Allow for arbitrary metadata to be encoded into a bundle (#334)

* * Add meta field to Bundle, allowing arbitrary metadata to be encoded into bundle
* Add unit-test for ensuring bundle serialization / deserialization works as expected
* Fix issue with incorrect default setting of BasicType in Value when parsing from JSON

* Update bundle-protobuf submodule

* Rewrite meta unit tests to work with Scala 2.10

bundle-ml/src/main/scala/ml/combust/bundle/BundleWriter.scala

@@ -11,9 +11,11 @@ import scala.util.Try
 case class BundleWriter[Context <: HasBundleRegistry,
 Transformer <: AnyRef](root: Transformer,
                        name: Option[String] = None,
-                       format: SerializationFormat = SerializationFormat.Json) {
+                       format: SerializationFormat = SerializationFormat.Json,
+                       meta: Option[ml.bundle.Attributes] = None) {
   def name(value: String): BundleWriter[Context, Transformer] = copy(name = Some(value))
   def format(value: SerializationFormat): BundleWriter[Context, Transformer] = copy(format = value)
+  def meta(value: ml.bundle.Attributes): BundleWriter[Context, Transformer] = copy(meta = Some(value))
 
   def save(file: BundleFile)
           (implicit context: Context): Try[Bundle[Transformer]] = {
@@ -23,6 +25,7 @@ Transformer <: AnyRef](root: Transformer,
 
     BundleSerializer(context, file).write(Bundle(name = n,
       format = format,
-      root = root))
+      root = root,
+      meta = meta))
   }
 }

bundle-ml/src/main/scala/ml/combust/bundle/dsl/Bundle.scala

@@ -101,12 +101,14 @@ object Bundle {
 
   def apply[Transformer <: AnyRef](name: String,
                                    format: SerializationFormat,
-                                   root: Transformer): Bundle[Transformer] = {
+                                   root: Transformer,
+                                   meta: Option[ml.bundle.Attributes] = None): Bundle[Transformer] = {
     apply(BundleInfo(uid = UUID.randomUUID(),
       name = name,
       format = format,
       version = Bundle.version,
-      timestamp = LocalDateTime.now().toString), root)
+      timestamp = LocalDateTime.now().toString,
+      meta = meta), root)
   }
 }
 
@@ -116,7 +118,8 @@ object BundleInfo {
       name = bundle.name,
       format = SerializationFormat.fromBundle(bundle.format),
       version = bundle.version,
-      timestamp = bundle.timestamp)
+      timestamp = bundle.timestamp,
+      meta = bundle.meta)
   }
 }
 
@@ -132,13 +135,16 @@ case class BundleInfo(uid: UUID,
                       name: String,
                       format: SerializationFormat,
                       version: String,
-                      timestamp: String) {
+                      timestamp: String,
+                      meta: Option[ml.bundle.Attributes]) {
   def asBundle: ml.bundle.Bundle = {
     ml.bundle.Bundle(uid = uid.toString,
       name = name,
       format = format.asBundle,
       version = version,
-      timestamp = timestamp)
+      timestamp = timestamp,
+      meta = meta
+    )
   }
 }
 

bundle-ml/src/main/scala/ml/combust/bundle/json/JsonSupport.scala

@@ -39,7 +39,7 @@ trait JsonSupport {
       case JsString("double") => BasicType.DOUBLE
       case JsString("string") => BasicType.STRING
       case JsString("byte_string") => BasicType.BYTE_STRING
-      case JsString("unknown") => BasicType.Unrecognized(100)
+      case JsString("unknown") => BasicType.UNDEFINED
       case _ => deserializationError("invalid basic type")
     }
 
@@ -290,6 +290,6 @@ trait JsonSupport {
 
   implicit val bundleNodeFormat: RootJsonFormat[Node] = jsonFormat2(Node.apply)
   implicit val bundleModelFormat: RootJsonFormat[Model] = jsonFormat2(Model.apply)
-  implicit val bundleBundleInfoFormat: RootJsonFormat[Bundle] = jsonFormat5(Bundle.apply)
+  implicit val bundleBundleInfoFormat: RootJsonFormat[Bundle] = jsonFormat6(Bundle.apply)
 }
 object JsonSupport extends JsonSupport

bundle-ml/src/test/scala/ml/combust/bundle/serializer/BundleSerializationSpec.scala

@@ -2,12 +2,13 @@ package ml.combust.bundle.serializer
 
 import java.net.URI
 
-import ml.combust.bundle.{BundleFile, BundleRegistry, TestUtil}
+import ml.bundle.Attributes
+import ml.combust.bundle.dsl.Bundle
+import ml.combust.bundle.test.TestSupport._
 import ml.combust.bundle.test._
 import ml.combust.bundle.test.ops._
+import ml.combust.bundle.{BundleFile, BundleRegistry, TestUtil, dsl}
 import org.scalatest.FunSpec
-import TestSupport._
-import ml.combust.bundle.dsl.Bundle
 import resource._
 
 import scala.util.Random
@@ -122,6 +123,29 @@ class BundleSerializationSpec extends FunSpec {
           }
         }
       }
+
+      describe("with metadata") {
+        it("serializes and deserializes any metadata we want to send along with the bundle") {
+          val uri = new URI(s"$prefix:${TestUtil.baseDir}/lr_bundle_meta.$format$suffix")
+          val meta = new Attributes().withList(Map(
+            "keyA" → dsl.Value.string("valA").value,
+            "keyB" → dsl.Value.double(1.2).value,
+            "keyC" → dsl.Value.stringList(Seq("listValA", "listValB")).value
+          ))
+          for(file <- managed(BundleFile(uri))) {
+            lr.writeBundle.name("my_bundle").
+              format(format).
+              meta(meta).
+              save(file)
+
+            val bundleRead = file.loadBundle().get
+
+            // These are written this way explicitly for compatibility with Scala 2.10
+            assert(!bundleRead.info.meta.isEmpty)
+            assert(bundleRead.info.meta.get.equals(meta))
+          }
+        }
+      }
     }
   }
 }