unique: further work

src/unique.nix

@@ -2,12 +2,43 @@ lib:
 let
   # O(n^2), assuming foldl' is O(n) and ++ causes a reallocation every time
   uniqueOld = builtins.foldl' (acc: e: if builtins.elem e acc then acc else acc ++ [ e ]) [];
-  # uniqueNew = xs: let
-  #   entries = lib.genAttrs xs (_: null);
-  # in builtins.concatMap (x: if );
 
-  bigdataMostlyDuplicates = (lib.replicate 999999 null) ++ ["unique"];
-  bigdataMostlyUnique = lib.imap0 (i: _: toString i) (lib.replicate 999999 null) ++ ["123" "456"];
+  # Not sure about the runtime of removing duplicates in attrsets and/or genericClosure,
+  # but it could theoretically be surpassed by the time it takes to create identifiers.
+  # O(a + b) where a is duplication removal runtime and b is identifier creation runtime
+  #
+  # This function performs well for large inputs with mostly unique values
+  uniqueHashable = let
+    identifier = x: {
+      "int" = "int:${toString x}";
+      "bool" = "bool:${toString x}";
+      "string" = "string:${x}";
+      "path" = "path:${toString x}";
+      "null" = "null";
+      "set" = "set:{${toString (lib.mapAttrsToList (name: value: "${name}=${identifier value},") x)}}";
+      "list" = "list:[${toString (map identifier x)}]";
+      "float" = "float:${toString x}";
+      # "lambda"
+    }.${builtins.typeOf x};
+
+  # Does not preserve order
+  # in list: builtins.attrValues (builtins.listToAttrs (map (x: {
+  #   name = identifier x;
+  #   value = x;
+  # }) list));
+
+  # Seemingly preserves order
+  in list: builtins.catAttrs "value" (builtins.genericClosure {
+    startSet = map (x: { key = identifier x; value = x; }) list;
+    operator = _: [];
+  });
+
+  bigdataCount = 9999;
+  bigdataMostlyDuplicates = (lib.replicate bigdataCount "") ++ ["unique" {a = 1;} [1 2 3]];
+  bigdataMostlyUnique = lib.imap0 (i: _: toString i) (lib.replicate bigdataCount null) ++ ["123" "456" {a = 1;} [1 2 3]];
 in {
-  old = uniqueOld bigdataMostlyDuplicates;
+  oldMostlyDuplicates = uniqueOld bigdataMostlyDuplicates;
+  oldMostlyUnique = uniqueOld bigdataMostlyUnique;
+  hashableMostlyDuplicates = uniqueHashable bigdataMostlyDuplicates;
+  hashableMostlyUnique = uniqueHashable bigdataMostlyUnique;
 }