sort code completions with rules

Fixups

Add levenshtein distance for comparisons, remove kind sort order, try to improve as many different use cases as possible

Trying again to improve code completion

Sort code autocompletion options by similarity based on input

To make it really brief, uses a combination `String.similiary`, the category system introduced in a previous PR, and some filtering to yield more predictable results, instead of scattering every completion option at seemingly random.

It also gives much higher priority to strings that contain the base in full, closer to the beginning or are perfect matches.

Also moves CodeCompletionOptionCompare to code_edit.cpp

Co-Authored-By: Micky <66727710+Mickeon@users.noreply.github.com>
Co-Authored-By: Eric M <41730826+EricEzaM@users.noreply.github.com>

2 files changed, 58 insertions, 3 deletions

diff --git a/core/object/script_language.cpp b/core/object/script_language.cpp
index 71f40660f4..6f047d80aa 100644
--- a/core/object/script_language.cpp
+++ b/core/object/script_language.cpp
@@ -34,7 +34,6 @@
 #include "core/core_string_names.h"
 #include "core/debugger/engine_debugger.h"
 #include "core/debugger/script_debugger.h"
-#include "core/variant/typed_array.h"
 
 #include <stdint.h>
 
@@ -461,6 +460,52 @@ void ScriptLanguage::get_core_type_words(List<String> *p_core_type_words) const
 void ScriptLanguage::frame() {
 }
 
+TypedArray<int> ScriptLanguage::CodeCompletionOption::get_option_characteristics(const String &p_base) {
+	// Return characacteristics of the match found by order of importance.
+	// Matches will be ranked by a lexicographical order on the vector returned by this function.
+	// The lower values indicate better matches and that they should go before in the order of appearance.
+	if (last_matches == matches) {
+		return charac;
+	}
+	charac.clear();
+	// Ensure base is not empty and at the same time that matches is not empty too.
+	if (p_base.length() == 0) {
+		last_matches = matches;
+		charac.push_back(location);
+		return charac;
+	}
+	charac.push_back(matches.size());
+	charac.push_back((matches[0].first == 0) ? 0 : 1);
+	charac.push_back(location);
+	const char32_t *target_char = &p_base[0];
+	int bad_case = 0;
+	for (const Pair<int, int> &match_segment : matches) {
+		const char32_t *string_to_complete_char = &display[match_segment.first];
+		for (int j = 0; j < match_segment.second; j++, string_to_complete_char++, target_char++) {
+			if (*string_to_complete_char != *target_char) {
+				bad_case++;
+			}
+		}
+	}
+	charac.push_back(bad_case);
+	charac.push_back(matches[0].first);
+	last_matches = matches;
+	return charac;
+}
+
+void ScriptLanguage::CodeCompletionOption::clear_characteristics() {
+	charac = TypedArray<int>();
+}
+
+TypedArray<int> ScriptLanguage::CodeCompletionOption::get_option_cached_characteristics() const {
+	// Only returns the cached value and warns if it was not updated since the last change of matches.
+	if (last_matches != matches) {
+		WARN_PRINT("Characteristics are not up to date.");
+	}
+
+	return charac;
+}
+
 bool PlaceHolderScriptInstance::set(const StringName &p_name, const Variant &p_value) {
 	if (script->is_placeholder_fallback_enabled()) {
 		return false;
diff --git a/core/object/script_language.h b/core/object/script_language.h
index 696c9a94a5..829f01fbcc 100644
--- a/core/object/script_language.h
+++ b/core/object/script_language.h
@@ -35,6 +35,7 @@
 #include "core/io/resource.h"
 #include "core/templates/pair.h"
 #include "core/templates/rb_map.h"
+#include "core/variant/typed_array.h"
 
 class ScriptLanguage;
 template <typename T>
@@ -305,8 +306,8 @@ public:
 	virtual Error open_in_external_editor(const Ref<Script> &p_script, int p_line, int p_col) { return ERR_UNAVAILABLE; }
 	virtual bool overrides_external_editor() { return false; }
 
-	/* Keep enum in Sync with:                               */
-	/* /scene/gui/code_edit.h - CodeEdit::CodeCompletionKind */
+	// Keep enums in sync with:
+	// scene/gui/code_edit.h - CodeEdit::CodeCompletionKind
 	enum CodeCompletionKind {
 		CODE_COMPLETION_KIND_CLASS,
 		CODE_COMPLETION_KIND_FUNCTION,
@@ -321,6 +322,7 @@ public:
 		CODE_COMPLETION_KIND_MAX
 	};
 
+	// scene/gui/code_edit.h - CodeEdit::CodeCompletionLocation
 	enum CodeCompletionLocation {
 		LOCATION_LOCAL = 0,
 		LOCATION_PARENT_MASK = 1 << 8,
@@ -336,6 +338,7 @@ public:
 		Ref<Resource> icon;
 		Variant default_value;
 		Vector<Pair<int, int>> matches;
+		Vector<Pair<int, int>> last_matches;
 		int location = LOCATION_OTHER;
 
 		CodeCompletionOption() {}
@@ -346,6 +349,13 @@ public:
 			kind = p_kind;
 			location = p_location;
 		}
+
+		TypedArray<int> get_option_characteristics(const String &p_base);
+		void clear_characteristics();
+		TypedArray<int> get_option_cached_characteristics() const;
+
+	private:
+		TypedArray<int> charac;
 	};
 
 	virtual Error complete_code(const String &p_code, const String &p_path, Object *p_owner, List<CodeCompletionOption> *r_options, bool &r_force, String &r_call_hint) { return ERR_UNAVAILABLE; }