summary refs log tree commit diff stats
path: root/lib/pure/json.nim
diff options
context:
space:
mode:
Diffstat (limited to 'lib/pure/json.nim')
-rw-r--r--lib/pure/json.nim1822
1 files changed, 1016 insertions, 806 deletions
diff --git a/lib/pure/json.nim b/lib/pure/json.nim
index c3db5bdf8..53fa7553a 100644
--- a/lib/pure/json.nim
+++ b/lib/pure/json.nim
@@ -14,514 +14,172 @@
 ## JSON is based on a subset of the JavaScript Programming Language,
 ## Standard ECMA-262 3rd Edition - December 1999.
 ##
-## Usage example:
+## See also
+## ========
+## * `std/parsejson <parsejson.html>`_
+## * `std/jsonutils <jsonutils.html>`_
+## * `std/marshal <marshal.html>`_
+## * `std/jscore <jscore.html>`_
 ##
-## .. code-block:: nim
-##  let
-##    small_json = """{"test": 1.3, "key2": true}"""
-##    jobj = parseJson(small_json)
-##  assert (jobj.kind == JObject)
-##  echo($jobj["test"].fnum)
-##  echo($jobj["key2"].bval)
 ##
-## Results in:
+## Overview
+## ========
 ##
-## .. code-block:: nim
+## Parsing JSON
+## ------------
 ##
-##   1.3000000000000000e+00
-##   true
+## JSON often arrives into your program (via an API or a file) as a `string`.
+## The first step is to change it from its serialized form into a nested object
+## structure called a `JsonNode`.
+##
+## The `parseJson` procedure takes a string containing JSON and returns a
+## `JsonNode` object. This is an object variant and it is either a
+## `JObject`, `JArray`, `JString`, `JInt`, `JFloat`, `JBool` or
+## `JNull`. You check the kind of this object variant by using the `kind`
+## accessor.
+##
+## For a `JsonNode` who's kind is `JObject`, you can access its fields using
+## the `[]` operator. The following example shows how to do this:
+##
+##   ```Nim
+##   import std/json
+##
+##   let jsonNode = parseJson("""{"key": 3.14}""")
+##
+##   doAssert jsonNode.kind == JObject
+##   doAssert jsonNode["key"].kind == JFloat
+##   ```
+##
+## Reading values
+## --------------
+##
+## Once you have a `JsonNode`, retrieving the values can then be achieved
+## by using one of the helper procedures, which include:
+##
+## * `getInt`
+## * `getFloat`
+## * `getStr`
+## * `getBool`
+##
+## To retrieve the value of `"key"` you can do the following:
+##
+##   ```Nim
+##   import std/json
+##
+##   let jsonNode = parseJson("""{"key": 3.14}""")
+##
+##   doAssert jsonNode["key"].getFloat() == 3.14
+##   ```
+##
+## **Important:** The `[]` operator will raise an exception when the
+## specified field does not exist.
+##
+## Handling optional keys
+## ----------------------
+##
+## By using the `{}` operator instead of `[]`, it will return `nil`
+## when the field is not found. The `get`-family of procedures will return a
+## type's default value when called on `nil`.
+##
+##   ```Nim
+##   import std/json
+##
+##   let jsonNode = parseJson("{}")
+##
+##   doAssert jsonNode{"nope"}.getInt() == 0
+##   doAssert jsonNode{"nope"}.getFloat() == 0
+##   doAssert jsonNode{"nope"}.getStr() == ""
+##   doAssert jsonNode{"nope"}.getBool() == false
+##   ```
+##
+## Using default values
+## --------------------
+##
+## The `get`-family helpers also accept an additional parameter which allow
+## you to fallback to a default value should the key's values be `null`:
+##
+##   ```Nim
+##   import std/json
+##
+##   let jsonNode = parseJson("""{"key": 3.14, "key2": null}""")
+##
+##   doAssert jsonNode["key"].getFloat(6.28) == 3.14
+##   doAssert jsonNode["key2"].getFloat(3.14) == 3.14
+##   doAssert jsonNode{"nope"}.getFloat(3.14) == 3.14 # note the {}
+##   ```
+##
+## Unmarshalling
+## -------------
+##
+## In addition to reading dynamic data, Nim can also unmarshal JSON directly
+## into a type with the `to` macro.
+##
+## Note: Use `Option <options.html#Option>`_ for keys sometimes missing in json
+## responses, and backticks around keys with a reserved keyword as name.
+##
+##   ```Nim
+##   import std/json
+##   import std/options
+##
+##   type
+##     User = object
+##       name: string
+##       age: int
+##       `type`: Option[string]
+##
+##   let userJson = parseJson("""{ "name": "Nim", "age": 12 }""")
+##   let user = to(userJson, User)
+##   if user.`type`.isSome():
+##     assert user.`type`.get() != "robot"
+##   ```
+##
+## Creating JSON
+## =============
 ##
 ## This module can also be used to comfortably create JSON using the `%*`
 ## operator:
 ##
-## .. code-block:: nim
+##   ```nim
+##   import std/json
 ##
 ##   var hisName = "John"
 ##   let herAge = 31
 ##   var j = %*
 ##     [
-##       {
-##         "name": hisName,
-##         "age": 30
-##       },
-##       {
-##         "name": "Susan",
-##         "age": herAge
-##       }
+##       { "name": hisName, "age": 30 },
+##       { "name": "Susan", "age": herAge }
 ##     ]
+##
+##   var j2 = %* {"name": "Isaac", "books": ["Robot Dreams"]}
+##   j2["details"] = %* {"age":35, "pi":3.1415}
+##   echo j2
+##   ```
+##
+## See also: std/jsonutils for hookable json serialization/deserialization
+## of arbitrary types.
 
-import
-  hashes, strutils, lexbase, streams, unicode, macros
+runnableExamples:
+  ## Note: for JObject, key ordering is preserved, unlike in some languages,
+  ## this is convenient for some use cases. Example:
+  type Foo = object
+    a1, a2, a0, a3, a4: int
+  doAssert $(%* Foo()) == """{"a1":0,"a2":0,"a0":0,"a3":0,"a4":0}"""
 
-type
-  JsonEventKind* = enum  ## enumeration of all events that may occur when parsing
-    jsonError,           ## an error occurred during parsing
-    jsonEof,             ## end of file reached
-    jsonString,          ## a string literal
-    jsonInt,             ## an integer literal
-    jsonFloat,           ## a float literal
-    jsonTrue,            ## the value ``true``
-    jsonFalse,           ## the value ``false``
-    jsonNull,            ## the value ``null``
-    jsonObjectStart,     ## start of an object: the ``{`` token
-    jsonObjectEnd,       ## end of an object: the ``}`` token
-    jsonArrayStart,      ## start of an array: the ``[`` token
-    jsonArrayEnd         ## start of an array: the ``]`` token
-
-  TTokKind = enum        # must be synchronized with TJsonEventKind!
-    tkError,
-    tkEof,
-    tkString,
-    tkInt,
-    tkFloat,
-    tkTrue,
-    tkFalse,
-    tkNull,
-    tkCurlyLe,
-    tkCurlyRi,
-    tkBracketLe,
-    tkBracketRi,
-    tkColon,
-    tkComma
-
-  JsonError* = enum        ## enumeration that lists all errors that can occur
-    errNone,               ## no error
-    errInvalidToken,       ## invalid token
-    errStringExpected,     ## string expected
-    errColonExpected,      ## ``:`` expected
-    errCommaExpected,      ## ``,`` expected
-    errBracketRiExpected,  ## ``]`` expected
-    errCurlyRiExpected,    ## ``}`` expected
-    errQuoteExpected,      ## ``"`` or ``'`` expected
-    errEOC_Expected,       ## ``*/`` expected
-    errEofExpected,        ## EOF expected
-    errExprExpected        ## expr expected
-
-  ParserState = enum
-    stateEof, stateStart, stateObject, stateArray, stateExpectArrayComma,
-    stateExpectObjectComma, stateExpectColon, stateExpectValue
-
-  JsonParser* = object of BaseLexer ## the parser object.
-    a: string
-    tok: TTokKind
-    kind: JsonEventKind
-    err: JsonError
-    state: seq[ParserState]
-    filename: string
-
-{.deprecated: [TJsonEventKind: JsonEventKind, TJsonError: JsonError,
-  TJsonParser: JsonParser].}
-
-const
-  errorMessages: array [JsonError, string] = [
-    "no error",
-    "invalid token",
-    "string expected",
-    "':' expected",
-    "',' expected",
-    "']' expected",
-    "'}' expected",
-    "'\"' or \"'\" expected",
-    "'*/' expected",
-    "EOF expected",
-    "expression expected"
-  ]
-  tokToStr: array [TTokKind, string] = [
-    "invalid token",
-    "EOF",
-    "string literal",
-    "int literal",
-    "float literal",
-    "true",
-    "false",
-    "null",
-    "{", "}", "[", "]", ":", ","
-  ]
-
-proc open*(my: var JsonParser, input: Stream, filename: string) =
-  ## initializes the parser with an input stream. `Filename` is only used
-  ## for nice error messages.
-  lexbase.open(my, input)
-  my.filename = filename
-  my.state = @[stateStart]
-  my.kind = jsonError
-  my.a = ""
-
-proc close*(my: var JsonParser) {.inline.} =
-  ## closes the parser `my` and its associated input stream.
-  lexbase.close(my)
-
-proc str*(my: JsonParser): string {.inline.} =
-  ## returns the character data for the events: ``jsonInt``, ``jsonFloat``,
-  ## ``jsonString``
-  assert(my.kind in {jsonInt, jsonFloat, jsonString})
-  return my.a
-
-proc getInt*(my: JsonParser): BiggestInt {.inline.} =
-  ## returns the number for the event: ``jsonInt``
-  assert(my.kind == jsonInt)
-  return parseBiggestInt(my.a)
-
-proc getFloat*(my: JsonParser): float {.inline.} =
-  ## returns the number for the event: ``jsonFloat``
-  assert(my.kind == jsonFloat)
-  return parseFloat(my.a)
-
-proc kind*(my: JsonParser): JsonEventKind {.inline.} =
-  ## returns the current event type for the JSON parser
-  return my.kind
-
-proc getColumn*(my: JsonParser): int {.inline.} =
-  ## get the current column the parser has arrived at.
-  result = getColNumber(my, my.bufpos)
-
-proc getLine*(my: JsonParser): int {.inline.} =
-  ## get the current line the parser has arrived at.
-  result = my.lineNumber
-
-proc getFilename*(my: JsonParser): string {.inline.} =
-  ## get the filename of the file that the parser processes.
-  result = my.filename
-
-proc errorMsg*(my: JsonParser): string =
-  ## returns a helpful error message for the event ``jsonError``
-  assert(my.kind == jsonError)
-  result = "$1($2, $3) Error: $4" % [
-    my.filename, $getLine(my), $getColumn(my), errorMessages[my.err]]
-
-proc errorMsgExpected*(my: JsonParser, e: string): string =
-  ## returns an error message "`e` expected" in the same format as the
-  ## other error messages
-  result = "$1($2, $3) Error: $4" % [
-    my.filename, $getLine(my), $getColumn(my), e & " expected"]
-
-proc handleHexChar(c: char, x: var int): bool =
-  result = true # Success
-  case c
-  of '0'..'9': x = (x shl 4) or (ord(c) - ord('0'))
-  of 'a'..'f': x = (x shl 4) or (ord(c) - ord('a') + 10)
-  of 'A'..'F': x = (x shl 4) or (ord(c) - ord('A') + 10)
-  else: result = false # error
-
-proc parseString(my: var JsonParser): TTokKind =
-  result = tkString
-  var pos = my.bufpos + 1
-  var buf = my.buf
-  while true:
-    case buf[pos]
-    of '\0':
-      my.err = errQuoteExpected
-      result = tkError
-      break
-    of '"':
-      inc(pos)
-      break
-    of '\\':
-      case buf[pos+1]
-      of '\\', '"', '\'', '/':
-        add(my.a, buf[pos+1])
-        inc(pos, 2)
-      of 'b':
-        add(my.a, '\b')
-        inc(pos, 2)
-      of 'f':
-        add(my.a, '\f')
-        inc(pos, 2)
-      of 'n':
-        add(my.a, '\L')
-        inc(pos, 2)
-      of 'r':
-        add(my.a, '\C')
-        inc(pos, 2)
-      of 't':
-        add(my.a, '\t')
-        inc(pos, 2)
-      of 'u':
-        inc(pos, 2)
-        var r: int
-        if handleHexChar(buf[pos], r): inc(pos)
-        if handleHexChar(buf[pos], r): inc(pos)
-        if handleHexChar(buf[pos], r): inc(pos)
-        if handleHexChar(buf[pos], r): inc(pos)
-        add(my.a, toUTF8(Rune(r)))
-      else:
-        # don't bother with the error
-        add(my.a, buf[pos])
-        inc(pos)
-    of '\c':
-      pos = lexbase.handleCR(my, pos)
-      buf = my.buf
-      add(my.a, '\c')
-    of '\L':
-      pos = lexbase.handleLF(my, pos)
-      buf = my.buf
-      add(my.a, '\L')
-    else:
-      add(my.a, buf[pos])
-      inc(pos)
-  my.bufpos = pos # store back
+import std/[hashes, tables, strutils, lexbase, streams, macros, parsejson]
 
-proc skip(my: var JsonParser) =
-  var pos = my.bufpos
-  var buf = my.buf
-  while true:
-    case buf[pos]
-    of '/':
-      if buf[pos+1] == '/':
-        # skip line comment:
-        inc(pos, 2)
-        while true:
-          case buf[pos]
-          of '\0':
-            break
-          of '\c':
-            pos = lexbase.handleCR(my, pos)
-            buf = my.buf
-            break
-          of '\L':
-            pos = lexbase.handleLF(my, pos)
-            buf = my.buf
-            break
-          else:
-            inc(pos)
-      elif buf[pos+1] == '*':
-        # skip long comment:
-        inc(pos, 2)
-        while true:
-          case buf[pos]
-          of '\0':
-            my.err = errEOC_Expected
-            break
-          of '\c':
-            pos = lexbase.handleCR(my, pos)
-            buf = my.buf
-          of '\L':
-            pos = lexbase.handleLF(my, pos)
-            buf = my.buf
-          of '*':
-            inc(pos)
-            if buf[pos] == '/':
-              inc(pos)
-              break
-          else:
-            inc(pos)
-      else:
-        break
-    of ' ', '\t':
-      inc(pos)
-    of '\c':
-      pos = lexbase.handleCR(my, pos)
-      buf = my.buf
-    of '\L':
-      pos = lexbase.handleLF(my, pos)
-      buf = my.buf
-    else:
-      break
-  my.bufpos = pos
-
-proc parseNumber(my: var JsonParser) =
-  var pos = my.bufpos
-  var buf = my.buf
-  if buf[pos] == '-':
-    add(my.a, '-')
-    inc(pos)
-  if buf[pos] == '.':
-    add(my.a, "0.")
-    inc(pos)
-  else:
-    while buf[pos] in Digits:
-      add(my.a, buf[pos])
-      inc(pos)
-    if buf[pos] == '.':
-      add(my.a, '.')
-      inc(pos)
-  # digits after the dot:
-  while buf[pos] in Digits:
-    add(my.a, buf[pos])
-    inc(pos)
-  if buf[pos] in {'E', 'e'}:
-    add(my.a, buf[pos])
-    inc(pos)
-    if buf[pos] in {'+', '-'}:
-      add(my.a, buf[pos])
-      inc(pos)
-    while buf[pos] in Digits:
-      add(my.a, buf[pos])
-      inc(pos)
-  my.bufpos = pos
-
-proc parseName(my: var JsonParser) =
-  var pos = my.bufpos
-  var buf = my.buf
-  if buf[pos] in IdentStartChars:
-    while buf[pos] in IdentChars:
-      add(my.a, buf[pos])
-      inc(pos)
-  my.bufpos = pos
-
-proc getTok(my: var JsonParser): TTokKind =
-  setLen(my.a, 0)
-  skip(my) # skip whitespace, comments
-  case my.buf[my.bufpos]
-  of '-', '.', '0'..'9':
-    parseNumber(my)
-    if {'.', 'e', 'E'} in my.a:
-      result = tkFloat
-    else:
-      result = tkInt
-  of '"':
-    result = parseString(my)
-  of '[':
-    inc(my.bufpos)
-    result = tkBracketLe
-  of '{':
-    inc(my.bufpos)
-    result = tkCurlyLe
-  of ']':
-    inc(my.bufpos)
-    result = tkBracketRi
-  of '}':
-    inc(my.bufpos)
-    result = tkCurlyRi
-  of ',':
-    inc(my.bufpos)
-    result = tkComma
-  of ':':
-    inc(my.bufpos)
-    result = tkColon
-  of '\0':
-    result = tkEof
-  of 'a'..'z', 'A'..'Z', '_':
-    parseName(my)
-    case my.a
-    of "null": result = tkNull
-    of "true": result = tkTrue
-    of "false": result = tkFalse
-    else: result = tkError
-  else:
-    inc(my.bufpos)
-    result = tkError
-  my.tok = result
-
-proc next*(my: var JsonParser) =
-  ## retrieves the first/next event. This controls the parser.
-  var tk = getTok(my)
-  var i = my.state.len-1
-  # the following code is a state machine. If we had proper coroutines,
-  # the code could be much simpler.
-  case my.state[i]
-  of stateEof:
-    if tk == tkEof:
-      my.kind = jsonEof
-    else:
-      my.kind = jsonError
-      my.err = errEofExpected
-  of stateStart:
-    # tokens allowed?
-    case tk
-    of tkString, tkInt, tkFloat, tkTrue, tkFalse, tkNull:
-      my.state[i] = stateEof # expect EOF next!
-      my.kind = JsonEventKind(ord(tk))
-    of tkBracketLe:
-      my.state.add(stateArray) # we expect any
-      my.kind = jsonArrayStart
-    of tkCurlyLe:
-      my.state.add(stateObject)
-      my.kind = jsonObjectStart
-    of tkEof:
-      my.kind = jsonEof
-    else:
-      my.kind = jsonError
-      my.err = errEofExpected
-  of stateObject:
-    case tk
-    of tkString, tkInt, tkFloat, tkTrue, tkFalse, tkNull:
-      my.state.add(stateExpectColon)
-      my.kind = JsonEventKind(ord(tk))
-    of tkBracketLe:
-      my.state.add(stateExpectColon)
-      my.state.add(stateArray)
-      my.kind = jsonArrayStart
-    of tkCurlyLe:
-      my.state.add(stateExpectColon)
-      my.state.add(stateObject)
-      my.kind = jsonObjectStart
-    of tkCurlyRi:
-      my.kind = jsonObjectEnd
-      discard my.state.pop()
-    else:
-      my.kind = jsonError
-      my.err = errCurlyRiExpected
-  of stateArray:
-    case tk
-    of tkString, tkInt, tkFloat, tkTrue, tkFalse, tkNull:
-      my.state.add(stateExpectArrayComma) # expect value next!
-      my.kind = JsonEventKind(ord(tk))
-    of tkBracketLe:
-      my.state.add(stateExpectArrayComma)
-      my.state.add(stateArray)
-      my.kind = jsonArrayStart
-    of tkCurlyLe:
-      my.state.add(stateExpectArrayComma)
-      my.state.add(stateObject)
-      my.kind = jsonObjectStart
-    of tkBracketRi:
-      my.kind = jsonArrayEnd
-      discard my.state.pop()
-    else:
-      my.kind = jsonError
-      my.err = errBracketRiExpected
-  of stateExpectArrayComma:
-    case tk
-    of tkComma:
-      discard my.state.pop()
-      next(my)
-    of tkBracketRi:
-      my.kind = jsonArrayEnd
-      discard my.state.pop() # pop stateExpectArrayComma
-      discard my.state.pop() # pop stateArray
-    else:
-      my.kind = jsonError
-      my.err = errBracketRiExpected
-  of stateExpectObjectComma:
-    case tk
-    of tkComma:
-      discard my.state.pop()
-      next(my)
-    of tkCurlyRi:
-      my.kind = jsonObjectEnd
-      discard my.state.pop() # pop stateExpectObjectComma
-      discard my.state.pop() # pop stateObject
-    else:
-      my.kind = jsonError
-      my.err = errCurlyRiExpected
-  of stateExpectColon:
-    case tk
-    of tkColon:
-      my.state[i] = stateExpectValue
-      next(my)
-    else:
-      my.kind = jsonError
-      my.err = errColonExpected
-  of stateExpectValue:
-    case tk
-    of tkString, tkInt, tkFloat, tkTrue, tkFalse, tkNull:
-      my.state[i] = stateExpectObjectComma
-      my.kind = JsonEventKind(ord(tk))
-    of tkBracketLe:
-      my.state[i] = stateExpectObjectComma
-      my.state.add(stateArray)
-      my.kind = jsonArrayStart
-    of tkCurlyLe:
-      my.state[i] = stateExpectObjectComma
-      my.state.add(stateObject)
-      my.kind = jsonObjectStart
-    else:
-      my.kind = jsonError
-      my.err = errExprExpected
+import std/options # xxx remove this dependency using same approach as https://github.com/nim-lang/Nim/pull/14563
+import std/private/since
+
+when defined(nimPreviewSlimSystem):
+  import std/[syncio, assertions, formatfloat]
 
+export
+  tables.`$`
 
-# ------------- higher level interface ---------------------------------------
+export
+  parsejson.JsonEventKind, parsejson.JsonError, JsonParser, JsonKindError,
+  open, close, str, getInt, getFloat, kind, getColumn, getLine, getFilename,
+  errorMsg, errorMsgExpected, next, JsonParsingError, raiseParseErr, nimIdentNormalize
 
 type
   JsonNodeKind* = enum ## possible JSON node types
@@ -535,6 +193,8 @@ type
 
   JsonNode* = ref JsonNodeObj ## JSON node
   JsonNodeObj* {.acyclic.} = object
+    isUnquoted: bool # the JString was a number-like token and
+                     # so shouldn't be quoted
     case kind*: JsonNodeKind
     of JString:
       str*: string
@@ -547,127 +207,244 @@ type
     of JNull:
       nil
     of JObject:
-      fields*: seq[tuple[key: string, val: JsonNode]]
+      fields*: OrderedTable[string, JsonNode]
     of JArray:
       elems*: seq[JsonNode]
 
-  JsonParsingError* = object of ValueError ## is raised for a JSON error
-
-{.deprecated: [EJsonParsingError: JsonParsingError, TJsonNode: JsonNodeObj,
-    PJsonNode: JsonNode, TJsonNodeKind: JsonNodeKind].}
-
-proc raiseParseErr*(p: JsonParser, msg: string) {.noinline, noreturn.} =
-  ## raises an `EJsonParsingError` exception.
-  raise newException(JsonParsingError, errorMsgExpected(p, msg))
+const DepthLimit = 1000
 
 proc newJString*(s: string): JsonNode =
   ## Creates a new `JString JsonNode`.
-  new(result)
-  result.kind = JString
-  result.str = s
+  result = JsonNode(kind: JString, str: s)
 
-proc newJStringMove(s: string): JsonNode =
-  new(result)
-  result.kind = JString
-  shallowCopy(result.str, s)
+proc newJRawNumber(s: string): JsonNode =
+  ## Creates a "raw JS number", that is a number that does not
+  ## fit into Nim's `BiggestInt` field. This is really a `JString`
+  ## with the additional information that it should be converted back
+  ## to the string representation without the quotes.
+  result = JsonNode(kind: JString, str: s, isUnquoted: true)
 
 proc newJInt*(n: BiggestInt): JsonNode =
   ## Creates a new `JInt JsonNode`.
-  new(result)
-  result.kind = JInt
-  result.num  = n
+  result = JsonNode(kind: JInt, num: n)
 
 proc newJFloat*(n: float): JsonNode =
   ## Creates a new `JFloat JsonNode`.
-  new(result)
-  result.kind = JFloat
-  result.fnum  = n
+  result = JsonNode(kind: JFloat, fnum: n)
 
 proc newJBool*(b: bool): JsonNode =
   ## Creates a new `JBool JsonNode`.
-  new(result)
-  result.kind = JBool
-  result.bval = b
+  result = JsonNode(kind: JBool, bval: b)
 
 proc newJNull*(): JsonNode =
   ## Creates a new `JNull JsonNode`.
-  new(result)
+  result = JsonNode(kind: JNull)
 
 proc newJObject*(): JsonNode =
   ## Creates a new `JObject JsonNode`
-  new(result)
-  result.kind = JObject
-  result.fields = @[]
+  result = JsonNode(kind: JObject, fields: initOrderedTable[string, JsonNode](2))
 
 proc newJArray*(): JsonNode =
   ## Creates a new `JArray JsonNode`
-  new(result)
-  result.kind = JArray
-  result.elems = @[]
+  result = JsonNode(kind: JArray, elems: @[])
+
+proc getStr*(n: JsonNode, default: string = ""): string =
+  ## Retrieves the string value of a `JString JsonNode`.
+  ##
+  ## Returns `default` if `n` is not a `JString`, or if `n` is nil.
+  if n.isNil or n.kind != JString: return default
+  else: return n.str
+
+proc getInt*(n: JsonNode, default: int = 0): int =
+  ## Retrieves the int value of a `JInt JsonNode`.
+  ##
+  ## Returns `default` if `n` is not a `JInt`, or if `n` is nil.
+  if n.isNil or n.kind != JInt: return default
+  else: return int(n.num)
+
+proc getBiggestInt*(n: JsonNode, default: BiggestInt = 0): BiggestInt =
+  ## Retrieves the BiggestInt value of a `JInt JsonNode`.
+  ##
+  ## Returns `default` if `n` is not a `JInt`, or if `n` is nil.
+  if n.isNil or n.kind != JInt: return default
+  else: return n.num
+
+proc getFloat*(n: JsonNode, default: float = 0.0): float =
+  ## Retrieves the float value of a `JFloat JsonNode`.
+  ##
+  ## Returns `default` if `n` is not a `JFloat` or `JInt`, or if `n` is nil.
+  if n.isNil: return default
+  case n.kind
+  of JFloat: return n.fnum
+  of JInt: return float(n.num)
+  else: return default
+
+proc getBool*(n: JsonNode, default: bool = false): bool =
+  ## Retrieves the bool value of a `JBool JsonNode`.
+  ##
+  ## Returns `default` if `n` is not a `JBool`, or if `n` is nil.
+  if n.isNil or n.kind != JBool: return default
+  else: return n.bval
+
+proc getFields*(n: JsonNode,
+    default = initOrderedTable[string, JsonNode](2)):
+        OrderedTable[string, JsonNode] =
+  ## Retrieves the key, value pairs of a `JObject JsonNode`.
+  ##
+  ## Returns `default` if `n` is not a `JObject`, or if `n` is nil.
+  if n.isNil or n.kind != JObject: return default
+  else: return n.fields
+
+proc getElems*(n: JsonNode, default: seq[JsonNode] = @[]): seq[JsonNode] =
+  ## Retrieves the array of a `JArray JsonNode`.
+  ##
+  ## Returns `default` if `n` is not a `JArray`, or if `n` is nil.
+  if n.isNil or n.kind != JArray: return default
+  else: return n.elems
 
+proc add*(father, child: JsonNode) =
+  ## Adds `child` to a JArray node `father`.
+  assert father.kind == JArray
+  father.elems.add(child)
+
+proc add*(obj: JsonNode, key: string, val: JsonNode) =
+  ## Sets a field from a `JObject`.
+  assert obj.kind == JObject
+  obj.fields[key] = val
 
 proc `%`*(s: string): JsonNode =
   ## Generic constructor for JSON data. Creates a new `JString JsonNode`.
-  new(result)
-  result.kind = JString
-  result.str = s
+  result = JsonNode(kind: JString, str: s)
+
+proc `%`*(n: uint): JsonNode =
+  ## Generic constructor for JSON data. Creates a new `JInt JsonNode`.
+  if n > cast[uint](int.high):
+    result = newJRawNumber($n)
+  else:
+    result = JsonNode(kind: JInt, num: BiggestInt(n))
+
+proc `%`*(n: int): JsonNode =
+  ## Generic constructor for JSON data. Creates a new `JInt JsonNode`.
+  result = JsonNode(kind: JInt, num: n)
+
+proc `%`*(n: BiggestUInt): JsonNode =
+  ## Generic constructor for JSON data. Creates a new `JInt JsonNode`.
+  if n > cast[BiggestUInt](BiggestInt.high):
+    result = newJRawNumber($n)
+  else:
+    result = JsonNode(kind: JInt, num: BiggestInt(n))
 
 proc `%`*(n: BiggestInt): JsonNode =
   ## Generic constructor for JSON data. Creates a new `JInt JsonNode`.
-  new(result)
-  result.kind = JInt
-  result.num  = n
+  result = JsonNode(kind: JInt, num: n)
 
 proc `%`*(n: float): JsonNode =
   ## Generic constructor for JSON data. Creates a new `JFloat JsonNode`.
-  new(result)
-  result.kind = JFloat
-  result.fnum  = n
+  runnableExamples:
+    assert $(%[NaN, Inf, -Inf, 0.0, -0.0, 1.0, 1e-2]) == """["nan","inf","-inf",0.0,-0.0,1.0,0.01]"""
+    assert (%NaN).kind == JString
+    assert (%0.0).kind == JFloat
+  # for those special cases, we could also have used `newJRawNumber` but then
+  # it would've been inconsisten with the case of `parseJson` vs `%` for representing them.
+  if n != n: newJString("nan")
+  elif n == Inf: newJString("inf")
+  elif n == -Inf: newJString("-inf")
+  else: JsonNode(kind: JFloat, fnum: n)
 
 proc `%`*(b: bool): JsonNode =
   ## Generic constructor for JSON data. Creates a new `JBool JsonNode`.
-  new(result)
-  result.kind = JBool
-  result.bval = b
+  result = JsonNode(kind: JBool, bval: b)
 
 proc `%`*(keyVals: openArray[tuple[key: string, val: JsonNode]]): JsonNode =
   ## Generic constructor for JSON data. Creates a new `JObject JsonNode`
-  new(result)
-  result.kind = JObject
-  newSeq(result.fields, keyVals.len)
-  for i, p in pairs(keyVals): result.fields[i] = p
+  if keyVals.len == 0: return newJArray()
+  result = newJObject()
+  for key, val in items(keyVals): result.fields[key] = val
+
+template `%`*(j: JsonNode): JsonNode = j
 
-proc `%`*(elements: openArray[JsonNode]): JsonNode =
+proc `%`*[T](elements: openArray[T]): JsonNode =
   ## Generic constructor for JSON data. Creates a new `JArray JsonNode`
-  new(result)
-  result.kind = JArray
-  newSeq(result.elems, elements.len)
-  for i, p in pairs(elements): result.elems[i] = p
+  result = newJArray()
+  for elem in elements: result.add(%elem)
 
-proc toJson(x: NimNode): NimNode {.compiletime.} =
-  case x.kind
-  of nnkBracket:
-    result = newNimNode(nnkBracket)
-    for i in 0 .. <x.len:
-      result.add(toJson(x[i]))
+proc `%`*[T](table: Table[string, T]|OrderedTable[string, T]): JsonNode =
+  ## Generic constructor for JSON data. Creates a new `JObject JsonNode`.
+  result = newJObject()
+  for k, v in table: result[k] = %v
 
-  of nnkTableConstr:
-    result = newNimNode(nnkTableConstr)
-    for i in 0 .. <x.len:
-      assert x[i].kind == nnkExprColonExpr
-      result.add(newNimNode(nnkExprColonExpr).add(x[i][0]).add(toJson(x[i][1])))
+proc `%`*[T](opt: Option[T]): JsonNode =
+  ## Generic constructor for JSON data. Creates a new `JNull JsonNode`
+  ## if `opt` is empty, otherwise it delegates to the underlying value.
+  if opt.isSome: %opt.get else: newJNull()
 
+when false:
+  # For 'consistency' we could do this, but that only pushes people further
+  # into that evil comfort zone where they can use Nim without understanding it
+  # causing problems later on.
+  proc `%`*(elements: set[bool]): JsonNode =
+    ## Generic constructor for JSON data. Creates a new `JObject JsonNode`.
+    ## This can only be used with the empty set `{}` and is supported
+    ## to prevent the gotcha `%*{}` which used to produce an empty
+    ## JSON array.
+    result = newJObject()
+    assert false notin elements, "usage error: only empty sets allowed"
+    assert true notin elements, "usage error: only empty sets allowed"
+
+proc `[]=`*(obj: JsonNode, key: string, val: JsonNode) {.inline.} =
+  ## Sets a field from a `JObject`.
+  assert(obj.kind == JObject)
+  obj.fields[key] = val
+
+proc `%`*[T: object](o: T): JsonNode =
+  ## Construct JsonNode from tuples and objects.
+  result = newJObject()
+  for k, v in o.fieldPairs: result[k] = %v
+
+proc `%`*(o: ref object): JsonNode =
+  ## Generic constructor for JSON data. Creates a new `JObject JsonNode`
+  if o.isNil:
+    result = newJNull()
   else:
-    result = x
+    result = %(o[])
+
+proc `%`*(o: enum): JsonNode =
+  ## Construct a JsonNode that represents the specified enum value as a
+  ## string. Creates a new `JString JsonNode`.
+  result = %($o)
 
-  result = prefix(result, "%")
+proc toJsonImpl(x: NimNode): NimNode =
+  case x.kind
+  of nnkBracket: # array
+    if x.len == 0: return newCall(bindSym"newJArray")
+    result = newNimNode(nnkBracket)
+    for i in 0 ..< x.len:
+      result.add(toJsonImpl(x[i]))
+    result = newCall(bindSym("%", brOpen), result)
+  of nnkTableConstr: # object
+    if x.len == 0: return newCall(bindSym"newJObject")
+    result = newNimNode(nnkTableConstr)
+    for i in 0 ..< x.len:
+      x[i].expectKind nnkExprColonExpr
+      result.add newTree(nnkExprColonExpr, x[i][0], toJsonImpl(x[i][1]))
+    result = newCall(bindSym("%", brOpen), result)
+  of nnkCurly: # empty object
+    x.expectLen(0)
+    result = newCall(bindSym"newJObject")
+  of nnkNilLit:
+    result = newCall(bindSym"newJNull")
+  of nnkPar:
+    if x.len == 1: result = toJsonImpl(x[0])
+    else: result = newCall(bindSym("%", brOpen), x)
+  else:
+    result = newCall(bindSym("%", brOpen), x)
 
-macro `%*`*(x: expr): expr =
+macro `%*`*(x: untyped): untyped =
   ## Convert an expression to a JsonNode directly, without having to specify
   ## `%` for every element.
-  result = toJson(x)
+  result = toJsonImpl(x)
 
-proc `==`* (a,b: JsonNode): bool =
+proc `==`*(a, b: JsonNode): bool {.noSideEffect, raises: [].} =
   ## Check two nodes for equality
   if a.isNil:
     if b.isNil: return true
@@ -675,23 +452,37 @@ proc `==`* (a,b: JsonNode): bool =
   elif b.isNil or a.kind != b.kind:
     return false
   else:
-    return case a.kind
+    case a.kind
     of JString:
-      a.str == b.str
+      result = a.str == b.str
     of JInt:
-      a.num == b.num
+      result = a.num == b.num
     of JFloat:
-      a.fnum == b.fnum
+      result = a.fnum == b.fnum
     of JBool:
-      a.bval == b.bval
+      result = a.bval == b.bval
     of JNull:
-      true
+      result = true
     of JArray:
-      a.elems == b.elems
+      {.cast(raises: []).}: # bug #19303
+        result = a.elems == b.elems
     of JObject:
-      a.fields == b.fields
+      # we cannot use OrderedTable's equality here as
+      # the order does not matter for equality here.
+      if a.fields.len != b.fields.len: return false
+      for key, val in a.fields:
+        if not b.fields.hasKey(key): return false
+        {.cast(raises: []).}:
+          when defined(nimHasEffectsOf):
+            {.noSideEffect.}:
+              if b.fields[key] != val: return false
+          else:
+            if b.fields[key] != val: return false
+      result = true
+
+proc hash*(n: OrderedTable[string, JsonNode]): Hash {.noSideEffect.}
 
-proc hash* (n:JsonNode): THash =
+proc hash*(n: JsonNode): Hash {.noSideEffect.} =
   ## Compute the hash for a JSON node
   case n.kind
   of JArray:
@@ -707,7 +498,12 @@ proc hash* (n:JsonNode): THash =
   of JString:
     result = hash(n.str)
   of JNull:
-    result = hash(0)
+    result = Hash(0)
+
+proc hash*(n: OrderedTable[string, JsonNode]): Hash =
+  for key, val in n:
+    result = result xor (hash(key) !& hash(val))
+  result = !$result
 
 proc len*(n: JsonNode): int =
   ## If `n` is a `JArray`, it returns the number of elements.
@@ -718,84 +514,131 @@ proc len*(n: JsonNode): int =
   of JObject: result = n.fields.len
   else: discard
 
-proc `[]`*(node: JsonNode, name: string): JsonNode =
+proc `[]`*(node: JsonNode, name: string): JsonNode {.inline.} =
   ## Gets a field from a `JObject`, which must not be nil.
-  ## If the value at `name` does not exist, returns nil
+  ## If the value at `name` does not exist, raises KeyError.
   assert(not isNil(node))
   assert(node.kind == JObject)
-  for key, item in items(node.fields):
-    if key == name:
-      return item
-  return nil
+  when defined(nimJsonGet):
+    if not node.fields.hasKey(name): return nil
+  result = node.fields[name]
 
-proc `[]`*(node: JsonNode, index: int): JsonNode =
+proc `[]`*(node: JsonNode, index: int): JsonNode {.inline.} =
   ## Gets the node at `index` in an Array. Result is undefined if `index`
-  ## is out of bounds
+  ## is out of bounds, but as long as array bound checks are enabled it will
+  ## result in an exception.
   assert(not isNil(node))
   assert(node.kind == JArray)
   return node.elems[index]
 
+proc `[]`*(node: JsonNode, index: BackwardsIndex): JsonNode {.inline, since: (1, 5, 1).} =
+  ## Gets the node at `array.len-i` in an array through the `^` operator.
+  ##
+  ## i.e. `j[^i]` is a shortcut for `j[j.len-i]`.
+  runnableExamples:
+    let
+      j = parseJson("[1,2,3,4,5]")
+
+    doAssert j[^1].getInt == 5
+    doAssert j[^2].getInt == 4
+
+  `[]`(node, node.len - int(index))
+
+proc `[]`*[U, V](a: JsonNode, x: HSlice[U, V]): JsonNode =
+  ## Slice operation for JArray.
+  ##
+  ## Returns the inclusive range `[a[x.a], a[x.b]]`:
+  runnableExamples:
+    import std/json
+    let arr = %[0,1,2,3,4,5]
+    doAssert arr[2..4] == %[2,3,4]
+    doAssert arr[2..^2] == %[2,3,4]
+    doAssert arr[^4..^2] == %[2,3,4]
+
+  assert(a.kind == JArray)
+  result = newJArray()
+  let xa = (when x.a is BackwardsIndex: a.len - int(x.a) else: int(x.a))
+  let L = (when x.b is BackwardsIndex: a.len - int(x.b) else: int(x.b)) - xa + 1
+  for i in 0..<L:
+    result.add(a[i + xa])
+
 proc hasKey*(node: JsonNode, key: string): bool =
   ## Checks if `key` exists in `node`.
   assert(node.kind == JObject)
-  for k, item in items(node.fields):
-    if k == key: return true
+  result = node.fields.hasKey(key)
 
-proc existsKey*(node: JsonNode, key: string): bool {.deprecated.} = node.hasKey(key)
-  ## Deprecated for `hasKey`
+proc contains*(node: JsonNode, key: string): bool =
+  ## Checks if `key` exists in `node`.
+  assert(node.kind == JObject)
+  node.fields.hasKey(key)
 
-proc add*(father, child: JsonNode) =
-  ## Adds `child` to a JArray node `father`.
-  assert father.kind == JArray
-  father.elems.add(child)
+proc contains*(node: JsonNode, val: JsonNode): bool =
+  ## Checks if `val` exists in array `node`.
+  assert(node.kind == JArray)
+  find(node.elems, val) >= 0
+
+proc `{}`*(node: JsonNode, keys: varargs[string]): JsonNode =
+  ## Traverses the node and gets the given value. If any of the
+  ## keys do not exist, returns `nil`. Also returns `nil` if one of the
+  ## intermediate data structures is not an object.
+  ##
+  ## This proc can be used to create tree structures on the
+  ## fly (sometimes called `autovivification`:idx:):
+  ##
+  runnableExamples:
+    var myjson = %* {"parent": {"child": {"grandchild": 1}}}
+    doAssert myjson{"parent", "child", "grandchild"} == newJInt(1)
 
-proc add*(obj: JsonNode, key: string, val: JsonNode) =
-  ## Adds ``(key, val)`` pair to the JObject node `obj`. For speed
-  ## reasons no check for duplicate keys is performed!
-  ## But ``[]=`` performs the check.
-  assert obj.kind == JObject
-  obj.fields.add((key, val))
+  result = node
+  for key in keys:
+    if isNil(result) or result.kind != JObject:
+      return nil
+    result = result.fields.getOrDefault(key)
+
+proc `{}`*(node: JsonNode, index: varargs[int]): JsonNode =
+  ## Traverses the node and gets the given value. If any of the
+  ## indexes do not exist, returns `nil`. Also returns `nil` if one of the
+  ## intermediate data structures is not an array.
+  result = node
+  for i in index:
+    if isNil(result) or result.kind != JArray or i >= node.len:
+      return nil
+    result = result.elems[i]
 
-proc `[]=`*(obj: JsonNode, key: string, val: JsonNode) =
-  ## Sets a field from a `JObject`. Performs a check for duplicate keys.
-  assert(obj.kind == JObject)
-  for i in 0..obj.fields.len-1:
-    if obj.fields[i].key == key:
-      obj.fields[i].val = val
-      return
-  obj.fields.add((key, val))
+proc getOrDefault*(node: JsonNode, key: string): JsonNode =
+  ## Gets a field from a `node`. If `node` is nil or not an object or
+  ## value at `key` does not exist, returns nil
+  if not isNil(node) and node.kind == JObject:
+    result = node.fields.getOrDefault(key)
 
 proc `{}`*(node: JsonNode, key: string): JsonNode =
-  ## Transverses the node and gets the given value. If any of the
-  ## names does not exist, returns nil
-  result = node
-  if isNil(node): return nil
-  result = result[key]
+  ## Gets a field from a `node`. If `node` is nil or not an object or
+  ## value at `key` does not exist, returns nil
+  node.getOrDefault(key)
 
-proc `{}=`*(node: JsonNode, names: varargs[string], value: JsonNode) =
-  ## Transverses the node and tries to set the value at the given location
-  ## to `value` If any of the names are missing, they are added
+proc `{}=`*(node: JsonNode, keys: varargs[string], value: JsonNode) =
+  ## Traverses the node and tries to set the value at the given location
+  ## to `value`. If any of the keys are missing, they are added.
   var node = node
-  for i in 0..(names.len-2):
-    if isNil(node[names[i]]):
-      node[names[i]] = newJObject()
-    node = node[names[i]]
-  node[names[names.len-1]] = value
+  for i in 0..(keys.len-2):
+    if not node.hasKey(keys[i]):
+      node[keys[i]] = newJObject()
+    node = node[keys[i]]
+  node[keys[keys.len-1]] = value
 
 proc delete*(obj: JsonNode, key: string) =
-  ## Deletes ``obj[key]`` preserving the order of the other (key, value)-pairs.
+  ## Deletes `obj[key]`.
   assert(obj.kind == JObject)
-  for i in 0..obj.fields.len-1:
-    if obj.fields[i].key == key:
-      obj.fields.delete(i)
-      return
-  raise newException(IndexError, "key not in object")
+  if not obj.fields.hasKey(key):
+    raise newException(KeyError, "key not in object")
+  obj.fields.del(key)
 
 proc copy*(p: JsonNode): JsonNode =
-  ## Performs a deep copy of `a`.
+  ## Performs a deep copy of `p`.
   case p.kind
   of JString:
     result = newJString(p.str)
+    result.isUnquoted = p.isUnquoted
   of JInt:
     result = newJInt(p.num)
   of JFloat:
@@ -806,8 +649,8 @@ proc copy*(p: JsonNode): JsonNode =
     result = newJNull()
   of JObject:
     result = newJObject()
-    for key, field in items(p.fields):
-      result.fields.add((key, copy(field)))
+    for key, val in pairs(p.fields):
+      result.fields[key] = copy(val)
   of JArray:
     result = newJArray()
     for i in items(p.elems):
@@ -823,43 +666,102 @@ proc newIndent(curr, indent: int, ml: bool): int =
   else: return indent
 
 proc nl(s: var string, ml: bool) =
-  if ml: s.add("\n")
+  s.add(if ml: "\n" else: " ")
+
+proc escapeJsonUnquoted*(s: string; result: var string) =
+  ## Converts a string `s` to its JSON representation without quotes.
+  ## Appends to `result`.
+  for c in s:
+    case c
+    of '\L': result.add("\\n")
+    of '\b': result.add("\\b")
+    of '\f': result.add("\\f")
+    of '\t': result.add("\\t")
+    of '\v': result.add("\\u000b")
+    of '\r': result.add("\\r")
+    of '"': result.add("\\\"")
+    of '\0'..'\7': result.add("\\u000" & $ord(c))
+    of '\14'..'\31': result.add("\\u00" & toHex(ord(c), 2))
+    of '\\': result.add("\\\\")
+    else: result.add(c)
+
+proc escapeJsonUnquoted*(s: string): string =
+  ## Converts a string `s` to its JSON representation without quotes.
+  result = newStringOfCap(s.len + s.len shr 3)
+  escapeJsonUnquoted(s, result)
+
+proc escapeJson*(s: string; result: var string) =
+  ## Converts a string `s` to its JSON representation with quotes.
+  ## Appends to `result`.
+  result.add("\"")
+  escapeJsonUnquoted(s, result)
+  result.add("\"")
 
 proc escapeJson*(s: string): string =
-  ## Converts a string `s` to its JSON representation.
+  ## Converts a string `s` to its JSON representation with quotes.
   result = newStringOfCap(s.len + s.len shr 3)
-  result.add("\"")
-  for x in runes(s):
-    var r = int(x)
-    if r >= 32 and r <= 127:
-      var c = chr(r)
-      case c
-      of '"': result.add("\\\"")
-      of '\\': result.add("\\\\")
-      else: result.add(c)
+  escapeJson(s, result)
+
+proc toUgly*(result: var string, node: JsonNode) =
+  ## Converts `node` to its JSON Representation, without
+  ## regard for human readability. Meant to improve `$` string
+  ## conversion performance.
+  ##
+  ## JSON representation is stored in the passed `result`
+  ##
+  ## This provides higher efficiency than the `pretty` procedure as it
+  ## does **not** attempt to format the resulting JSON to make it human readable.
+  var comma = false
+  case node.kind:
+  of JArray:
+    result.add "["
+    for child in node.elems:
+      if comma: result.add ","
+      else: comma = true
+      result.toUgly child
+    result.add "]"
+  of JObject:
+    result.add "{"
+    for key, value in pairs(node.fields):
+      if comma: result.add ","
+      else: comma = true
+      key.escapeJson(result)
+      result.add ":"
+      result.toUgly value
+    result.add "}"
+  of JString:
+    if node.isUnquoted:
+      result.add node.str
     else:
-      result.add("\\u")
-      result.add(toHex(r, 4))
-  result.add("\"")
+      escapeJson(node.str, result)
+  of JInt:
+    result.addInt(node.num)
+  of JFloat:
+    result.addFloat(node.fnum)
+  of JBool:
+    result.add(if node.bval: "true" else: "false")
+  of JNull:
+    result.add "null"
 
 proc toPretty(result: var string, node: JsonNode, indent = 2, ml = true,
               lstArr = false, currIndent = 0) =
   case node.kind
   of JObject:
-    if currIndent != 0 and not lstArr: result.nl(ml)
-    result.indent(currIndent) # Indentation
+    if lstArr: result.indent(currIndent) # Indentation
     if node.fields.len > 0:
       result.add("{")
       result.nl(ml) # New line
-      for i in 0..len(node.fields)-1:
+      var i = 0
+      for key, val in pairs(node.fields):
         if i > 0:
-          result.add(", ")
+          result.add(",")
           result.nl(ml) # New Line
+        inc i
         # Need to indent more than {
         result.indent(newIndent(currIndent, indent, ml))
-        result.add(escapeJson(node.fields[i].key))
+        escapeJson(key, result)
         result.add(": ")
-        toPretty(result, node.fields[i].val, indent, ml, false,
+        toPretty(result, val, indent, ml, false,
                  newIndent(currIndent, indent, ml))
       result.nl(ml)
       result.indent(currIndent) # indent the same as {
@@ -868,16 +770,16 @@ proc toPretty(result: var string, node: JsonNode, indent = 2, ml = true,
       result.add("{}")
   of JString:
     if lstArr: result.indent(currIndent)
-    result.add(escapeJson(node.str))
+    toUgly(result, node)
   of JInt:
     if lstArr: result.indent(currIndent)
-    result.add($node.num)
+    result.addInt(node.num)
   of JFloat:
     if lstArr: result.indent(currIndent)
-    result.add($node.fnum)
+    result.addFloat(node.fnum)
   of JBool:
     if lstArr: result.indent(currIndent)
-    result.add($node.bval)
+    result.add(if node.bval: "true" else: "false")
   of JArray:
     if lstArr: result.indent(currIndent)
     if len(node.elems) != 0:
@@ -885,7 +787,7 @@ proc toPretty(result: var string, node: JsonNode, indent = 2, ml = true,
       result.nl(ml)
       for i in 0..len(node.elems)-1:
         if i > 0:
-          result.add(", ")
+          result.add(",")
           result.nl(ml) # New Line
         toPretty(result, node.elems[i], indent, ml,
             true, newIndent(currIndent, indent, ml))
@@ -898,59 +800,93 @@ proc toPretty(result: var string, node: JsonNode, indent = 2, ml = true,
     result.add("null")
 
 proc pretty*(node: JsonNode, indent = 2): string =
-  ## Converts `node` to its JSON Representation, with indentation and
+  ## Returns a JSON Representation of `node`, with indentation and
   ## on multiple lines.
+  ##
+  ## Similar to prettyprint in Python.
+  runnableExamples:
+    let j = %* {"name": "Isaac", "books": ["Robot Dreams"],
+                "details": {"age": 35, "pi": 3.1415}}
+    doAssert pretty(j) == """
+{
+  "name": "Isaac",
+  "books": [
+    "Robot Dreams"
+  ],
+  "details": {
+    "age": 35,
+    "pi": 3.1415
+  }
+}"""
   result = ""
   toPretty(result, node, indent)
 
 proc `$`*(node: JsonNode): string =
   ## Converts `node` to its JSON Representation on one line.
-  result = ""
-  toPretty(result, node, 0, false)
+  result = newStringOfCap(node.len shl 1)
+  toUgly(result, node)
 
 iterator items*(node: JsonNode): JsonNode =
   ## Iterator for the items of `node`. `node` has to be a JArray.
-  assert node.kind == JArray
+  assert node.kind == JArray, ": items() can not iterate a JsonNode of kind " & $node.kind
   for i in items(node.elems):
     yield i
 
 iterator mitems*(node: var JsonNode): var JsonNode =
   ## Iterator for the items of `node`. `node` has to be a JArray. Items can be
   ## modified.
-  assert node.kind == JArray
+  assert node.kind == JArray, ": mitems() can not iterate a JsonNode of kind " & $node.kind
   for i in mitems(node.elems):
     yield i
 
 iterator pairs*(node: JsonNode): tuple[key: string, val: JsonNode] =
   ## Iterator for the child elements of `node`. `node` has to be a JObject.
-  assert node.kind == JObject
-  for key, val in items(node.fields):
+  assert node.kind == JObject, ": pairs() can not iterate a JsonNode of kind " & $node.kind
+  for key, val in pairs(node.fields):
     yield (key, val)
 
-iterator mpairs*(node: var JsonNode): var tuple[key: string, val: JsonNode] =
-  ## Iterator for the child elements of `node`. `node` has to be a JObject.
-  ## Items can be modified
-  assert node.kind == JObject
-  for keyVal in mitems(node.fields):
-    yield keyVal
+iterator keys*(node: JsonNode): string =
+  ## Iterator for the keys in `node`. `node` has to be a JObject.
+  assert node.kind == JObject, ": keys() can not iterate a JsonNode of kind " & $node.kind
+  for key in node.fields.keys:
+    yield key
 
-proc eat(p: var JsonParser, tok: TTokKind) =
-  if p.tok == tok: discard getTok(p)
-  else: raiseParseErr(p, tokToStr[tok])
+iterator mpairs*(node: var JsonNode): tuple[key: string, val: var JsonNode] =
+  ## Iterator for the child elements of `node`. `node` has to be a JObject.
+  ## Values can be modified
+  assert node.kind == JObject, ": mpairs() can not iterate a JsonNode of kind " & $node.kind
+  for key, val in mpairs(node.fields):
+    yield (key, val)
 
-proc parseJson(p: var JsonParser): JsonNode =
+proc parseJson(p: var JsonParser; rawIntegers, rawFloats: bool, depth = 0): JsonNode =
   ## Parses JSON from a JSON Parser `p`.
   case p.tok
   of tkString:
     # we capture 'p.a' here, so we need to give it a fresh buffer afterwards:
-    result = newJStringMove(p.a)
-    p.a = ""
+    when defined(gcArc) or defined(gcOrc) or defined(gcAtomicArc):
+      result = JsonNode(kind: JString, str: move p.a)
+    else:
+      result = JsonNode(kind: JString)
+      shallowCopy(result.str, p.a)
+      p.a = ""
     discard getTok(p)
   of tkInt:
-    result = newJInt(parseBiggestInt(p.a))
+    if rawIntegers:
+      result = newJRawNumber(p.a)
+    else:
+      try:
+        result = newJInt(parseBiggestInt(p.a))
+      except ValueError:
+        result = newJRawNumber(p.a)
     discard getTok(p)
   of tkFloat:
-    result = newJFloat(parseFloat(p.a))
+    if rawFloats:
+      result = newJRawNumber(p.a)
+    else:
+      try:
+        result = newJFloat(parseFloat(p.a))
+      except ValueError:
+        result = newJRawNumber(p.a)
     discard getTok(p)
   of tkTrue:
     result = newJBool(true)
@@ -962,66 +898,91 @@ proc parseJson(p: var JsonParser): JsonNode =
     result = newJNull()
     discard getTok(p)
   of tkCurlyLe:
+    if depth > DepthLimit:
+      raiseParseErr(p, "}")
     result = newJObject()
     discard getTok(p)
     while p.tok != tkCurlyRi:
       if p.tok != tkString:
-        raiseParseErr(p, "string literal as key expected")
+        raiseParseErr(p, "string literal as key")
       var key = p.a
       discard getTok(p)
       eat(p, tkColon)
-      var val = parseJson(p)
+      var val = parseJson(p, rawIntegers, rawFloats, depth+1)
       result[key] = val
       if p.tok != tkComma: break
       discard getTok(p)
     eat(p, tkCurlyRi)
   of tkBracketLe:
+    if depth > DepthLimit:
+      raiseParseErr(p, "]")
     result = newJArray()
     discard getTok(p)
     while p.tok != tkBracketRi:
-      result.add(parseJson(p))
+      result.add(parseJson(p, rawIntegers, rawFloats, depth+1))
       if p.tok != tkComma: break
       discard getTok(p)
     eat(p, tkBracketRi)
   of tkError, tkCurlyRi, tkBracketRi, tkColon, tkComma, tkEof:
     raiseParseErr(p, "{")
 
-when not defined(js):
-  proc parseJson*(s: Stream, filename: string): JsonNode =
-    ## Parses from a stream `s` into a `JsonNode`. `filename` is only needed
-    ## for nice error messages.
-    var p: JsonParser
-    p.open(s, filename)
+iterator parseJsonFragments*(s: Stream, filename: string = ""; rawIntegers = false, rawFloats = false): JsonNode =
+  ## Parses from a stream `s` into `JsonNodes`. `filename` is only needed
+  ## for nice error messages.
+  ## The JSON fragments are separated by whitespace. This can be substantially
+  ## faster than the comparable loop
+  ## `for x in splitWhitespace(s): yield parseJson(x)`.
+  ## This closes the stream `s` after it's done.
+  ## If `rawIntegers` is true, integer literals will not be converted to a `JInt`
+  ## field but kept as raw numbers via `JString`.
+  ## If `rawFloats` is true, floating point literals will not be converted to a `JFloat`
+  ## field but kept as raw numbers via `JString`.
+  var p: JsonParser
+  p.open(s, filename)
+  try:
+    discard getTok(p) # read first token
+    while p.tok != tkEof:
+      yield p.parseJson(rawIntegers, rawFloats)
+  finally:
+    p.close()
+
+proc parseJson*(s: Stream, filename: string = ""; rawIntegers = false, rawFloats = false): JsonNode =
+  ## Parses from a stream `s` into a `JsonNode`. `filename` is only needed
+  ## for nice error messages.
+  ## If `s` contains extra data, it will raise `JsonParsingError`.
+  ## This closes the stream `s` after it's done.
+  ## If `rawIntegers` is true, integer literals will not be converted to a `JInt`
+  ## field but kept as raw numbers via `JString`.
+  ## If `rawFloats` is true, floating point literals will not be converted to a `JFloat`
+  ## field but kept as raw numbers via `JString`.
+  var p: JsonParser
+  p.open(s, filename)
+  try:
     discard getTok(p) # read first token
-    result = p.parseJson()
+    result = p.parseJson(rawIntegers, rawFloats)
+    eat(p, tkEof) # check if there is no extra data
+  finally:
     p.close()
 
-  proc parseJson*(buffer: string): JsonNode =
-    ## Parses JSON from `buffer`.
-    result = parseJson(newStringStream(buffer), "input")
+when defined(js):
+  from std/math import `mod`
+  from std/jsffi import JsObject, `[]`, to
+  from std/private/jsutils import getProtoName, isInteger, isSafeInteger
 
-  proc parseFile*(filename: string): JsonNode =
-    ## Parses `file` into a `JsonNode`.
-    var stream = newFileStream(filename, fmRead)
-    if stream == nil:
-      raise newException(IOError, "cannot read from file: " & filename)
-    result = parseJson(stream, filename)
-else:
-  from math import `mod`
-  type
-    TJSObject = object
-  proc parseNativeJson(x: cstring): TJSObject {.importc: "JSON.parse".}
+  proc parseNativeJson(x: cstring): JsObject {.importjs: "JSON.parse(#)".}
 
-  proc getVarType(x): JsonNodeKind =
+  proc getVarType(x: JsObject, isRawNumber: var bool): JsonNodeKind =
     result = JNull
-    proc getProtoName(y): cstring
-      {.importc: "Object.prototype.toString.call".}
     case $getProtoName(x) # TODO: Implicit returns fail here.
     of "[object Array]": return JArray
     of "[object Object]": return JObject
     of "[object Number]":
-      if cast[float](x) mod 1.0 == 0:
-        return JInt
+      if isInteger(x) and 1.0 / cast[float](x) != -Inf: # preserve -0.0 as float
+        if isSafeInteger(x):
+          return JInt
+        else:
+          isRawNumber = true
+          return JString
       else:
         return JFloat
     of "[object Boolean]": return JBool
@@ -1029,56 +990,384 @@ else:
     of "[object String]": return JString
     else: assert false
 
-  proc len(x: TJSObject): int =
-    assert x.getVarType == JArray
-    asm """
-      return `x`.length;
-    """
-
-  proc `[]`(x: TJSObject, y: string): TJSObject =
-    assert x.getVarType == JObject
-    asm """
-      return `x`[`y`];
-    """
-
-  proc `[]`(x: TJSObject, y: int): TJSObject =
-    assert x.getVarType == JArray
-    asm """
-      return `x`[`y`];
-    """
-
-  proc convertObject(x: TJSObject): JsonNode =
-    case getVarType(x)
+  proc len(x: JsObject): int =
+    {.emit: """
+      `result` = `x`.length;
+    """.}
+
+  proc convertObject(x: JsObject): JsonNode =
+    var isRawNumber = false
+    case getVarType(x, isRawNumber)
     of JArray:
       result = newJArray()
-      for i in 0 .. <x.len:
+      for i in 0 ..< x.len:
         result.add(x[i].convertObject())
     of JObject:
       result = newJObject()
-      asm """for (property in `x`) {
+      {.emit: """for (var property in `x`) {
         if (`x`.hasOwnProperty(property)) {
-      """
+      """.}
+
       var nimProperty: cstring
-      var nimValue: TJSObject
-      asm "`nimProperty` = property; `nimValue` = `x`[property];"
+      var nimValue: JsObject
+      {.emit: "`nimProperty` = property; `nimValue` = `x`[property];".}
       result[$nimProperty] = nimValue.convertObject()
-      asm "}}"
+      {.emit: "}}".}
     of JInt:
-      result = newJInt(cast[int](x))
+      result = newJInt(x.to(int))
     of JFloat:
-      result = newJFloat(cast[float](x))
+      result = newJFloat(x.to(float))
     of JString:
-      result = newJString($cast[cstring](x))
+      # Dunno what to do with isUnquoted here
+      if isRawNumber:
+        var value: cstring
+        {.emit: "`value` = `x`.toString();".}
+        result = newJRawNumber($value)
+      else:
+        result = newJString($x.to(cstring))
     of JBool:
-      result = newJBool(cast[bool](x))
+      result = newJBool(x.to(bool))
     of JNull:
       result = newJNull()
 
   proc parseJson*(buffer: string): JsonNode =
-    return parseNativeJson(buffer).convertObject()
+    when nimvm:
+      return parseJson(newStringStream(buffer), "input")
+    else:
+      return parseNativeJson(buffer).convertObject()
+
+else:
+  proc parseJson*(buffer: string; rawIntegers = false, rawFloats = false): JsonNode =
+    ## Parses JSON from `buffer`.
+    ## If `buffer` contains extra data, it will raise `JsonParsingError`.
+    ## If `rawIntegers` is true, integer literals will not be converted to a `JInt`
+    ## field but kept as raw numbers via `JString`.
+    ## If `rawFloats` is true, floating point literals will not be converted to a `JFloat`
+    ## field but kept as raw numbers via `JString`.
+    result = parseJson(newStringStream(buffer), "input", rawIntegers, rawFloats)
+
+  proc parseFile*(filename: string): JsonNode =
+    ## Parses `file` into a `JsonNode`.
+    ## If `file` contains extra data, it will raise `JsonParsingError`.
+    var stream = newFileStream(filename, fmRead)
+    if stream == nil:
+      raise newException(IOError, "cannot read from file: " & filename)
+    result = parseJson(stream, filename, rawIntegers=false, rawFloats=false)
+
+# -- Json deserialiser. --
+
+template verifyJsonKind(node: JsonNode, kinds: set[JsonNodeKind],
+                        ast: string) =
+  if node == nil:
+    raise newException(KeyError, "key not found: " & ast)
+  elif  node.kind notin kinds:
+    let msg = "Incorrect JSON kind. Wanted '$1' in '$2' but got '$3'." % [
+      $kinds,
+      ast,
+      $node.kind
+    ]
+    raise newException(JsonKindError, msg)
+
+macro isRefSkipDistinct*(arg: typed): untyped =
+  ## internal only, do not use
+  var impl = getTypeImpl(arg)
+  if impl.kind == nnkBracketExpr and impl[0].eqIdent("typeDesc"):
+    impl = getTypeImpl(impl[1])
+  while impl.kind == nnkDistinctTy:
+    impl = getTypeImpl(impl[0])
+  result = newLit(impl.kind == nnkRefTy)
+
+# The following forward declarations don't work in older versions of Nim
+
+# forward declare all initFromJson
+
+proc initFromJson(dst: var string; jsonNode: JsonNode; jsonPath: var string)
+proc initFromJson(dst: var bool; jsonNode: JsonNode; jsonPath: var string)
+proc initFromJson(dst: var JsonNode; jsonNode: JsonNode; jsonPath: var string)
+proc initFromJson[T: SomeInteger](dst: var T; jsonNode: JsonNode, jsonPath: var string)
+proc initFromJson[T: SomeFloat](dst: var T; jsonNode: JsonNode; jsonPath: var string)
+proc initFromJson[T: enum](dst: var T; jsonNode: JsonNode; jsonPath: var string)
+proc initFromJson[T](dst: var seq[T]; jsonNode: JsonNode; jsonPath: var string)
+proc initFromJson[S, T](dst: var array[S, T]; jsonNode: JsonNode; jsonPath: var string)
+proc initFromJson[T](dst: var Table[string, T]; jsonNode: JsonNode; jsonPath: var string)
+proc initFromJson[T](dst: var OrderedTable[string, T]; jsonNode: JsonNode; jsonPath: var string)
+proc initFromJson[T](dst: var ref T; jsonNode: JsonNode; jsonPath: var string)
+proc initFromJson[T](dst: var Option[T]; jsonNode: JsonNode; jsonPath: var string)
+proc initFromJson[T: distinct](dst: var T; jsonNode: JsonNode; jsonPath: var string)
+proc initFromJson[T: object|tuple](dst: var T; jsonNode: JsonNode; jsonPath: var string)
+
+# initFromJson definitions
+
+proc initFromJson(dst: var string; jsonNode: JsonNode; jsonPath: var string) =
+  verifyJsonKind(jsonNode, {JString, JNull}, jsonPath)
+  # since strings don't have a nil state anymore, this mapping of
+  # JNull to the default string is questionable. `none(string)` and
+  # `some("")` have the same potentional json value `JNull`.
+  if jsonNode.kind == JNull:
+    dst = ""
+  else:
+    dst = jsonNode.str
+
+proc initFromJson(dst: var bool; jsonNode: JsonNode; jsonPath: var string) =
+  verifyJsonKind(jsonNode, {JBool}, jsonPath)
+  dst = jsonNode.bval
+
+proc initFromJson(dst: var JsonNode; jsonNode: JsonNode; jsonPath: var string) =
+  if jsonNode == nil:
+    raise newException(KeyError, "key not found: " & jsonPath)
+  dst = jsonNode.copy
+
+proc initFromJson[T: SomeInteger](dst: var T; jsonNode: JsonNode, jsonPath: var string) =
+  when T is uint|uint64 or int.sizeof == 4:
+    verifyJsonKind(jsonNode, {JInt, JString}, jsonPath)
+    case jsonNode.kind
+    of JString:
+      let x = parseBiggestUInt(jsonNode.str)
+      dst = cast[T](x)
+    else:
+      dst = T(jsonNode.num)
+  else:
+    verifyJsonKind(jsonNode, {JInt}, jsonPath)
+    dst = cast[T](jsonNode.num)
+
+proc initFromJson[T: SomeFloat](dst: var T; jsonNode: JsonNode; jsonPath: var string) =
+  verifyJsonKind(jsonNode, {JInt, JFloat, JString}, jsonPath)
+  if jsonNode.kind == JString:
+    case jsonNode.str
+    of "nan":
+      let b = NaN
+      dst = T(b)
+      # dst = NaN # would fail some tests because range conversions would cause CT error
+      # in some cases; but this is not a hot-spot inside this branch and backend can optimize this.
+    of "inf":
+      let b = Inf
+      dst = T(b)
+    of "-inf":
+      let b = -Inf
+      dst = T(b)
+    else: raise newException(JsonKindError, "expected 'nan|inf|-inf', got " & jsonNode.str)
+  else:
+    if jsonNode.kind == JFloat:
+      dst = T(jsonNode.fnum)
+    else:
+      dst = T(jsonNode.num)
+
+proc initFromJson[T: enum](dst: var T; jsonNode: JsonNode; jsonPath: var string) =
+  verifyJsonKind(jsonNode, {JString}, jsonPath)
+  dst = parseEnum[T](jsonNode.getStr)
+
+proc initFromJson[T](dst: var seq[T]; jsonNode: JsonNode; jsonPath: var string) =
+  verifyJsonKind(jsonNode, {JArray}, jsonPath)
+  dst.setLen jsonNode.len
+  let orignalJsonPathLen = jsonPath.len
+  for i in 0 ..< jsonNode.len:
+    jsonPath.add '['
+    jsonPath.addInt i
+    jsonPath.add ']'
+    initFromJson(dst[i], jsonNode[i], jsonPath)
+    jsonPath.setLen orignalJsonPathLen
+
+proc initFromJson[S,T](dst: var array[S,T]; jsonNode: JsonNode; jsonPath: var string) =
+  verifyJsonKind(jsonNode, {JArray}, jsonPath)
+  let originalJsonPathLen = jsonPath.len
+  for i in 0 ..< jsonNode.len:
+    jsonPath.add '['
+    jsonPath.addInt i
+    jsonPath.add ']'
+    initFromJson(dst[i.S], jsonNode[i], jsonPath) # `.S` for enum indexed arrays
+    jsonPath.setLen originalJsonPathLen
+
+proc initFromJson[T](dst: var Table[string,T]; jsonNode: JsonNode; jsonPath: var string) =
+  dst = initTable[string, T]()
+  verifyJsonKind(jsonNode, {JObject}, jsonPath)
+  let originalJsonPathLen = jsonPath.len
+  for key in keys(jsonNode.fields):
+    jsonPath.add '.'
+    jsonPath.add key
+    initFromJson(mgetOrPut(dst, key, default(T)), jsonNode[key], jsonPath)
+    jsonPath.setLen originalJsonPathLen
+
+proc initFromJson[T](dst: var OrderedTable[string,T]; jsonNode: JsonNode; jsonPath: var string) =
+  dst = initOrderedTable[string,T]()
+  verifyJsonKind(jsonNode, {JObject}, jsonPath)
+  let originalJsonPathLen = jsonPath.len
+  for key in keys(jsonNode.fields):
+    jsonPath.add '.'
+    jsonPath.add key
+    initFromJson(mgetOrPut(dst, key, default(T)), jsonNode[key], jsonPath)
+    jsonPath.setLen originalJsonPathLen
+
+proc initFromJson[T](dst: var ref T; jsonNode: JsonNode; jsonPath: var string) =
+  verifyJsonKind(jsonNode, {JObject, JNull}, jsonPath)
+  if jsonNode.kind == JNull:
+    dst = nil
+  else:
+    dst = new(T)
+    initFromJson(dst[], jsonNode, jsonPath)
+
+proc initFromJson[T](dst: var Option[T]; jsonNode: JsonNode; jsonPath: var string) =
+  if jsonNode != nil and jsonNode.kind != JNull:
+    when T is ref:
+      dst = some(new(T))
+    else:
+      dst = some(default(T))
+    initFromJson(dst.get, jsonNode, jsonPath)
+
+macro assignDistinctImpl[T: distinct](dst: var T;jsonNode: JsonNode; jsonPath: var string) =
+  let typInst = getTypeInst(dst)
+  let typImpl = getTypeImpl(dst)
+  let baseTyp = typImpl[0]
+
+  result = quote do:
+    initFromJson(`baseTyp`(`dst`), `jsonNode`, `jsonPath`)
+
+proc initFromJson[T: distinct](dst: var T; jsonNode: JsonNode; jsonPath: var string) =
+  assignDistinctImpl(dst, jsonNode, jsonPath)
+
+proc detectIncompatibleType(typeExpr, lineinfoNode: NimNode) =
+  if typeExpr.kind == nnkTupleConstr:
+    error("Use a named tuple instead of: " & typeExpr.repr, lineinfoNode)
+
+proc foldObjectBody(dst, typeNode, tmpSym, jsonNode, jsonPath, originalJsonPathLen: NimNode) =
+  case typeNode.kind
+  of nnkEmpty:
+    discard
+  of nnkRecList, nnkTupleTy:
+    for it in typeNode:
+      foldObjectBody(dst, it, tmpSym, jsonNode, jsonPath, originalJsonPathLen)
+
+  of nnkIdentDefs:
+    typeNode.expectLen 3
+    let fieldSym = typeNode[0]
+    let fieldNameLit = newLit(fieldSym.strVal)
+    let fieldPathLit = newLit("." & fieldSym.strVal)
+    let fieldType = typeNode[1]
+
+    # Detecting incompatiple tuple types in `assignObjectImpl` only
+    # would be much cleaner, but the ast for tuple types does not
+    # contain usable type information.
+    detectIncompatibleType(fieldType, fieldSym)
+
+    dst.add quote do:
+      jsonPath.add `fieldPathLit`
+      when nimvm:
+        when isRefSkipDistinct(`tmpSym`.`fieldSym`):
+          # workaround #12489
+          var tmp: `fieldType`
+          initFromJson(tmp, getOrDefault(`jsonNode`,`fieldNameLit`), `jsonPath`)
+          `tmpSym`.`fieldSym` = tmp
+        else:
+          initFromJson(`tmpSym`.`fieldSym`, getOrDefault(`jsonNode`,`fieldNameLit`), `jsonPath`)
+      else:
+        initFromJson(`tmpSym`.`fieldSym`, getOrDefault(`jsonNode`,`fieldNameLit`), `jsonPath`)
+      jsonPath.setLen `originalJsonPathLen`
+
+  of nnkRecCase:
+    let kindSym = typeNode[0][0]
+    let kindNameLit = newLit(kindSym.strVal)
+    let kindPathLit = newLit("." & kindSym.strVal)
+    let kindType = typeNode[0][1]
+    let kindOffsetLit = newLit(uint(getOffset(kindSym)))
+    dst.add quote do:
+      var kindTmp: `kindType`
+      jsonPath.add `kindPathLit`
+      initFromJson(kindTmp, `jsonNode`[`kindNameLit`], `jsonPath`)
+      jsonPath.setLen `originalJsonPathLen`
+      when defined js:
+        `tmpSym`.`kindSym` = kindTmp
+      else:
+        when nimvm:
+          `tmpSym`.`kindSym` = kindTmp
+        else:
+          # fuck it, assign kind field anyway
+          ((cast[ptr `kindType`](cast[uint](`tmpSym`.addr) + `kindOffsetLit`))[]) = kindTmp
+    dst.add nnkCaseStmt.newTree(nnkDotExpr.newTree(tmpSym, kindSym))
+    for i in 1 ..< typeNode.len:
+      foldObjectBody(dst, typeNode[i], tmpSym, jsonNode, jsonPath, originalJsonPathLen)
+
+  of nnkOfBranch, nnkElse:
+    let ofBranch = newNimNode(typeNode.kind)
+    for i in 0 ..< typeNode.len-1:
+      ofBranch.add copyNimTree(typeNode[i])
+    let dstInner = newNimNode(nnkStmtListExpr)
+    foldObjectBody(dstInner, typeNode[^1], tmpSym, jsonNode, jsonPath, originalJsonPathLen)
+    # resOuter now contains the inner stmtList
+    ofBranch.add dstInner
+    dst[^1].expectKind nnkCaseStmt
+    dst[^1].add ofBranch
+
+  of nnkObjectTy:
+    typeNode[0].expectKind nnkEmpty
+    typeNode[1].expectKind {nnkEmpty, nnkOfInherit}
+    if typeNode[1].kind == nnkOfInherit:
+      let base = typeNode[1][0]
+      var impl = getTypeImpl(base)
+      while impl.kind in {nnkRefTy, nnkPtrTy}:
+        impl = getTypeImpl(impl[0])
+      foldObjectBody(dst, impl, tmpSym, jsonNode, jsonPath, originalJsonPathLen)
+    let body = typeNode[2]
+    foldObjectBody(dst, body, tmpSym, jsonNode, jsonPath, originalJsonPathLen)
+
+  else:
+    error("unhandled kind: " & $typeNode.kind, typeNode)
+
+macro assignObjectImpl[T](dst: var T; jsonNode: JsonNode; jsonPath: var string) =
+  let typeSym = getTypeInst(dst)
+  let originalJsonPathLen = genSym(nskLet, "originalJsonPathLen")
+  result = newStmtList()
+  result.add quote do:
+    let `originalJsonPathLen` = len(`jsonPath`)
+  if typeSym.kind in {nnkTupleTy, nnkTupleConstr}:
+    # both, `dst` and `typeSym` don't have good lineinfo. But nothing
+    # else is available here.
+    detectIncompatibleType(typeSym, dst)
+    foldObjectBody(result, typeSym, dst, jsonNode, jsonPath, originalJsonPathLen)
+  else:
+    foldObjectBody(result, typeSym.getTypeImpl, dst, jsonNode, jsonPath, originalJsonPathLen)
+
+proc initFromJson[T: object|tuple](dst: var T; jsonNode: JsonNode; jsonPath: var string) =
+  assignObjectImpl(dst, jsonNode, jsonPath)
+
+proc to*[T](node: JsonNode, t: typedesc[T]): T =
+  ## `Unmarshals`:idx: the specified node into the object type specified.
+  ##
+  ## Known limitations:
+  ##
+  ##   * Heterogeneous arrays are not supported.
+  ##   * Sets in object variants are not supported.
+  ##   * Not nil annotations are not supported.
+  ##
+  runnableExamples:
+    let jsonNode = parseJson("""
+      {
+        "person": {
+          "name": "Nimmer",
+          "age": 21
+        },
+        "list": [1, 2, 3, 4]
+      }
+    """)
+
+    type
+      Person = object
+        name: string
+        age: int
+
+      Data = object
+        person: Person
+        list: seq[int]
+
+    var data = to(jsonNode, Data)
+    doAssert data.person.name == "Nimmer"
+    doAssert data.person.age == 21
+    doAssert data.list == @[1, 2, 3, 4]
+
+  var jsonPath = ""
+  result = default(T)
+  initFromJson(result, node, jsonPath)
 
 when false:
-  import os
+  import std/os
   var s = newFileStream(paramStr(1), fmRead)
   if s == nil: quit("cannot open the file" & paramStr(1))
   var x: JsonParser
@@ -1103,82 +1392,3 @@ when false:
 
 # { "json": 5 }
 # To get that we shall use, obj["json"]
-
-when isMainModule:
-  #var node = parse("{ \"test\": null }")
-  #echo(node.existsKey("test56"))
-  var parsed = parseFile("tests/testdata/jsontest.json")
-  var parsed2 = parseFile("tests/testdata/jsontest2.json")
-  echo(parsed)
-  echo()
-  echo(pretty(parsed, 2))
-  echo()
-  echo(parsed["keyÄÖöoßß"])
-  echo()
-  echo(pretty(parsed2))
-  try:
-    echo(parsed["key2"][12123])
-    raise newException(ValueError, "That line was expected to fail")
-  except IndexError: echo()
-
-  let testJson = parseJson"""{ "a": [1, 2, 3, 4], "b": "asd" }"""
-  # nil passthrough
-  assert(testJson{"doesnt_exist"}{"anything"}.isNil)
-  testJson{["c", "d"]} = %true
-  assert(testJson["c"]["d"].bval)
-
-  # Bounds checking
-  try:
-    let a = testJson["a"][9]
-    assert(false, "EInvalidIndex not thrown")
-  except IndexError:
-    discard
-  try:
-    let a = testJson["a"][-1]
-    assert(false, "EInvalidIndex not thrown")
-  except IndexError:
-    discard
-  try:
-    assert(testJson["a"][0].num == 1, "Index doesn't correspond to its value")
-  except:
-    assert(false, "EInvalidIndex thrown for valid index")
-
-  # Generator:
-  var j = %* [{"name": "John", "age": 30}, {"name": "Susan", "age": 31}]
-  assert j == %[%{"name": %"John", "age": %30}, %{"name": %"Susan", "age": %31}]
-
-  var j2 = %*
-    [
-      {
-        "name": "John",
-        "age": 30
-      },
-      {
-        "name": "Susan",
-        "age": 31
-      }
-    ]
-  assert j2 == %[%{"name": %"John", "age": %30}, %{"name": %"Susan", "age": %31}]
-
-  var name = "John"
-  let herAge = 30
-  const hisAge = 31
-
-  var j3 = %*
-    [ { "name": "John"
-      , "age": herAge
-      }
-    , { "name": "Susan"
-      , "age": hisAge
-      }
-    ]
-  assert j3 == %[%{"name": %"John", "age": %30}, %{"name": %"Susan", "age": %31}]
-
-  discard """
-  while true:
-    var json = stdin.readLine()
-    var node = parse(json)
-    echo(node)
-    echo()
-    echo()
-  """