for a match. - const Prefixes prefixes_; + const history::Prefixes prefixes_; // Params for the current query. The provider should not free this directly; // instead, it is passed as a parameter through the history backend, and the diff --git a/chrome/browser/history/history.cc b/chrome/browser/history/history.cc index aaec4207..c001298 100644 --- a/chrome/browser/history/history.cc +++ b/chrome/browser/history/history.cc @@ -222,6 +222,16 @@ history::URLDatabase* HistoryService::InMemoryDatabase() { return NULL; } +history::InMemoryURLIndex* HistoryService::InMemoryIndex() { + // NOTE: See comments in BackendLoaded() as to why we call + // LoadBackendIfNecessary() here even though it won't affect the return value + // for this call. + LoadBackendIfNecessary(); + if (in_memory_backend_.get()) + return in_memory_backend_->index(); + return NULL; +} + void HistoryService::SetSegmentPresentationIndex(int64 segment_id, int index) { ScheduleAndForget(PRIORITY_UI, &HistoryBackend::SetSegmentPresentationIndex, diff --git a/chrome/browser/history/history.h b/chrome/browser/history/history.h index 8da542c..121d573 100644 --- a/chrome/browser/history/history.h +++ b/chrome/browser/history/history.h @@ -51,6 +51,7 @@ class TypedUrlDataTypeController; namespace history { class InMemoryHistoryBackend; +class InMemoryURLIndex; class HistoryBackend; class HistoryDatabase; struct HistoryDetails; @@ -151,6 +152,8 @@ class HistoryService : public CancelableRequestProvider, // TODO(brettw) this should return the InMemoryHistoryBackend. history::URLDatabase* InMemoryDatabase(); + history::InMemoryURLIndex* InMemoryIndex(); + // Navigation ---------------------------------------------------------------- // Adds the given canonical URL to history with the current time as the visit diff --git a/chrome/browser/history/history_backend.cc b/chrome/browser/history/history_backend.cc index 0f512db..8e1c5fb 100644 --- a/chrome/browser/history/history_backend.cc +++ b/chrome/browser/history/history_backend.cc @@ -548,7 +548,7 @@ void HistoryBackend::InitImpl() { // Fill the in-memory database and send it back to the history service on the // main thread. InMemoryHistoryBackend* mem_backend = new InMemoryHistoryBackend; - if (mem_backend->Init(history_name)) + if (mem_backend->Init(history_name, db_.get())) delegate_->SetInMemoryBackend(mem_backend); // Takes ownership of pointer. else delete mem_backend; // Error case, run without the in-memory DB. diff --git a/chrome/browser/history/history_types.cc b/chrome/browser/history/history_types.cc index 50395aa..305fa35 100644 --- a/chrome/browser/history/history_types.cc +++ b/chrome/browser/history/history_types.cc @@ -10,6 +10,7 @@ #include "base/stl_util-inl.h" using base::Time; +using base::TimeDelta; namespace history { @@ -237,4 +238,10 @@ void QueryResults::AdjustResultMap(size_t begin, size_t end, ptrdiff_t delta) { } } +Time AutocompleteAgeThreshold() { + Time recent_threshold = + Time::Now() - TimeDelta::FromDays(kLowQualityMatchAgeLimitInDays); + return recent_threshold; +} + } // namespace history diff --git a/chrome/browser/history/history_types.h b/chrome/browser/history/history_types.h index 4619792..0c283b1 100644 --- a/chrome/browser/history/history_types.h +++ b/chrome/browser/history/history_types.h @@ -6,6 +6,7 @@ #define CHROME_BROWSER_HISTORY_HISTORY_TYPES_H_ #pragma once +#include <deque> #include <map> #include <set> #include <string> @@ -311,7 +312,7 @@ struct StarredEntry { // If type == URL, this is the ID of the URL of the primary page that was // starred. - history::URLID url_id; + URLID url_id; // Time the entry was last modified. This is only used for groups and // indicates the last time a URL was added as a child to the group. @@ -528,6 +529,80 @@ struct MostVisitedURL { typedef std::vector<MostVisitedURL> MostVisitedURLList; -} // history +// Used for intermediate history result operations. +struct HistoryMatch { + // Required for STL, we don't use this directly. + HistoryMatch() + : url_info(), + input_location(std::wstring::npos), + match_in_scheme(false), + innermost_match(true) { + } + + HistoryMatch(const URLRow& url_info, + size_t input_location, + bool match_in_scheme, + bool innermost_match) + : url_info(url_info), + input_location(input_location), + match_in_scheme(match_in_scheme), + innermost_match(innermost_match) { + } + + bool operator==(const GURL& url) const { + return url_info.url() == url; + } + + URLRow url_info; + + // The offset of the user's input within the URL. + size_t input_location; + + // Whether this is a match in the scheme. This determines whether we'll go + // ahead and show a scheme on the URL even if the user didn't type one. + // If our best match was in the scheme, not showing the scheme is both + // confusing and, for inline autocomplete of the fill_into_edit, dangerous. + // (If the user types "h" and we match "http://foo/", we need to inline + // autocomplete that, not "foo/", which won't show anything at all, and + // will mislead the user into thinking the What You Typed match is what's + // selected.) + bool match_in_scheme; + + // A match after any scheme/"www.", if the user input could match at both + // locations. If the user types "w", an innermost match ("website.com") is + // better than a non-innermost match ("www.google.com"). If the user types + // "x", no scheme in our prefix list (or "www.") begins with x, so all + // matches are, vacuously, "innermost matches". + bool innermost_match; +}; +typedef std::deque<HistoryMatch> HistoryMatches; + +struct Prefix { + Prefix(std::wstring const& prefix, int num_components) + : prefix(prefix), + num_components(num_components) {} + + std::wstring prefix; + + // The number of "components" in the prefix. The scheme is a component, + // and the initial "www." or "ftp." is a component. So "http://foo.com" + // and "www.bar.com" each have one component, "ftp://ftp.ftp.com" has two, + // and "mysite.com" has none. This is used to tell whether the user's + // input is an innermost match or not. See comments in HistoryMatch. + int num_components; +}; +typedef std::vector<Prefix> Prefixes; + +// Constants which specify, when considered altogether, 'significant' +// history items. These are used to filter out insignificant items +// for consideration as autocomplete candidates. +const int kLowQualityMatchTypedLimit = 1; +const int kLowQualityMatchVisitLimit = 3; +const int kLowQualityMatchAgeLimitInDays = 3; + +// Returns the date threshold for considering an history item as significant. +base::Time AutocompleteAgeThreshold(); + +} // namespace history #endif // CHROME_BROWSER_HISTORY_HISTORY_TYPES_H_ diff --git a/chrome/browser/history/in_memory_history_backend.cc b/chrome/browser/history/in_memory_history_backend.cc index 9f3e7be..1350667 100644 --- a/chrome/browser/history/in_memory_history_backend.cc +++ b/chrome/browser/history/in_memory_history_backend.cc @@ -5,10 +5,13 @@ #include "chrome/browser/history/in_memory_history_backend.h" #include "base/command_line.h" +#include "base/histogram.h" +#include "base/time.h" #include "chrome/browser/browser_process.h" #include "chrome/browser/history/history_notifications.h" #include "chrome/browser/history/in_memory_database.h" #include "chrome/browser/history/in_memory_url_index.h" +#include "chrome/browser/history/url_database.h" #include "chrome/browser/profile.h" #include "chrome/common/chrome_switches.h" #include "chrome/common/notification_service.h" @@ -26,14 +29,18 @@ InMemoryHistoryBackend::InMemoryHistoryBackend() InMemoryHistoryBackend::~InMemoryHistoryBackend() { } -bool InMemoryHistoryBackend::Init(const FilePath& history_filename) { +bool InMemoryHistoryBackend::Init(const FilePath& history_filename, + URLDatabase* db) { db_.reset(new InMemoryDatabase); bool success = db_->InitFromDisk(history_filename); if (CommandLine::ForCurrentProcess()->HasSwitch( switches::kEnableInMemoryURLIndex)) { index_.reset(new InMemoryURLIndex); - // TODO(rohitrao): Load index. + base::TimeTicks beginning_time = base::TimeTicks::Now(); + index_->Init(db); + UMA_HISTOGRAM_TIMES("Autocomplete.HistoryDatabaseIndexingTime", + base::TimeTicks::Now() - beginning_time); } return success; diff --git a/chrome/browser/history/in_memory_history_backend.h b/chrome/browser/history/in_memory_history_backend.h index b10bb31..ac4c185 100644 --- a/chrome/browser/history/in_memory_history_backend.h +++ b/chrome/browser/history/in_memory_history_backend.h @@ -29,6 +29,7 @@ namespace history { class InMemoryDatabase; class InMemoryURLIndex; +class URLDatabase; struct URLsDeletedDetails; struct URLsModifiedDetails; @@ -38,7 +39,7 @@ class InMemoryHistoryBackend : public NotificationObserver { ~InMemoryHistoryBackend(); // Initializes with data from the given history database. - bool Init(const FilePath& history_filename); + bool Init(const FilePath& history_filename, URLDatabase* db); // Does initialization work when this object is attached to the history // system on the main thread. The argument is the profile with which the diff --git a/chrome/browser/history/in_memory_url_index.cc b/chrome/browser/history/in_memory_url_index.cc index 83c401f..7f249df2 100644 --- a/chrome/browser/history/in_memory_url_index.cc +++ b/chrome/browser/history/in_memory_url_index.cc @@ -4,10 +4,16 @@ #include "chrome/browser/history/in_memory_url_index.h" +#include "chrome/browser/history/url_database.h" + namespace history { -InMemoryURLIndex::InMemoryURLIndex() {} +// Indexing -InMemoryURLIndex::~InMemoryURLIndex() {} +bool InMemoryURLIndex::Init(history::URLDatabase* history_db) { + bool success = true; + // TODO(mrossetti): Implement. + return success; +} } // namespace history diff --git a/chrome/browser/history/in_memory_url_index.h b/chrome/browser/history/in_memory_url_index.h index 166c400..5b7861a 100644 --- a/chrome/browser/history/in_memory_url_index.h +++ b/chrome/browser/history/in_memory_url_index.h @@ -8,6 +8,8 @@ namespace history { +class URLDatabase; + // The URL history source. // Holds portions of the URL database in memory in an indexed form. Used to // quickly look up matching URLs for a given query string. Used by @@ -15,8 +17,11 @@ namespace history { // matches to the omnibox. class InMemoryURLIndex { public: - InMemoryURLIndex(); - ~InMemoryURLIndex(); + InMemoryURLIndex() {} + ~InMemoryURLIndex() {} + + // Open and index the URL history database. + bool Init(URLDatabase* history_db); }; } // namespace history diff --git a/chrome/chrome_browser.gypi b/chrome/chrome_browser.gypi index 468dc3d..5b92731 100644 --- a/chrome/chrome_browser.gypi +++ b/chrome/chrome_browser.gypi @@ -107,6 +107,8 @@ 'browser/autocomplete/autocomplete_popup_view_mac.mm', 'browser/autocomplete/history_contents_provider.cc', 'browser/autocomplete/history_contents_provider.h', + 'browser/autocomplete/history_quick_provider.cc', + 'browser/autocomplete/history_quick_provider.h', 'browser/autocomplete/history_url_provider.cc', 'browser/autocomplete/history_url_provider.h', 'browser/autocomplete/keyword_provider.cc', diff --git a/chrome/chrome_tests.gypi b/chrome/chrome_tests.gypi index 7418532..c33e615 100644 --- a/chrome/chrome_tests.gypi +++ b/chrome/chrome_tests.gypi @@ -677,6 +677,7 @@ 'browser/autocomplete/autocomplete_popup_view_mac_unittest.mm', 'browser/autocomplete/autocomplete_unittest.cc', 'browser/autocomplete/history_contents_provider_unittest.cc', + 'browser/autocomplete/history_quick_provider_unittest.cc', 'browser/autocomplete/history_url_provider_unittest.cc', 'browser/autocomplete/keyword_provider_unittest.cc', 'browser/autocomplete/search_provider_unittest.cc', |