Re #1786 Minor changes and comment to be nice and push up the build

_test/test_utilities_herbert/test_fast_map.m

@@ -11,6 +11,7 @@
             obj = obj@TestCase(name);
         end
         %------------------------------------------------------------------
+        %------------------------------------------------------------------
         function test_insertion_in_optimized(~)
             n_keys = 100;
             base_key = 10+round(rand(1,10*n_keys)*(10*n_keys-1));
@@ -46,6 +47,18 @@ function test_optimization(~)
             end
         end
         %------------------------------------------------------------------
+        function test_map_loadobj(~)
+            keys = uint32(1:10);
+            keys = num2cell(keys);
+            val = num2cell(10:-1:1);
+            fm = fast_map(keys,val );
+
+            struc = fm.to_struct();
+            rec = serializable.loadobj(struc);
+
+            assertEqual(fm,rec);
+        end
+
         function test_fast_map_accepts_addition(~)
             keys = 10:-1:1;
 

herbert_core/utilities/classes/@fast_map/fast_map.m

@@ -15,7 +15,10 @@
     %
     % WARNING: intentianally disabled multiple reliability checks and
     % convenience properties in favour of access speed.
-
+    %
+    % See fast_map_vs_map_performance in test_herbert_utilities
+    % to compare speed and optimize fast_map operations.
+    %
     properties
         % map optimization for doing fast access limit
         %
@@ -184,20 +187,17 @@
         end
         %
         function obj = optimize(obj)
-            % place values into expanded cellrarray, containing
-            % empty where keys are missing and values where
-            obj.min_max_key_val_ = min_max(obj.keys_);
-            obj.key_shif_ = obj.min_max_key_val_(1)-1;
-            n_places = obj.min_max_key_val_(2)-obj.min_max_key_val_(1)+1;
-            obj.keyval_optimized_ = nan(1,n_places);
-            keys_shifted = obj.keys_-obj.min_max_key_val_(1)+1;
-            obj.keyval_optimized_(keys_shifted) = obj.values_(:);
-            obj.optimized_ = true;
+            % place values into expanded array or cellarray, containing
+            % NaN or empty where keys are missing and values where
+            % keys are present. This array/cellarray is optimal for fast
+            % access to the values as function of keys.
+            %
+            obj = optimize_(obj);
         end
     end
     %----------------------------------------------------------------------
-    % Overloaded indexers. DESPITE NICE, adding them makes fast_map 40-60
-    % times slower even without using indexes itself. Disabled for this
+    % Overloaded indexers. DESPITE LOOKING NICE, adding them makes fast_map
+    % 40-60 times slower even without using indexes itself. Disabled for this
     % reason, until, may be mex is written which would deal with fast part
     % of indices.
     methods

herbert_core/utilities/classes/@fast_map/private/check_combo_arg_.m

@@ -1,6 +1,7 @@
 function  obj = check_combo_arg_(obj)
 %CHECK_COMBO_ARG_ validate consistency of interdependent properties, namely
-%key/value properties
+%key/value properties. If conditions right, the function also builds
+%cash for optimized access to values
 
 if numel(obj.keys_) ~= numel(obj.values_)
     error('HERBERT:fast_map:invalid_argument', ...

herbert_core/utilities/classes/@fast_map/private/optimize_.m

@@ -0,0 +1,13 @@
+function obj = optimize_(obj)
+%OPTIMIZE_  % place values into expanded array or cellarray, containing
+% NaN or empty where keys are missing and values where
+% keys are present. This array/cellarray is optimal for fast
+% access to the values as function of keys.
+
+obj.min_max_key_val_ = min_max(obj.keys_);
+obj.key_shif_ = obj.min_max_key_val_(1)-1;
+n_places = obj.min_max_key_val_(2)-obj.min_max_key_val_(1)+1;
+obj.keyval_optimized_ = nan(1,n_places);
+keys_shifted = obj.keys_-obj.min_max_key_val_(1)+1;
+obj.keyval_optimized_(keys_shifted) = obj.values_(:);
+obj.optimized_ = true;