August 14, 2009

Add Client Side Caching to Groovlets and Separate Logic and View

We are going to see how we can implement client side caching for our Groovlet we made in the previous blog. Client side caching decreases bandwidth usage and also lessens the load on our application. Besides a faster response for the users we can even save costs if we have to pay for bandwidth usage if the appliation is running in a cloud for example. We will also refactor the Groovlet and use a view template for displaying the data and remove all HTML generation from the Groovlet. This way we have a better separation between the code and the HTML.

The current Groovlet display a list of activities and each activity has a certain date. We use this date to print out nice human readable, relative timestamps like we see on Twitter. For example 2 days ago, just now, 1 week ago. Each time we refresh the Groovlet we want to update this information. Because the time is relative compared to the current date and time we cannot cache this data. Even if all activity information stays the same we still have to update the human readable timestamps, otherwise the timestamps don't make sense. To solve this and make the data cachable we move the code to generate the human readable timestamps from the server to the client. On the client side we use Javascript to generate the relative date and time values. Now we can implement client side caching for our Groovlet output, because the uncachable date and time values are generated on the client side and not on the server.

Here is the code for the Groovlet. We place it in a file war/WEB-INF/groovy/viewactivities.groovy:

import net.sf.json.groovy.JsonSlurper
import net.sf.json.JSONException
import org.apache.commons.codec.digest.DigestUtils

// We keep track of the last time the content of the Yahoo! Pipe
// has changed in the application context.
now = new Date()
lastUpdated = application.getAttribute('lastUpdated')
if (!lastUpdated) {
    lastUpdated = now
    application.setAttribute('lastUpdated', lastUpdated)

// Get the contents of the Yahoo! Pipe.
pipeUrl = 'http://pipes.yahoo.com/pipes/pipe.run?_id=UtVVPkx83hGZ2wKUKX1_0w&_render=json'
result = urlFetchService.fetch(new URL(pipeUrl))
def resultString = new String(result.content)

// The Yahoo! Pipe contains a pubDate element, which contains the date and time
// the Pipe is requested. This changes each time we make a request, even all other
// data stays the same. We filter out this date, so we can make a MD5 hash without
// the dynamic date and time value for better comparison.
resultStringNoPubDate = resultString.replaceFirst(/"pubDate":"\w{3}, \d{2,} \w{3} \d{4} \d{2}:\d{2}:\d{2} [+-]\d{4}",/, '')

// Determine MD5 hash for the Pipe's content.
etag = DigestUtils.md5Hex(resultStringNoPubDate)

// If the request header contains the 'If-None-Match' header we must see if the 
// value is the same as we calculated above. If so the content of the Pipe has not changed
// and we can return a 304 Not Modified response.
// Or if the request header contains a 'If-Modified-Since' header we see if the value
// is the same as the last time the content has changed. We return a 304 Not Modified
// if this is the case.
// The response is send directly to the client without any further content.
if (headers['If-None-Match'] == etag || headers['If-Modified-Since'] == lastUpdated.time) {
    response.sendError javax.servlet.http.HttpServletResponse.SC_NOT_MODIFIED

// The content has changed so we update the lastUpdated value.
lastUpdated = now
application.setAttribute('lastUpdated', lastUpdated)

// We set the response headers 'ETag', 'Last-modified'
// and 'Cache-Control'. Cache-control is set to 5 minutes.
response.setHeader('ETag', etag)
response.setDateHeader('Last-modified', lastUpdated.time)
response.setHeader('Cache-Control', 'max-age=' + 5 * 60)

// Set request attributes with values we can use in the 'activities.gtpl' view.
if (resultString) {
    try {
        def jsonReader = new JsonSlurper()
        def json = jsonReader.parseText(resultString)
        request.setAttribute('json', json)

    } catch (JSONException jsonException) {
        request.setAttribute('error', jsonException.getMessage() + "
") } } else { request.setAttribute('error', 'No results found. Try again later.') } // Go to the view 'activities.gtpl'. forward 'activities.gtpl'

At line 22 we remove the value for the pubDate element from the Yahoo! Pipe. This value is updated each time we make a request. If we want to be able to compare the MD5 hash of the content from a previous request with the current result we must exclude the dynamic date and time value. If we don't remove it we always get a different MD5 hash value and we wouldn't be able to send back a 304 Not Modified response.

In line 25 we calculate the MD5 hash with the Apache Commons Codec library. We need to place the JAR file in the war/WEB-INF/lib directory of our Google App Engine application.

At line 34 we see if the request header If-None-Match is available. The value of this header is the MD5 hash of the content from a previous request. We compare this value to the freshly calculated value in line 25. If they are equal we know the content of the Pipe hasn't changed, so the client can use a cached version of the page. As a fallback we also check the If-Modified-Since header in case the client doesn't support If-None-Match.

We set the response header Cache-Control to five minutes at line 47. This means the client can use a cached copy of the page for five minutes before it needs to check the server again to see if something has changed. This means if the content has changed in five minutes after the last request, the client will not see it. For this application that is acceptable, but we always need to check if such a period is acceptable or not for the users.

At lines 50-61 we populate the request with attributes we can use in the view template. The JSON results is stored in a request variable named json. If any error has occured we save it in the request variable error.

At the last line we leave the Groovlet and continue on to the view activites.gtpl.

We create a view activites.gtpl in the directory war to display the results:

import java.text.SimpleDateFormat
import java.text.ParseException

// Closure to parse the date string from the activities.
def parseActivityDate = { dateString ->
    def result
    dateString = dateString.replaceAll(/(\d{2}):(\d{2})$/, '$1$2')
    def parsers = [ 
        new SimpleDateFormat("EEE, dd MMM yyyy HH:mm:ss Z", Locale.US),
        new SimpleDateFormat("yyyy-MM-dd'T'HH:mm:ss'Z'", Locale.US),
        new SimpleDateFormat("yyyy-MM-dd'T'HH:mm:ss.SSSZ", Locale.US)
    for (SimpleDateFormat parser : parsers) {
        try {
            result = parser.parse(dateString)
        } catch (ParseException e) {

    <title>Social Activities</title>
    <link type='text/css' rel='stylesheet' href='css/activities.css'/>

    <% if (request.getAttribute('error')) { %>
    <% } else { %>
      <div id="activities">

      <% request.getAttribute('json').value.items.each { item -> %>
        <div class="item">
          <img src="images/socialicons/32x32/${item.source}.png" title="${item.source}" alt="${item.source}"/>
          <a href="${item.link}">${item.title}</a>
          <br />
          <span class="date" title="${parseActivityDate(item.pubDate).format("yyyy-MM-dd'T'HH:mm:ss")}"> </span>
    <% } %>
    <script src='js/jquery.js' type='text/javascript'> </script>
    <script src='js/humane.js' type='text/javascript'> </script>
    <script type='text/javascript'>
     \$(document).ready(function() {
       setInterval(function() {
       }, 5000);
   <% } %>


At the top we define a closure to parse the date strings from the JSON results. We use it at line 43 to format the date for the title attribute of the span element. At line 50 we import the Yet Another Pretty Date Javascript library. This library will use the value of the title attribute of an element and create a human readable timestamp. At lines 52-57 we define Javascript to calculate the timestamp values and display them on the page. We also define an interval of five seconds to update the information. This way the information is always up-to-date without refreshing the page. Notice how we must escape the $ sign to have it placed literally in the output, otherwise Groovy thinks it is an Groovy expression.

The rest of the code is straightforward. At line 38 we loop through the JSON results and display a div element with the information for each item.

We have seen how we can add reponse headers to our Groovlet to support client side caching. We set the Cache-Control header to minimize the number of requests to the server. And if a request hits our Groovlet we check if the contents has changed. If there is no change we return a 304 Not Modified response, so a cached version on the client side can be used. Furthermore we refactored the previous Groovlet to separate the logic from the view. A more spiced up example of the application is available on Google App Engine.