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_INPUT_H_ 6 #define COMPONENTS_OMNIBOX_AUTOCOMPLETE_INPUT_H_ 7 8 #include <string> 9 10 #include "base/basictypes.h" 11 #include "base/gtest_prod_util.h" 12 #include "base/strings/string16.h" 13 #include "components/metrics/proto/omnibox_event.pb.h" 14 #include "components/metrics/proto/omnibox_input_type.pb.h" 15 #include "url/gurl.h" 16 #include "url/url_parse.h" 17 18 class AutocompleteSchemeClassifier; 19 20 // The user input for an autocomplete query. Allows copying. 21 class AutocompleteInput { 22 public: 23 AutocompleteInput(); 24 // |text| and |cursor_position| represent the input query and location of 25 // the cursor with the query respectively. |cursor_position| may be set to 26 // base::string16::npos if the input |text| doesn't come directly from the 27 // user's typing. 28 // 29 // |desired_tld| is the user's desired TLD, if one is not already present in 30 // the text to autocomplete. When this is non-empty, it also implies that 31 // "www." should be prepended to the domain where possible. The |desired_tld| 32 // should not contain a leading '.' (use "com" instead of ".com"). 33 // 34 // If |current_url| is set to a valid search result page URL, providers can 35 // use it to perform query refinement. For example, if it is set to an image 36 // search result page, the search provider may generate an image search URL. 37 // Query refinement is only used by mobile ports, so only these set 38 // |current_url| to a non-empty string. 39 // 40 // |current_page_classification| represents the type of page the user is 41 // viewing and manner in which the user is accessing the omnibox; it's 42 // more than simply the URL. It includes, for example, whether the page 43 // is a search result page doing search term replacement or not. 44 // 45 // |prevent_inline_autocomplete| is true if the generated result set should 46 // not require inline autocomplete for the default match. This is difficult 47 // to explain in the abstract; the practical use case is that after the user 48 // deletes text in the edit, the HistoryURLProvider should make sure not to 49 // promote a match requiring inline autocomplete too highly. 50 // 51 // |prefer_keyword| should be true when the keyword UI is onscreen; this will 52 // bias the autocomplete result set toward the keyword provider when the input 53 // string is a bare keyword. 54 // 55 // |allow_exact_keyword_match| should be false when triggering keyword mode on 56 // the input string would be surprising or wrong, e.g. when highlighting text 57 // in a page and telling the browser to search for it or navigate to it. This 58 // parameter only applies to substituting keywords. 59 // 60 // If |want_asynchronous_matches| is false the controller asks the providers 61 // to only return matches which are synchronously available, which should mean 62 // that all providers will be done immediately. 63 // 64 // |scheme_classifier| is passed to Parse() to help determine the type of 65 // input this is; see comments there. 66 AutocompleteInput(const base::string16& text, 67 size_t cursor_position, 68 const base::string16& desired_tld, 69 const GURL& current_url, 70 metrics::OmniboxEventProto::PageClassification 71 current_page_classification, 72 bool prevent_inline_autocomplete, 73 bool prefer_keyword, 74 bool allow_exact_keyword_match, 75 bool want_asynchronous_matches, 76 const AutocompleteSchemeClassifier& scheme_classifier); 77 ~AutocompleteInput(); 78 79 // If type is |FORCED_QUERY| and |text| starts with '?', it is removed. 80 // Returns number of leading characters removed. 81 static size_t RemoveForcedQueryStringIfNecessary( 82 metrics::OmniboxInputType::Type type, 83 base::string16* text); 84 85 // Converts |type| to a string representation. Used in logging. 86 static std::string TypeToString(metrics::OmniboxInputType::Type type); 87 88 // Parses |text| (including an optional |desired_tld|) and returns the type of 89 // input this will be interpreted as. |scheme_classifier| is used to check 90 // the scheme in |text| is known and registered in the current environment. 91 // The components of the input are stored in the output parameter |parts|, if 92 // it is non-NULL. The scheme is stored in |scheme| if it is non-NULL. The 93 // canonicalized URL is stored in |canonicalized_url|; however, this URL is 94 // not guaranteed to be valid, especially if the parsed type is, e.g., QUERY. 95 static metrics::OmniboxInputType::Type Parse( 96 const base::string16& text, 97 const base::string16& desired_tld, 98 const AutocompleteSchemeClassifier& scheme_classifier, 99 url::Parsed* parts, 100 base::string16* scheme, 101 GURL* canonicalized_url); 102 103 // Parses |text| and fill |scheme| and |host| by the positions of them. 104 // The results are almost as same as the result of Parse(), but if the scheme 105 // is view-source, this function returns the positions of scheme and host 106 // in the URL qualified by "view-source:" prefix. 107 static void ParseForEmphasizeComponents( 108 const base::string16& text, 109 const AutocompleteSchemeClassifier& scheme_classifier, 110 url::Component* scheme, 111 url::Component* host); 112 113 // Code that wants to format URLs with a format flag including 114 // net::kFormatUrlOmitTrailingSlashOnBareHostname risk changing the meaning if 115 // the result is then parsed as AutocompleteInput. Such code can call this 116 // function with the URL and its formatted string, and it will return a 117 // formatted string with the same meaning as the original URL (i.e. it will 118 // re-append a slash if necessary). Because this uses Parse() under the hood 119 // to determine the meaning of the different strings, callers need to supply a 120 // |scheme_classifier| to pass to Parse(). 121 static base::string16 FormattedStringWithEquivalentMeaning( 122 const GURL& url, 123 const base::string16& formatted_url, 124 const AutocompleteSchemeClassifier& scheme_classifier); 125 126 // Returns the number of non-empty components in |parts| besides the host. 127 static int NumNonHostComponents(const url::Parsed& parts); 128 129 // Returns whether |text| begins "http:" or "view-source:http:". 130 static bool HasHTTPScheme(const base::string16& text); 131 132 // User-provided text to be completed. 133 const base::string16& text() const { return text_; } 134 135 // Returns 0-based cursor position within |text_| or base::string16::npos if 136 // not used. 137 size_t cursor_position() const { return cursor_position_; } 138 139 // Use of this setter is risky, since no other internal state is updated 140 // besides |text_|, |cursor_position_| and |parts_|. Only callers who know 141 // that they're not changing the type/scheme/etc. should use this. 142 void UpdateText(const base::string16& text, 143 size_t cursor_position, 144 const url::Parsed& parts); 145 146 // The current URL, or an invalid GURL if query refinement is not desired. 147 const GURL& current_url() const { return current_url_; } 148 149 // The type of page that is currently behind displayed and how it is 150 // displayed (e.g., with search term replacement or without). 151 metrics::OmniboxEventProto::PageClassification current_page_classification() 152 const { 153 return current_page_classification_; 154 } 155 156 // The type of input supplied. 157 metrics::OmniboxInputType::Type type() const { return type_; } 158 159 // Returns parsed URL components. 160 const url::Parsed& parts() const { return parts_; } 161 162 // The scheme parsed from the provided text; only meaningful when type_ is 163 // URL. 164 const base::string16& scheme() const { return scheme_; } 165 166 // The input as an URL to navigate to, if possible. 167 const GURL& canonicalized_url() const { return canonicalized_url_; } 168 169 // Returns whether inline autocompletion should be prevented. 170 bool prevent_inline_autocomplete() const { 171 return prevent_inline_autocomplete_; 172 } 173 174 // Returns whether, given an input string consisting solely of a substituting 175 // keyword, we should score it like a non-substituting keyword. 176 bool prefer_keyword() const { return prefer_keyword_; } 177 178 // Returns whether this input is allowed to be treated as an exact 179 // keyword match. If not, the default result is guaranteed not to be a 180 // keyword search, even if the input is "<keyword> <search string>". 181 bool allow_exact_keyword_match() const { return allow_exact_keyword_match_; } 182 183 // Returns whether providers should be allowed to make asynchronous requests 184 // when processing this input. 185 bool want_asynchronous_matches() const { return want_asynchronous_matches_; } 186 187 // Resets all internal variables to the null-constructed state. 188 void Clear(); 189 190 private: 191 FRIEND_TEST_ALL_PREFIXES(AutocompleteProviderTest, GetDestinationURL); 192 193 // NOTE: Whenever adding a new field here, please make sure to update Clear() 194 // method. 195 base::string16 text_; 196 size_t cursor_position_; 197 GURL current_url_; 198 metrics::OmniboxEventProto::PageClassification current_page_classification_; 199 metrics::OmniboxInputType::Type type_; 200 url::Parsed parts_; 201 base::string16 scheme_; 202 GURL canonicalized_url_; 203 bool prevent_inline_autocomplete_; 204 bool prefer_keyword_; 205 bool allow_exact_keyword_match_; 206 bool want_asynchronous_matches_; 207 }; 208 209 #endif // COMPONENTS_OMNIBOX_AUTOCOMPLETE_INPUT_H_ 210
