Merge pull request #97 from fission/workflow-env-headers

Add support for headers and query params to workflow invocations

Author: erwinvaneyk

examples/misc/inputs.wf.yaml

@@ -0,0 +1,9 @@
+# The Inputs Workflow - simply prints the inputs it received, including body, query params and headers.
+#
+# Example: curl -XPUT -H "hello: world" http://$FISSION_ROUTER/fission-function/inputs?a=b
+apiVersion: 1
+output: Printer
+tasks:
+  Printer:
+    run: compose
+    inputs: "{$.Invocation.Inputs}"

examples/whales/metadatawhale.wf.yaml

@@ -5,4 +5,6 @@ tasks:
   PrefixedFortune:
     run: fortune
     inputs:
-      header_prefix: "Whale says:"
+      headers:
+        # If the 'prefix' header is non-empty, we use that. Otherwise we default to "whale says"
+        prefix: "{ $.Invocation.Inputs.headers.Prefix || 'Whale says: ' }"

pkg/fnenv/fission/envproxy.go

@@ -25,7 +25,8 @@ import (
 	"strings"
 )
 
-// Proxy between Fission and Workflow to ensure that workflowInvocations comply with Fission function interface
+// Proxy between Fission and Workflow to ensure that workflowInvocations comply with Fission function interface. This
+// ensures that workflows can be executed exactly like Fission functions are executed.
 type Proxy struct {
 	invocationServer apiserver.WorkflowInvocationAPIServer
 	workflowServer   apiserver.WorkflowAPIServer
@@ -77,57 +78,57 @@ func (fp *Proxy) handleRequest(w http.ResponseWriter, r *http.Request) {
 		fp.fissionIds[fnId] = true
 	}
 
-	// Map Inputs to function parameters
+	// Map request to workflow inputs
 	inputs := map[string]*types.TypedValue{}
-	err := ParseRequest(r, inputs)
+	err := parseRequest(r, inputs)
 	if err != nil {
 		logrus.Errorf("Failed to parse inputs: %v", err)
 		http.Error(w, "Failed to parse inputs", 400)
 		return
 	}
 
+	wfSpec := &types.WorkflowInvocationSpec{
+		WorkflowId: fnId,
+		Inputs:     inputs,
+	}
+
 	// Temporary: in case of query header 'X-Async' being present, make request async
 	if len(r.Header.Get("X-Async")) > 0 {
-		invocatinId, err := fp.invocationServer.Invoke(ctx, &types.WorkflowInvocationSpec{
-			WorkflowId: fnId,
-			Inputs:     inputs,
-		})
+		invocationId, err := fp.invocationServer.Invoke(ctx, wfSpec)
 		if err != nil {
 			logrus.Errorf("Failed to invoke: %v", err)
 			http.Error(w, err.Error(), 500)
 			return
 		}
 		w.WriteHeader(200)
-		w.Write([]byte(invocatinId.Id))
+		w.Write([]byte(invocationId.Id))
 		return
 	}
 
 	// Otherwise, the request synchronous like other Fission functions
-	invocation, err := fp.invocationServer.InvokeSync(ctx, &types.WorkflowInvocationSpec{
-		WorkflowId: fnId,
-		Inputs:     inputs,
-	})
+	invocation, err := fp.invocationServer.InvokeSync(ctx, wfSpec)
 	if err != nil {
 		logrus.Errorf("Failed to invoke: %v", err)
-		http.Error(w, err.Error(), 500)
+		http.Error(w, err.Error(), http.StatusInternalServerError)
 		return
 	}
 
+	// In case of an error, create an error response corresponding to Fission function errors
 	if !invocation.Status.Status.Successful() {
 		logrus.Errorf("Invocation not successful, was '%v'", invocation.Status.Status.String())
 		http.Error(w, invocation.Status.Status.String(), 500)
 		return
 	}
 
-	// TODO determine header based on the output value
+	// Otherwise, create a response corresponding to Fission function responses.
 	var resp []byte
 	if invocation.Status.Output != nil {
 		resp = invocation.Status.Output.Value
 		w.Header().Add("Content-Type", inferContentType(invocation.Status.Output, defaultContentType))
 	} else {
 		logrus.Infof("Invocation '%v' has no output.", fnId)
 	}
-	w.WriteHeader(200)
+	w.WriteHeader(http.StatusOK)
 	w.Write(resp)
 }