The following code:
import scala.pickling.{FastTypeTag, Pickler, Unpickler}
import scala.pickling.binary._
import scala.pickling.Defaults._
class Serializer[T : Pickler : FastTypeTag] {
def serialize(data: T): Array[Byte] = {
Foo.bar(data)
}
}
object Foo {
def bar[T: Pickler: FastTypeTag](t: T): Array[Byte] = t.pickle.value
def unbar[T: Unpickler: FastTypeTag](bytes: Array[Byte]): T = bytes.unpickle[T]
}
class Message(message: String)
implicit object messageSerializer extends Serializer[Message]
def test[A: Pickler: FastTypeTag: Serializer](message: A): Array[Byte] = {
implicitly[Serializer[A]].serialize(message)
}
val message = new Message("message")
test(message)
Evaluates to:
import scala.pickling.{FastTypeTag, Pickler, Unpickler}
import scala.pickling.binary._
import scala.pickling.Defaults._
defined class Serializer
defined module Foo
defined class Message
defined module messageSerializer
test: test[A](val message: A)(implicit <synthetic> val evidence$10: scala.pickling.Pickler[A],implicit <synthetic> val evidence$11: pickling.FastTypeTag[A],implicit <synthetic> val evidence$12: Serializer[A]) => Array[Byte]
message: Message = com.impresign.hub.core.A$A12$A$A12$Message@6c7df044
java.lang.VerifyError: Bad type on operand stack
Exception Details:
Location:
com/impresign/hub/core/A$A12$A$A12$messageSerializer$.<init>(Lcom/impresign/hub/core/A$A12$A$A12;)V @212: invokespecial
Reason:
Type uninitializedThis (current frame, stack[3]) is not assignable to 'com/impresign/hub/core/A$A12$A$A12$messageSerializer$'
Current Frame:
bci: @212
flags: { flagThisUninit }
locals: { uninitializedThis, 'com/impresign/hub/core/A$A12$A$A12', 'scala/runtime/VolatileObjectRef', 'scala/Tuple2' }
stack: { uninitializedThis, 'com/impresign/hub/core/A$A12$A$A12', 'scala/Predef$', uninitializedThis, 'scala/runtime/VolatileObjectRef', 'com/impresign/hub/core/A$A12$A$A12' }
Bytecode:
0x0000000: 2a2b b200 2db8 0031 4db2 0036 b600 3ab9
0x0000010: 0040 0100 b900 4601 0099 00c5 bb00 4859
0x0000020: b200 36b6 003a b900 4001 00b2 004d 124f
0x0000030: b200 54b6 0058 b900 5e01 00b9 0062 0200
0x0000040: b200 36b6 003a b900 4001 00b2 004d 124f
0x0000050: b200 54b6 0058 b900 5e01 00b9 0065 0200
0x0000060: b700 684e 2dc6 006c 2db6 006c c000 6e3a
0x0000070: 042d b600 71c0 006e 3a05 1904 c100 7399
0x0000080: 0052 1904 c000 733a 0619 06b6 0076 c000
0x0000090: 783a 0719 05c1 0073 9900 3919 05c0 0073
0x00000a0: 3a08 1908 b600 76c0 007a 3a09 1907 1909
0x00000b0: 3a0a 59c7 000c 5719 0ac6 000e a700 1519
0x00000c0: 0ab6 0080 9900 0d19 07c0 0082 3a0b a700
0x00000d0: 0b2a 2c2b b700 843a 0b19 0ba7 0009 2a2c
0x00000e0: 2bb7 0084 b600 88c0 0078 b200 4d12 4fb2
0x00000f0: 0054 b600 58b7 008b b1
Stackmap Table:
full_frame(@191,{UninitializedThis,Object[#147],Object[#10],Object[#72],Object[#110],Object[#110],Object[#115],Object[#120],Object[#115],Object[#122],Object[#122]},{UninitializedThis,Object[#147],Object[#41],Object[#120]})
full_frame(@199,{UninitializedThis,Object[#147],Object[#10],Object[#72],Object[#110],Object[#110],Object[#115],Object[#120],Object[#115],Object[#122],Object[#122]},{UninitializedThis,Object[#147],Object[#41]})
full_frame(@209,{UninitializedThis,Object[#147],Object[#10],Object[#72]},{UninitializedThis,Object[#147],Object[#41]})
full_frame(@217,{UninitializedThis,Object[#147],Object[#10],Object[#72],Top,Top,Top,Top,Top,Top,Top,Object[#130]},{UninitializedThis,Object[#147],Object[#41]})
full_frame(@222,{UninitializedThis,Object[#147],Object[#10]},{UninitializedThis,Object[#147],Object[#41]})
full_frame(@228,{UninitializedThis,Object[#147],Object[#10]},{UninitializedThis,Object[#147],Object[#41],Object[#130]})
Output exceeds cutoff limit.
Never saw anything like it. I'd like to use implicit Serializers with Scala Pickler 0.10 under the hood. Is this achievable? Scala 2.11.8.
If you change your implicit object declaration to a val
implicit val mssgSerializer: Serializer[Message] = new Serializer[Message]
everything works nicely. As @laughedelic comment mentions, this is probably due to some funny interaction between macros and objects initialization.
Tested with Scala 2.11.8 and Scala Pickling 0.10.1.