Merge lp:~tombeckmann/+junk/libcolumbus into lp:~elementary-pantheon/slingshot/trunk

[Tom Beckmann]

This adds support for libcolumbus, it replaces the previous search and weighting algorithm, which is still used if columbus is disabled before compilation. If I'm not completely mistaken, the app relevancies as provided by Zeitgeist should still be used to sort the results returned by libcolubmus.

Please test if the results shown are indeed sensible, otherwise I'll go and adjust the index weights or check which part of the API I misunderstood.

[Danielle Foré (danrabbit)]

Hrm I'm getting a build error: make[2]: *** No rule to make target `../src/Widgets/SearchViewInterface.vala'

In my head I would think User Input > LibColumbus > Zeitgeist > Search Results. I think that's what you're saying you did right?

[Tom Beckmann (tombeckmann)]

You can just remove src/Widgets/SearchViewInterface.vala to make it compile, I'll fix it once I get back to my computer, which will take a while.
Theoretically that's what should happen, yes.

[lp:~tombeckmann/+junk/libcolumbus]

409. By Tom Beckmann on 2014-03-04

remove reference to non existent file

[Tom Beckmann (tombeckmann)]

Should work fine now.

[Rico Tzschichholz (ricotz)]

Relative header-paths in vapis aren't a good idea.
Aren't there bindings for libcolumbus itself to avoid using ccode with a primitive vapi?

[lp:~tombeckmann/+junk/libcolumbus]

410. By Tom Beckmann on 2014-03-04

add real libcolumbus vapi by ricotz, improve search results

[Tom Beckmann (tombeckmann)]

Thanks to ricotz providing a vapi for libcolumbus I could drop the C code now. Also the search results should be a lot better now.

[lp:~tombeckmann/+junk/libcolumbus]

411. By Tom Beckmann on 2014-03-04

forgot most important class -.-

412. By Tom Beckmann on 2014-03-04

add some docs, remove debug print

[Danielle Foré (danrabbit)]

Hrm, I feel like the top result isn't as good as it was before.

It's pretty cool getting results even when you have typos, but when I don't make typos (such as typing "mi") I'm not getting the result I expect ("Midori" or "Minecraft")

Since typos shouldn't return results, can we list results returned from the actual typed string first and results returned from libcolumbus after those?

That way, we'll get "minecraft" listed before "terminal" (command line) when I type "mine"

review: Needs Fixing

Unmerged revisions

412. By Tom Beckmann on 2014-03-04

add some docs, remove debug print

411. By Tom Beckmann on 2014-03-04

forgot most important class -.-

410. By Tom Beckmann on 2014-03-04

add real libcolumbus vapi by ricotz, improve search results

409. By Tom Beckmann on 2014-03-04

remove reference to non existent file

408. By Tom Beckmann on 2014-03-03

add support for libcolumbus

=== modified file 'CMakeLists.txt'
--- CMakeLists.txt	2014-02-23 19:56:35 +0000
+++ CMakeLists.txt	2014-03-04 23:27:45 +0000
@@ -54,8 +54,19 @@
   message ("-- Zeitgeist integration disabled")
 endif ()
 
+option (USE_LIBCOLUMBUS "Use libcolumbus for fuzzy matching" ON)
+
+if (USE_LIBCOLUMBUS)
+    message ("-- libcolumbus integration enabled")
+    set (LIBCOLUMBUS_DEPS libcolumbus)
+    set (LIBCOLUMBUS_OPTIONS --define=HAVE_LIBCOLUMBUS --vapidir=${CMAKE_SOURCE_DIR}/vapi)
+    set (LIBCOLUMBUS_SOURCES src/Backend/ColumbusSearchModel.vala)
+else ()
+    message ("-- libcolumbus integration disabled")
+endif ()
+
 set (CORE_DEPS "gobject-2.0;glib-2.0;gio-2.0;gio-unix-2.0;gee-0.8;libgnome-menu-3.0;libwnck-3.0;gdk-x11-3.0;${UNITY_DEPS};")
-set (UI_DEPS "gtk+-3.0>=3.2.0;granite;${ZEITGEIST_DEPS};")
+set (UI_DEPS "gtk+-3.0>=3.2.0;granite;${ZEITGEIST_DEPS};${LIBCOLUMBUS_DEPS};")
 
 find_package (PkgConfig)
 pkg_check_modules (DEPS REQUIRED "${CORE_DEPS}${UI_DEPS}" gthread-2.0)
@@ -81,16 +92,19 @@
     src/Widgets/SearchItem.vala
     src/Widgets/Sidebar.vala
     src/Widgets/CategoryView.vala
+    ${LIBCOLUMBUS_SOURCES}
 PACKAGES
     ${CORE_DEPS}
     ${UI_DEPS}
 CUSTOM_VAPIS
     vapi/config.vapi
+    ${LIBCOLUMBUS_VAPI}
 OPTIONS
     --thread
     -g
     ${UNITY_OPTIONS}
     ${ZEITGEIST_OPTIONS}
+    ${LIBCOLUMBUS_OPTIONS}
 )
 
 # Comment this out to enable C compiler warnings
 
=== modified file 'src/Backend/AppSystem.vala'
--- src/Backend/AppSystem.vala	2013-01-10 16:53:50 +0000
+++ src/Backend/AppSystem.vala	2014-03-04 23:27:45 +0000
@@ -26,6 +26,10 @@
     private RelevancyService rl_service;
 #endif
 
+#if HAVE_LIBCOLUMBUS
+    private ColumbusSearch.Model search_model;
+#endif
+
     public signal void changed ();
 
     construct {
@@ -35,6 +39,10 @@
         rl_service.update_complete.connect (update_popularity);
 #endif
 
+#if HAVE_LIBCOLUMBUS
+        search_model = new ColumbusSearch.Model ();
+#endif
+
         apps_menu = new GMenu.Tree ("pantheon-applications.menu", GMenu.TreeFlags.INCLUDE_EXCLUDED | GMenu.TreeFlags.SORT_DISPLAY_NAME);
         apps_menu.changed.connect (update_app_system);
         
@@ -90,9 +98,24 @@
 
         apps.clear ();
 
-        foreach (var cat in categories)
-            apps.set (cat.get_name (), get_apps_by_category (cat));
-
+#if HAVE_LIBCOLUMBUS
+        search_model.clear ();
+#endif
+
+        foreach (var cat in categories) {
+            var category_apps = get_apps_by_category (cat);
+            apps.set (cat.get_name (), category_apps);
+#if HAVE_LIBCOLUMBUS
+            foreach (var app in category_apps) {
+                search_model.add_entry (app, app.name,
+                    app.description, app.exec);
+            }
+#endif
+        }
+
+#if HAVE_LIBCOLUMBUS
+        search_model.finished_adding ();
+#endif
     }
 
     public Gee.ArrayList<GMenu.TreeDirectory> get_categories () {
@@ -171,6 +194,12 @@
 
         var filtered = new Gee.ArrayList<App> ();
 
+#if HAVE_LIBCOLUMBUS
+        List<void*> results = search_model.match (search);
+        foreach (var result in results) {
+            filtered.add (result as App);
+        }
+#else
         /** It's a bit stupid algorithm, simply check if the char is present
          * some of the App values, then assign it a double. This is very simple:
          * if an App name coincide with the search string they have obvious the
@@ -213,6 +242,7 @@
                 }
             }
         }
+#endif
 
         filtered.sort ((a, b) => Utils.sort_apps_by_relevancy ((App) a, (App) b));
 
 
=== added file 'src/Backend/ColumbusSearchModel.vala'
--- src/Backend/ColumbusSearchModel.vala	1970-01-01 00:00:00 +0000
+++ src/Backend/ColumbusSearchModel.vala	2014-03-04 23:27:45 +0000
@@ -0,0 +1,88 @@
+
+#if HAVE_LIBCOLUMBUS
+using Columbus;
+
+namespace ColumbusSearch
+{
+    public class Model : Object {
+        Corpus corpus;
+        Matcher matcher;
+
+        Word field_name;
+        Word field_description;
+        Word field_exec;
+
+        public Model () {
+            corpus = new Corpus ();
+            matcher = new Matcher ();
+
+            field_name = new Word ("name");
+            field_description = new Word ("description");
+            field_exec = new Word ("exec");
+        }
+
+        /**
+         * Add an entry to the model.
+         *
+         * @param id          A pointer to the app which is identified by these values. It
+         *                    is internally used as an id.
+         * @param name        The app's name
+         * @param description The app's description
+         * @param exec        The app's executable line
+         */
+        public void add_entry (void *id, string name, string description, string exec) {
+            var doc = new Document ((DocumentID)id);
+
+            doc.add_text (field_name, name);
+            doc.add_text (field_description, description);
+            doc.add_text (field_exec, exec);
+
+            corpus.add_document (doc);
+        }
+
+        /**
+         * Call this function after you added all the entries to build the index
+         */
+        public void finished_adding ()
+        {
+            matcher.index (corpus);
+
+            var errors = matcher.get_error_values ();
+            errors.add_standard_errors ();
+            errors.set_substring_mode ();
+
+            var weights = matcher.get_index_weights ();
+            weights.set_weight (field_description, 0.5);
+        }
+
+        /**
+         * Clear the current model
+         */
+        public void clear () {
+            corpus = new Corpus ();
+        }
+
+        /**
+         * Search in the model. Don't forget to call finished_adding before calling this
+         * 
+         * @param query The query string to be searched for
+         * @return      Returns a list of the ids which represent pointers to the app
+         *              objects as added earlier.
+         */
+        public List<void*> match (string query) {
+            var result = matcher.match (query);
+            var len = result.size ();
+
+            var list = new List<void*> ();
+            for (var i = 0; i < len; i++) {
+                list.prepend ((void*)result.get_id (i));
+            }
+
+            list.reverse ();
+
+            return list;
+        }
+    }
+}
+
+#endif
 
=== added file 'vapi/libcolumbus.vapi'
--- vapi/libcolumbus.vapi	1970-01-01 00:00:00 +0000
+++ vapi/libcolumbus.vapi	2014-03-04 23:27:45 +0000
@@ -0,0 +1,72 @@
+/* libcolumbus.vapi generated by vapigen, do not modify. */
+
+[CCode (cprefix = "Col", lower_case_cprefix = "col_")]
+namespace Columbus {
+	[CCode (cheader_filename = "columbus.h", free_function = "col_corpus_delete")]
+	[Compact]
+	public class Corpus {
+		[CCode (has_construct_function = false, type = "ColCorpus")]
+		public Corpus ();
+		public void add_document (Columbus.Document d);
+		public void @delete ();
+	}
+	[CCode (cheader_filename = "columbus.h", free_function = "col_document_delete")]
+	[Compact]
+	public class Document {
+		[CCode (has_construct_function = false, type = "ColDocument")]
+		public Document (Columbus.DocumentID id);
+		public void add_text (Columbus.Word field_name, string text_as_utf8);
+		public void @delete ();
+		public Columbus.DocumentID get_id ();
+	}
+	[CCode (cheader_filename = "columbus.h", free_function = "")]
+	[Compact]
+	public class ErrorValues {
+		public void add_standard_errors ();
+		public void set_substring_mode ();
+	}
+	[CCode (cheader_filename = "columbus.h", free_function = "")]
+	[Compact]
+	public class IndexWeights {
+		public double get_weight (Columbus.Word field);
+		public void set_weight (Columbus.Word field, double new_weight);
+	}
+	[CCode (cheader_filename = "columbus.h", free_function = "col_match_results_delete")]
+	[Compact]
+	public class MatchResults {
+		[CCode (has_construct_function = false, type = "ColMatchResults")]
+		public MatchResults ();
+		public void @delete ();
+		public Columbus.DocumentID get_id (size_t i);
+		public double get_relevancy (size_t i);
+		public size_t size ();
+	}
+	[CCode (cheader_filename = "columbus.h", free_function = "col_matcher_delete")]
+	[Compact]
+	public class Matcher {
+		[CCode (has_construct_function = false, type = "ColMatcher")]
+		public Matcher ();
+		public void @delete ();
+		public Columbus.ErrorValues get_error_values ();
+		public Columbus.IndexWeights get_index_weights ();
+		public void index (Columbus.Corpus c);
+		public Columbus.MatchResults match (string query_as_utf8);
+	}
+	[CCode (cheader_filename = "columbus.h", free_function = "col_word_delete")]
+	[Compact]
+	public class Word {
+		[CCode (has_construct_function = false, type = "ColWord")]
+		public Word (string utf8_word);
+		public void as_utf8 (string buf, uint bufSize);
+		public void @delete ();
+		public size_t length ();
+	}
+	[CCode (cname = "uintptr_t", cheader_filename = "columbus.h", cprefix = "")]
+	[SimpleType]
+	public struct DocumentID {
+	}
+	[CCode (cname = "uint32_t", cheader_filename = "columbus.h", cprefix = "")]
+	[SimpleType]
+	public struct WordID {
+	}
+}