Fixed up lambda-node to work with multi-stage changes.

docs/lambda/README.md

@@ -6,10 +6,10 @@ anywhere. You should be able to take your code and run them without any changes.
 ## Creating Lambda Functions
 
 Creating Lambda functions is not much different than using regular functions, just use
-the `lambda-node` runtime.
+the `lambda-node-4` runtime.
 
 ```sh
-fn init --runtime lambda-node <DOCKER_HUB_USERNAME>/lambda-node
+fn init --runtime lambda-node-4 <DOCKER_HUB_USERNAME>/lambda-node
 ```
 
 Be sure the filename for your main handler is `func.js`.

examples/lambda/node/func.js

@@ -7,6 +7,6 @@ exports.handler = (event, context, callback) => {
     console.log('value1 =', event.key1);
     console.log('value2 =', event.key2);
     console.log('value3 =', event.key3);
-    callback(null, event.key1);  // Echo back the first key value
+    callback(null, "hello " + event.key1);  // Echo back the first key value
     //callback('Something went wrong');
 };

fn/lambda/node-4/build.sh

@@ -0,0 +1,3 @@
+set -ex
+
+docker build --build-arg HTTP_PROXY -t funcy/lambda:node-4 .

fn/lambda/node-4/release.sh

@@ -0,0 +1,5 @@
+set -ex
+
+./build.sh
+
+docker push funcy/lambda:node-4

fn/lambda/node-6/Dockerfile

@@ -0,0 +1,13 @@
+FROM node:6-alpine
+
+WORKDIR /function
+
+# Install ImageMagick and AWS SDK as provided by Lambda.
+RUN apk update && apk --no-cache add imagemagick
+RUN npm install aws-sdk@2.2.32 imagemagick && npm cache clear
+
+# cli should forbid this name
+ADD bootstrap.js /function/lambda-bootstrap.js
+
+# Run the handler, with a payload in the future.
+ENTRYPOINT ["node", "./lambda-bootstrap"]

fn/lambda/node-6/bootstrap.js

@@ -0,0 +1,330 @@
+'use strict';
+
+var fs = require('fs');
+
+var oldlog = console.log
+console.log = console.error
+
+// Some notes on the semantics of the succeed(), fail() and done() methods.
+// Tests are the source of truth!
+// First call wins in terms of deciding the result of the function. BUT,
+// subsequent calls also log. Further, code execution does not stop, even where
+// for done(), the docs say that the "function terminates". It seems though
+// that further cycles of the event loop do not run. For example:
+// index.handler = function(event, context) {
+//   context.fail("FAIL")
+//   process.nextTick(function() {
+//     console.log("This does not get logged")
+//   })
+//   console.log("This does get logged")
+// }
+// on the other hand:
+// index.handler = function(event, context) {
+//   process.nextTick(function() {
+//     console.log("This also gets logged")
+//     context.fail("FAIL")
+//   })
+//   console.log("This does get logged")
+// }
+//
+// The same is true for context.succeed() and done() captures the semantics of
+// both. It seems this is implemented simply by having process.nextTick() cause
+// process.exit() or similar, because the following:
+// exports.handler = function(event, context) {
+//     process.nextTick(function() {console.log("This gets logged")})
+//     process.nextTick(function() {console.log("This also gets logged")})
+//     context.succeed("END")
+//     process.nextTick(function() {console.log("This does not get logged")})
+// };
+//
+// So the context object needs to have some sort of hidden boolean that is only
+// flipped once, by the first call, and dictates the behavior on the next tick.
+//
+// In addition, the response behaviour depends on the invocation type. If we
+// are to only support the async type, succeed() must return a 202 response
+// code, not sure how to do this.
+//
+// Only the first 256kb, followed by a truncation message, should be logged.
+//
+// Also, the error log is always in a json literal
+// { "errorMessage": "<message>" }
+var Context = function() {
+  var concluded = false;
+
+  var contextSelf = this;
+
+  // The succeed, fail and done functions are public, but access a private
+  // member (concluded). Hence this ugly nested definition.
+  this.succeed = function(result) {
+    if (concluded) {
+      return
+    }
+
+    // We have to process the result before we can conclude, because otherwise
+    // we have to fail. This means NO EARLY RETURNS from this function without
+    // review!
+    if (result === undefined) {
+      result = null
+    }
+
+    var failed = false;
+    try {
+      // Output result to log
+      oldlog(JSON.stringify(result));
+    } catch(e) {
+      // Set X-Amz-Function-Error: Unhandled header
+      console.log("Unable to stringify body as json: " + e);
+      failed = true;
+    }
+
+    // FIXME(nikhil): Return 202 or 200 based on invocation type and set response
+    // to result. Should probably be handled externally by the runner/swapi.
+
+    // OK, everything good.
+    concluded = true;
+    process.nextTick(function() { process.exit(failed ? 1 : 0) })
+  }
+
+  this.fail = function(error) {
+    if (concluded) {
+      return
+    }
+
+    concluded = true
+    process.nextTick(function() { process.exit(1) })
+
+    if (error === undefined) {
+      error = null
+    }
+
+    // FIXME(nikhil): Truncated log of error, plus non-truncated response body
+    var errstr = "fail() called with argument but a problem was encountered while converting it to a to string";
+
+    // The semantics of fail() are weird. If the error is something that can be
+    // converted to a string, the log output wraps the string in a JSON literal
+    // with key "errorMessage". If toString() fails, then the output is only
+    // the error string.
+    try {
+      if (error === null) {
+        errstr = null
+      } else {
+        errstr = error.toString()
+      }
+      oldlog(JSON.stringify({"errorMessage": errstr }))
+    } catch(e) {
+      // Set X-Amz-Function-Error: Unhandled header
+      oldlog(errstr)
+    }
+  }
+
+  this.done = function() {
+    var error = arguments[0];
+    var result = arguments[1];
+    if (error) {
+      contextSelf.fail(error)
+    } else {
+      contextSelf.succeed(result)
+    }
+  }
+
+  var plannedEnd = Date.now() + (getTimeoutInSeconds() * 1000);
+  this.getRemainingTimeInMillis = function() {
+    return Math.max(plannedEnd - Date.now(), 0);
+  }
+}
+
+function getTimeoutInSeconds() {
+  var t = parseInt(getEnv("TASK_TIMEOUT"));
+  if (Number.isNaN(t)) {
+    return 3600;
+  }
+
+  return t;
+}
+
+var getEnv = function(name) {
+  return process.env[name] || "";
+}
+
+var makeCtx = function() {
+  var fnname = getEnv("AWS_LAMBDA_FUNCTION_NAME");
+  // FIXME(nikhil): Generate UUID.
+  var taskID = getEnv("TASK_ID");
+
+  var mem = getEnv("TASK_MAXRAM").toLowerCase();
+  var bytes = 300 * 1024 * 1024;
+
+  var scale = { 'b': 1, 'k': 1024, 'm': 1024*1024, 'g': 1024*1024*1024 };
+  // We don't bother validating too much, if the last character is not a number
+  // and not in the scale table we just return a default value.
+  // We use slice instead of indexing so that we always get an empty string,
+  // instead of undefined.
+  if (mem.slice(-1).match(/[0-9]/)) {
+    var a = parseInt(mem);
+    if (!Number.isNaN(a)) {
+      bytes = a;
+    }
+  } else {
+    var rem = parseInt(mem.slice(0, -1));
+    if (!Number.isNaN(rem)) {
+      var multiplier = scale[mem.slice(-1)];
+      if (multiplier) {
+        bytes = rem * multiplier
+      }
+    }
+  }
+
+  var memoryMB = bytes / (1024 * 1024);
+
+  var ctx = new Context();
+  Object.defineProperties(ctx, {
+    "functionName": {
+      value: fnname,
+      enumerable: true,
+    },
+    "functionVersion": {
+      value: "$LATEST",
+      enumerable: true,
+    },
+    "invokedFunctionArn": {
+      // FIXME(nikhil): Should be filled in.
+      value: "",
+      enumerable: true,
+    },
+    "memoryLimitInMB": {
+      // Sigh, yes it is a string.
+      value: ""+memoryMB,
+      enumerable: true,
+    },
+    "awsRequestId": {
+      value: taskID,
+      enumerable: true,
+    },
+    "logGroupName": {
+      // FIXME(nikhil): Should be filled in.
+      value: "",
+      enumerable: true,
+    },
+    "logStreamName": {
+      // FIXME(nikhil): Should be filled in.
+      value: "",
+      enumerable: true,
+    },
+    "identity": {
+      // FIXME(nikhil): Should be filled in.
+      value: null,
+      enumerable: true,
+    },
+    "clientContext": {
+      // FIXME(nikhil): Should be filled in.
+      value: null,
+      enumerable: true,
+    },
+  });
+
+  return ctx;
+}
+
+var setEnvFromHeader = function () {
+  var headerPrefix = "CONFIG_";
+  var newEnvVars = {};
+  for (var key in process.env) {
+    if (key.indexOf(headerPrefix) == 0) {
+      newEnvVars[key.slice(headerPrefix.length)] = process.env[key];
+    }
+  }
+
+  for (var key in newEnvVars) {
+    process.env[key] = newEnvVars[key];
+  }
+}
+
+
+function run() {
+  setEnvFromHeader();
+  // FIXME(nikhil): Check for file existence and allow non-payload.
+  var path = process.env["PAYLOAD_FILE"];
+  var stream = process.stdin;
+  if (path) {
+    try {
+      stream = fs.createReadStream(path);
+    } catch(e) {
+      console.error("bootstrap: Error opening payload file", e)
+      process.exit(1);
+    }
+  }
+
+  var input = "";
+  stream.setEncoding('utf8');
+  stream.on('data', function(chunk) {
+    input += chunk;
+  });
+
+  stream.on('error', function(err) {
+    console.error("bootstrap: Error reading payload stream", err);
+    process.exit(1);
+  });
+
+  stream.on('end', function() {
+    var payload = {}
+    try {
+      if (input.length > 0) {
+        payload = JSON.parse(input);
+      }
+    } catch(e) {
+      console.error("bootstrap: Error parsing JSON", e);
+      process.exit(1);
+    }
+
+    if (process.argv.length > 2) {
+      var handler = process.argv[2];
+      var parts = handler.split('.');
+      // FIXME(nikhil): Error checking.
+      var script = parts[0];
+      var entry = parts[1];
+      var started = false;
+      try {
+        var mod = require('./'+script);
+        var func = mod[entry];
+        if (func === undefined) {
+          oldlog("Handler '" + entry + "' missing on module '" + script + "'");
+          return;
+        }
+
+        if (typeof func !== 'function') {
+          throw "TypeError: " + (typeof func) + " is not a function";
+        }
+        started = true;
+        var cback 
+        // RUN THE FUNCTION:
+        mod[entry](payload, makeCtx(), functionCallback)
+      } catch(e) {
+        if (typeof e === 'string') {
+          oldlog(e)
+        } else {
+          oldlog(e.message)
+        }
+        if (!started) {
+          oldlog("Process exited before completing request\n")
+        }
+      }
+    } else {
+      console.error("bootstrap: No script specified")
+      process.exit(1);
+    }
+  })
+}
+
+function functionCallback(err, result) {
+    if (err != null) {
+        // then user returned error and we should respond with error
+        // http://docs.aws.amazon.com/lambda/latest/dg/nodejs-prog-mode-exceptions.html
+        oldlog(JSON.stringify({"errorMessage": errstr }))
+        return
+    }
+    if (result != null) {
+        oldlog(JSON.stringify(result))
+    }
+}
+
+run()

fn/lambda/node-6/build.sh

@@ -0,0 +1,3 @@
+set -ex
+
+docker build --build-arg HTTP_PROXY -t funcy/lambda:node-6 .

fn/lambda/node-6/release.sh

@@ -0,0 +1,5 @@
+set -ex
+
+./build.sh
+
+docker push funcy/lambda:node-6

fn/lambda/node/build.sh

@@ -1,3 +0,0 @@
-set -ex
-
-docker build --build-arg HTTP_PROXY -t treeder/functions-lambda:nodejs4.3 .

fn/lambda/node/release.sh

@@ -1,5 +0,0 @@
-set -ex
-
-./build.sh
-# TODO: where to push?
-# docker push treeder/functions-lambda:nodejs4.3

fn/lambda/release.sh

@@ -1,5 +1,9 @@
 set -ex
 
-cd node
+cd node-4
+./release.sh
+cd ..
+
+cd node-6
 ./release.sh
 cd ..

fn/langs/base.go

@@ -27,7 +27,7 @@ func GetLangHelper(lang string) LangHelper {
 		return &RustLangHelper{}
 	case "dotnet":
 		return &DotNetLangHelper{}
-	case "lambda-nodejs4.3":
+	case "lambda-nodejs4.3", "lambda-node-4":
 		return &LambdaNodeHelper{}
 	case "java":
 		return &JavaLangHelper{}
@@ -69,10 +69,11 @@ func (h *BaseHelper) RunFromImage() string          { return h.BuildFromImage()
 func (h *BaseHelper) IsMultiStage() bool            { return true }
 func (h *BaseHelper) DockerfileBuildCmds() []string { return []string{} }
 func (h *BaseHelper) DockerfileCopyCmds() []string  { return []string{} }
+func (h *BaseHelper) Entrypoint() string            { return "" }
 func (h *BaseHelper) Cmd() string                   { return "" }
-func (lh *BaseHelper) HasPreBuild() bool            { return false }
-func (lh *BaseHelper) PreBuild() error              { return nil }
-func (lh *BaseHelper) AfterBuild() error            { return nil }
+func (h *BaseHelper) HasPreBuild() bool             { return false }
+func (h *BaseHelper) PreBuild() error               { return nil }
+func (h *BaseHelper) AfterBuild() error             { return nil }
 func (h *BaseHelper) HasBoilerplate() bool          { return false }
 func (h *BaseHelper) GenerateBoilerplate() error    { return nil }
 

fn/langs/lambda_node.go

@@ -5,26 +5,26 @@ type LambdaNodeHelper struct {
 }
 
 func (lh *LambdaNodeHelper) BuildFromImage() string {
-	return "funcy/functions-lambda:nodejs4.3"
+	return "funcy/lambda:node-4"
 }
 
-func (lh *LambdaNodeHelper) Entrypoint() string {
-	return ""
+func (lh *LambdaNodeHelper) IsMultiStage() bool {
+	return false
 }
 
 func (lh *LambdaNodeHelper) Cmd() string {
 	return "func.handler"
 }
 
-func (lh *LambdaNodeHelper) HasPreBuild() bool {
-	return false
-}
-
-// PreBuild for Go builds the binary so the final image can be as small as possible
-func (lh *LambdaNodeHelper) PreBuild() error {
-	return nil
-}
-
-func (lh *LambdaNodeHelper) AfterBuild() error {
-	return nil
+func (h *LambdaNodeHelper) DockerfileBuildCmds() []string {
+	r := []string{}
+	if exists("package.json") {
+		r = append(r,
+			"ADD package.json /function/",
+			"RUN npm install",
+		)
+	}
+	// single stage build for this one, so add files
+	r = append(r, "ADD . /function/")
+	return r
 }