new bench results

README.md

@@ -208,17 +208,19 @@ various operations on a JSON document containing 1,000 entries.
 
 | Op\Lib | jsonpak | std/json |
 |----------|----------|----------|
-| Extract | 0.2805 | 0.7552 |
-| toString | 0.8243 | 0.6683 |
-| fromJson | 0.0033 | 0.0009 |
+| Extract | 0.2679 | 0.6892 |
+| toString | 0.8314 | 0.6853 |
+| fromJson | 0.0040 | 0.0008 |
 | toJson | 0.0007 | 0.0005 |
-| Parse | 1.5065 | 1.7205 |
-| Test | 0.0035 | 0.0005 |
-| Replace | 0.0036 | 0.0006 |
-| Remove | 0.0124 | 0.0009 |
-| Add | 0.0035 | 0.0006 |
-| Copy | 0.0120 | 0.0007 |
-| Move | 0.0212 | 0.0011 |
+| Parse | 0.9767 | 1.4560 |
+| Test | 0.0041 | 0.0005 |
+| Replace | 0.0043 | 0.0005 |
+| Remove | 0.0131 | 0.0008 |
+| Add | 0.0043 | 0.0005 |
+| Copy | 0.0086 | 0.0005 |
+| Move | 0.0179 | 0.0008 |
+| Sort | 0.8483 | 0.1953 |
+| Hash | 0.1045 | 0.1484 |
 
 However, the standard library's representation occupies approximately 13.4MiB,
 whereas ours only takes up 2.8MiB. Therefore, this library aims to optimize

bench/benchmark.nim

@@ -26,6 +26,20 @@ template bench(name, tree, code: untyped) =
  let globalDuration = cpuTime() - globalStart
  printStats(name, stats, globalDuration)
 
+import std/tables
+
+proc sort(n: JsonNode) =
+ ## Sort a JSON node
+ case n.kind
+ of JArray:
+ for e in mitems(n.elems):
+ sort(e)
+ of JObject:
+ sort(n.fields, proc (x, y: (string, JsonNode)): int = cmp[string](x[0], y[0]))
+ for k, v in mpairs(n.fields):
+ sort(v)
+ else: discard
+
 type
  UserRecord = object
  id: int
@@ -118,6 +132,9 @@ proc main() =
  t["records"][500]["location"] = t["records"][0]["city"]
  t["records"][0].delete("city")
 
+ bench "stdlib - sort", stdTree:
+ t.sort
+
  echo "used Mem: ", formatSize getOccupiedMem()
 
 main()