Lightweight Groovy toolkit for Google App Engine

A lightweight
Groovy toolkit
for developing web applications
on Google App Engine

Guillaume Laforge / @glaforge

• Groovy Project Manager
• Head of Groovy Development
at SpringSource
• Initiator of the Grails framework
• Founder of the Gaelyk toolkit
• Co-author of Groovy in Action
• Member of «Les Cast Codeurs» podcast

• Speaker: JavaOne, QCon, JavaZone, Sun TechDays,
Devoxx, The Spring Experience, SpringOne2GX,
JAX, Dynamic Language World, IJTC, and more...

Copyright 2010 SpringSource. Copying, publishing or distributing without express written permission is prohibited. 2

Et la suite du le Paris GGUG ?

• Aujourd’hui, nous allons parler de Gaelyk

• Et demain ?
– Spock — Mathilde Lemée
– Transformation d’AST — Hamlet D’Arcy
– Gradle — Grégory Boissinot

• Et vous ?
– Racontez-nous votre utilisation de Groovy,
Grails, Gradle, Spock, etc.


IaaS, PaaS, SaaS

• Software as a Service
– Gmail, SalesForce.com SaaS
• Platform as a Service
– Google App Engine
PaaS
• Infrastructure as a Service
– Amazon EC2
IaaS

Google App Engine

• Google’s PaaS solution
• Run your app on Google’s infrastructure

• Initially just Python supported

• Java supported added too
– Sandboxed JVM
– Jetty servlet container

• Several JVM-compatible language supported


Key aspects

• You can use most of your usual web frameworks for
developping apps on App Engine Java
– A WAR file, basically! (an exploded war actually)
– Uploading to the cloud by sending deltas of changes

• No OS image, or software to install
– Unlike with Amazon EC2

• All the scaling aspects are handled for you
– Database / session replication, load balancing...

• There are quotas, but you need a high traffic
application to start being charged
– Free to get started


Available services

• Memcache – Send / receive Jabber
messages (GTalk)
– JCache implementation
– Save on CPU and DB • User
– Use Google’s user/
• URL Fetch authentication system
– Access remote resources – OAuth support
– HttpUrlConnection
• Cron & Task queues
• Mail – Schedule tasks at regular
– Support both incoming and intervals
outgoing emails – Queue units of work
• Images • Blobstore
– Resize, crop, rotate... – For storing large content
• XMPP • And much more...

Limitations

•Not our usual relational database
– key / value datastore

•30 seconds request duration limit
• Forbidden to
– write on the file system
– create threads
– use raw sockets
– issue system calls
– use IO / Swing / etc. directly
•There’s a whitelist of classes allowed

• Number of files and their size are limited

Quotas


Quotas (1/2)

•Bandwidth •Mail
– 1,3M requests/day – 7K calls/day
– 1GB/day in/out – 2K recepients/day
– 6.5 CPU hours/day – 5K emails/day
– 2K attachments
– 100MB of attachments
•Datastore
– 10M calls
– 1GB/day •URL Fetch
– 12GB in / 115GB out – 657K calls/day
– 60 CPU hours/day – 4GB in/out /day


Quotas (2/2)

•XMPP •Memcache
– 657K calls/day – 8.6M calls/day
– 4GB data sent/day – 10GB in
– 657K recepients/day – 50GB out
– 1MB invitations/day
•Task queues
•Image manipulation – 100K calls

–864 calls/day
–1GB in / 5GB out
–2.5M transforms


Quotas (2/2)

•XMPP •Memcache
– 657K calls/day – 8.6M calls/day
– 4GB data sent/day – 10GB in
– 657K recepients/day – 50GB out
– 1MB invitations/day
•Task queues
•Image manipulation – 100K calls

–864 calls/day
ur es
–1GB in / 5GB out d fig
da te
–2.5M transforms
O ut


The datastore...

• Not your father’s relational database! «NoSQL»

• Distributed key / value store
– Based on Google’s «BigTable»
– Schema-less approach
• Supporting
– Transactions and partitioning
– Hierarchies through entity groups
• Data access APIs
– JPA and JDO
• but adds a big request load time factor
– Direct low-level APIs


...and its «limitations»

• You’re not using SQL
–No joins
– No database constraints
– No aggregation functions (count, avg...)

• In a query, you can only filter on one column
for inequality

• Transactions only available in entity groups

• You can only update an entity once in a transaction


http://gaelyk.appspot.com

F

Gaelyk 0.6 released!


• Gaelyk is a lightweight Groovy toolkit on top of
the Google App Engine Java SDK

• Gaelyk builds on Groovy’s servlet support
– Groovlets: Groovy scripts instead of raw servlets!
– Groovy templates: JSP-like template engine
– Both allow for a clean separation of views and logic

• Gaelyk provides several enhancements around the
GAE Java SDK to make life easier, thanks to
Groovy’s dynamic nature


Why Groovy?

• Groovy is a dynamic language for the JVM
– very flexible, malleable, expressive and concise syntax
– easy to learn for Java developers
• deriving from the Java 5 grammar
– provides powerful APIs to simplify the life of developers
• possibility to dynamically enrich existing APIs
– support for Groovlets and its own template engine


First steps...

• Go to http://gaelyk.appspot.com
• Download the template project
• Put your ﬁrst Groovlet in /WEB-INF/groovy
• And your templates in /WEB-INF/pages
• And you’re ready to go!
• Launch dev_appserver.sh
• Go to http://localhost:8080/

The web.xml

">
<web-app xmlns="http://java.sun.com/xml/ns/javaee" version="2.5
<listener>
/servlet-class>
<listener-class>groovyx.gaelyk.GaelykServletContextListener<
</listener>
<servlet>
<servlet-name>GroovletServlet</servlet-name>
<servlet-class>groovyx.gaelyk.GaelykServlet</servlet-class>
</servlet>
<servlet>
<servlet-name>TemplateServlet</servlet-name>
-class>
<servlet-class>groovyx.gaelyk.GaelykTemplateServlet</servlet
</servlet>

<servlet-mapping>
<servlet-name>GroovletServlet</servlet-name>
<url-pattern>*.groovy</url-pattern>
</servlet-mapping>
<servlet-mapping>
<servlet-name>TemplateServlet</servlet-name>
<url-pattern>*.gtpl</url-pattern>
</servlet-mapping>
</web-app>

MVC: Groovlets and templates

Groovlets Templates
(controllers) (views)

Entities
(domain)

Copyright 2010 SpringSource. Copying, publishing or distributing without express written permission is prohibited.

A groovlet

• Instead of writing full-blown servlets, just write
Groovy scripts (aka Groovlets)

def numbers = [1, 2, 3, 4]
def now = new Date()

html.html {
body {
numbers.each { number -> p number }
p now
}
}


A groovlet

• Instead of writing full-blown servlets, just write
Groovy scripts (aka Groovlets)

def numbers = [1, 2, 3, 4]
def now = new Date()

html.html {
body {
numbers.each { number -> p number }
p now
}
di ng
oa
}

-rel
to
Au

A template

<html>
<body>
<%
def message = "Hello World!"
print message %>

<%= message %>
${message}
<ul>
<% 3.times { %>
<li>${message}</li>
<% } %>
</ul>
</body>
</html>

A template

<html>
<body>
<%
def message = "Hello World!"
print message %>

<%= message %>
${message}
<ul>
<% 3.times { %>
<li>${message}</li>
<% } %>
di ng
</ul>
el oa
</body>
to -r
Au
</html>

Shortcuts

• Google services • Variables available
–datastore –request / response
–blobstore –context / application
–memcache –sessions
–capabilities –params / headers
–images –out / sout / html
–urlFetch –localMode / app.*
–mail
–userService / user • Methods available
–defaultQueue / queues –include / forward /
–xmpp redirect
–namespace –println / print

Sending emails with Gaelyk

mail.send to: 'to@gmail.com',
from: 'other@gmail.com',
subject: 'Hello World',
htmlBody: '<bold>Hello</bold>'

...compared to Java

Properties props = new Properties();
, null);
Sessio n session = Session.getDefaultInstance(props

try {
Message msg = new MimeMessage(session);
om", "Admin"));
msg .setFrom(new InternetAddress("other@gmail.c
O,
msg.addRecipient(Message.RecipientType.T
, "Mr. User"));
new InternetAddress("to@example.com"
msg.setSubject("Hello World");
msg.setText("<bold>Hello</bold>");
Transport.send(msg);
} catch (AddressException e) {}
} catch (MessagingException e) {}

Accessing the datastore

• Direct interaction with the low-level datastore API
Entity entity = new Entity("person")

map
// subscript notation, like when accessing a
entity['name'] = "Guillaume Laforge"

// normal property access notation
entity.age = 32

entity.save() // asyncSave()
entity.delete() // asyncDelete()

datastore.withTransaction {
// do stuff with your entities
// within the transaction
}

// use the asynchronous datastore service
datastore.async.put(entity)


Querying to be improved...

import com.google.appengine.api.datastore.*
der.*
import static com.google.appengine.api.datastore.FetchOptions.Buil

// query the scripts stored in the datastore
def query = new Query("savedscript")

// sort results by descending order of the creation date
query.addSort("dateCreated", Query.SortDirection.DESCENDING)

author
// filters the entities so as to return only scripts by a certain
r)
query.addFilter("author", Query.FilterOperator.EQUAL, params.autho

PreparedQuery preparedQuery = datastore.prepare(query)

// return only the first 10 results
def entities = preparedQuery.asList( withLimit(10) )


...into something groovier?

def entities = datastore.query {
select all from savedscript
sort desc by dateCreated
where author == params.author
limit 10
} as List


...into something groovier?

def entities = datastore.query {
select all from savedscript
sort desc by dateCreated
where author == params.author
limit 10 te d!
} as List m en
p le
et Im
o tY
N


URL Routing system (1/4)

• You can have friendly URL mappings with the URL
routing system
all "/aboutus",
redirect: "/blog/2008/10/20/about-us"

all "/blog/@year/@month/@day/@title",
@day@title=@title"
forward: "/blog.groovy?year=@year&month=@month@day=

get "/blog/@year/@month/@day",
@day"
forward: "/blog.groovy?year=@year&month=@month@day=

get "/book/isbn/@isbn",
forward: "/book.groovy?isbn=@isbn",
validate: { isbn ==~ /d{9}(d|X)/ }

– You’ll have to configure the RouteFilter in web.xml



• You can also define caching times

get "/aboutus", cache: 24.hours,
forward: "/aboutus.gtpl"

get "/breaking-news", cache: 1.minute,
forward: "/news.groovy?last=10"

– Nice for GAE’s infamous «loading requests»
•less critical with GAE SDK 1.4
– ability to reserve 3 instances (billing)
– warmup requests



• Special routes for specifying
– incoming emails
– jabber messages

email to "/incomingMail.groovy"

jabber to "/incomingXmpp.groovy"



• Namespace awareness: nice for multitenancy
• Capability awareness: for graceful degradation

// @cust customer variable could be « acme »
post "/@cust/update", forward: "/update.groovy",
namespace: { "ns-$cust" }

es status
// different destinations depending on the GAE servic
get "/speakers", forward {
to "/speakers.groovy" // default destination
// when the datastore is not available
to "/unavailable.gtpl" on DATASTORE not ENABLED
// show some maintenance is upcoming

to "/speakers.groovy?maintenance=true" on DATASTORE
is SCHEDULED_MAINTENANCE
}


Capabilities

• Google App Engine allows you to know the status
and availability of the various services
– DATASTORE, DATESTORE_WRITE, MEMCACHE...
– ENABLED, DISABLED, UNKNOWN,
SCHEDULED_MAINTENANCE
– is() and not() methods

if (capabilities[DATASTORE_WRITE].is(ENABLED)) {
// write some content in the datastore
} else {
// otherwise, redirect to some maintenance page
}


Task queue API

// access a configured queue using the subscript notation
queues['dailyEmailQueue']

// or using the property access notation
queues.dailyEmailQueue

// you can also access the default queue with:
queues.default
defaultQueue

// add a task to the queue
queue << [
countdownMillis: 1000, url: "/task/dailyEmail",
taskName: "sendDailyEmailNewsletter",
method: 'PUT', params: [date: '20090914'],
payload: content
]

Jabber / XMPP support (1/3)

• Sending instant messages

String recipient = "someone@gmail.com"

// check if the user is online
if (xmpp.getPresence(recipient).isAvailable()) {
// send the message
def status = xmpp.send(to: recipient,
body: "Hello, how are you?")

// checks the message was successfully
// delivered to all the recipients
assert status.isSuccessful()
}



• Sending instant messages with an XML payload
String recipient = "service@gmail.com"

// check if the service is online
// send the message
def status = xmpp.send(to: recipient, xml: {
customers {
customer(id: 1) {
name 'Google'
}
}
})

// checks the message was successfully delivered to the service
}


• Sending instant messages with an XML payload
String recipient = "service@gmail.com"

// check if the service is online
// send the message
def status = xmpp.send(to: recipient, xml: {
customers {
customer(id: 1) { <customers>
name 'Google' <customer id=’1’>
} <name>Google</name>
} </customer>
</customers>
})

// checks the message was successfully delivered to the service
}


• Receving incoming instant messages
– Once you’ve configured a route for jabber messages
– Add the inbound message service in appengine-web.xml

def message = xmpp.parseMessage(request)
// get the body of the message
message.body
// get the sender Jabber ID
message.from
// get the list of recipients Jabber IDs
message.recipients

// if the message is an XML document instead of a raw string message
if (message.isXml()) {
// get the raw XML
message.stanza
// get a document parsed with XmlSlurper
message.xml
}

Memcache service

• Map notation access to the cache

}
class Country implements Serialzable { String name

def countryFr = new Country(name: 'France')

in the cache
// use the subscript notation to put a country object
// (you can also use non-string keys)
memcache['FR'] = countryFr

// check that a key is present in the cache
if ('FR' in memcache) {
the cache using a key
// use the subscript notation to get an entry from
def countryFromCache = memcache['FR']
}


Closure memoization

• Cache the return values for each dinstinct
invocation (for a given arguments set)

->
def countEntities = memcache.memoize { String kind
datastore.prepare( new Query(kind) )
.countEntities()
}

// first call
def totalPics = countEntityes('photos')

// second call, hitting the cache
totalPics = countEntityes('photos')


Blobstore enhancements

• The blobstore allows to store some large content
– images, videos, etc.
def blob = ...
print blob.filename // contentType, creation, size

// output the content of the blob to the response
blob.serve response

// read the content of the blob
blob.withReader { Reader r -> ... }
blob.withStream { InputStream is -> ... }

// delete the blob
blob.delete()


Images service

• Readable DSL for manipulating images

def blobKey = ...

def image = blobKey.image.transform {
resize 1600, 1200
crop 0.1, 0.1, 0.9, 0.9
horizontal flip // vertical flip too
rotate 90
feeling lucky // image corrections
}

def thumbnail = image.resize(100, 100)


Channel Service

• For Comet-style applications

def token = hannel.createChannel('clientID')
channel.send 'clientID', 'Hi!'

• Then in the view, in JavaScript...

channel = new goog.appengine.Channel(token);
socket = channel.open();
socket.onmessage = function(msg) { ... }


Simple plugin system (1/3)

• Gaelyk features a simple plugin system for
extending your apps and share commonalities

• A plugin lets you
– provide additional groovlets and templates
– contribute new URL routes
– add new categories
– define variables in the binding
– override existing binding variables
– provide any static content
– add new libraries
– do any initialization
– add before/after request hooks


• A plugin is actually just a zip file!
– Basically, just a Gaelyk application, minus...
•the Groovy / Gaelyk / GAE JARs
•the web.xml and appengine-web.xml descriptors
– But with a /WEB-INF/plugins/myplugin.groovy
descriptor

• A plugin must be referenced in /WEB-INF/
plugins.groovy with
– install myplugin
•shortcut to /WEB-INF/plugins/myplugin.groovy



• An example plugin descriptor
– /WEB-INF/plugins/jsonPlugin.groovy
import net.sf.json.*
import net.sf.json.groovy.*
 
// add new variables in the binding
binding {
le
jsonLibVersion = "2.3"          // a simple string variab
 of a 3rd‐party JAR
json = new JsonGroovyBuilder()  // an instance of a class
}
 
// add new routes with the usual routing system format
routes {
get "/json", forward: "/json.groovy"
}
 
// install a category you've developped
categories jsonlib.JsonlibCategory

before { req, resp ‐> ... } // or after
 
// any other initialization code you'd need


a ry
m m
S u

Copyright 2010 SpringSource. Copying, publishing or distributing without express written permission is prohibited.

Summary

• Easy access to a cloud solution
–Deploying Java apps, as easily as you would with PHP
• Familiar to Java folks
–Your good old Servlet centric webapps style
• Pretty cheap
–You need a high-trafficed website to reach the quotas
• Gaelyk provides a simplified approach to
creating Servlet centric webapps in a
productive manner
–Leveraging Groovy’s servlet / template support and
dynamic capabilities


What’s coming next?

• Expect more sugar around the Datastore
– An SQL-like query DSL
– Easier relationship management (builder?)

• Perhaps pre-compiled groovlets and templates
– Less needed since SDK 1.4
• (reserve 3 hot instances, warmup requests)

• Testing facilities specific to Gaelyk
–testing utility class, Spock-specific ones...

• Anything that’ll come up in newer GAE SDK versions


Thanks for your attention!

e
e Laforg velopment
Gu illaum oovy De m
of Gr e@gmail.co
Head laforg
E mail: g glaforge
@
T witter:

• References:
• http://gaelyk.appspot.com/
• http://groovy.codehaus.org/
• http://code.google.com/
appengine/


Picture credits

Light bulb: https://newsline.llnl.gov/retooling/mar/03.28.08_images/lightBulb.png
Speed limit : http://www.morguefile.com/archive/display/18492
Warehouse : http://www.morguefile.com/archive/display/85628
Check mark: http://www.lnl.infn.it/~epics/WikiDumps/localhost/600px-symbol_ok.svg.png
Puzzle: http://www.everystockphoto.com/photo.php?imageId=263521
Light bulb: https://newsline.llnl.gov/retooling/mar/03.28.08_images/lightBulb.png
Clouds http://www.morguefile.com/archive/display/627059
http://www.morguefile.com/archive/display/625552
http://www.morguefile.com/archive/display/629785
Duke ok GAE http://weblogs.java.net/blog/felipegaucho/archive/ae_gwt_java.png
Python logo : http://python.org/images/python-logo.gif
Gaelyc cross with clouds : http://www.morguefile.com/archive/display/37889
Speed limit : http://www.morguefile.com/archive/display/18492
Warehouse : http://www.morguefile.com/archive/display/85628
Snow foot steps : http://www.flickr.com/photos/robinvanmourik/2875929243/
Sugar : http://www.flickr.com/photos/ayelie/441101223/sizes/l/
Press release: http://www.napleswebdesign.net/wp-content/uploads/2009/06/press_release_11.jpg


Lightweight Groovy toolkit for Google App Engine

Recommended

Recommended

More Related Content

What's hot

What's hot (12)

Viewers also liked

Viewers also liked (6)

Similar to Lightweight Groovy toolkit for Google App Engine

Similar to Lightweight Groovy toolkit for Google App Engine (20)

More from Guillaume Laforge

More from Guillaume Laforge (20)

Recently uploaded

Recently uploaded (20)

Lightweight Groovy toolkit for Google App Engine