All actions performed within ff4j can be stored in a EventRepository (event are already published asynchronously). https://github.com/clun/ff4j/blob/master/ff4j-core/src/main/java/org/ff4j/audit/repository/EventRepository.java

The implement must be provided for EventRepository for database (JDBC) in the same way as PropertyStore.

We need to have a auditing capability so that we always can trace who did which change.

One way of doing it would be a write through hazelcast to a GIT-repo with the user as committer and/or author.

Hi ,
The artifact in the Maven central are not java 1.6 compatible .

How we can use ff4j with java 1.6 project , without compiling ff4j-core in local ?

Thanks in Advance

If the eventpublisher self-initialises, there is no code which eventually stops the thread pool. You will then have something like:

"ff4j-monitoring-pool-1-thread-1" prio=10 tid=0x00007fbc10084000 nid=0x3c4d waiting on condition [0x00007fbb7e140000]
   java.lang.Thread.State: WAITING (parking)
        at sun.misc.Unsafe.park(Native Method)
        - parking to wait for  <0x0000000790456708> (a java.util.concurrent.locks.AbstractQueuedSynchronizer$ConditionObject)
        at java.util.concurrent.locks.LockSupport.park(LockSupport.java:186)
        at java.util.concurrent.locks.AbstractQueuedSynchronizer$ConditionObject.await(AbstractQueuedSynchronizer.java:2043)
        at java.util.concurrent.ArrayBlockingQueue.take(ArrayBlockingQueue.java:374)
        at java.util.concurrent.ThreadPoolExecutor.getTask(ThreadPoolExecutor.java:1068)
        at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1130)
        at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:615)
        at java.lang.Thread.run(Thread.java:745)

which prevents cleanly stopping the application

There are some corrections to be made one the website:

• Link the pdf from github instead of sourceforge
• The images and links of the use-cases do not work correctly

Since it is a static site, I recommend using github to host the site.

Simplify configuration

It looks like when Strategy class name is long enough, the view console is chopped off.

Some interface allow to implements your own behavior as the FlippingStrategy and it should be more visible

• Today ff4j-jmx expose only few operations and not a whole "FeatureStore"
• Today ff4j-jmx relies on Spring-JMX where is could be full JMX

Hi,
The date format specified in the ReleaseDateStrategy must be set to "yyyy-MM-dd-HH:mm" instead of "yyyy-MM-DD-HH:mm", since DD is day in year instead of day in month.

Thank you,
Arsen

When using the ff4j taglib, I encountered the following error :

The absolute uri: http://www.ff4j.org/taglibs/ff4j cannot be resolved in either web.xml or the jar files deployed with this application

In documentation it says that the taglib should be discovered if the file is present in META-INF. I'm not sure whether the file should be present in the root of META-INF folder or not.

For information, I use jetty.

Might just be a documentation issue, but have added ff4j-web as a dependency and included servlet AdministrationConsoleServlet as per https://github.com/clun/ff4j#administr-ation-servlet.

However, this gives the following response when /ff4j-console request is made:

2014-03-26 16:34:41.440:WARN:oejs.ServletHandler:qtp765511918-56: /ff4j-console
java.lang.IllegalArgumentException: FF4J has not been initialized.
at org.ff4j.web.FF4jWebContextHolder.getFf4j(FF4jWebContextHolder.java:57)

Is there something else that needs to be done to initialize FF4J, other than the Spring component scan?

Hi,

In the embedded console, the js file is served with a content type "text/plain"
So if I deploy this servlet in an application with spring security, I get the following error when I get to the page :

Refused to execute script from 'https://localhost:8080/admin/ff4j?rsc=js' because its MIME type ('text/plain') is not executable, and strict MIME type checking is enabled.

Is there any reason to serve the file in "text/plain" instead of "application/javascript" ?

• Graceful Degradation
• Zero Downtime Design
• A/B Testing

It would be useful to be able to decorate any FeatureStore with caching and persistence.

I need to be able to use a remote client that continues working even if the central server is unavailable. Even if you start the application when the central store is down you should be able to fall back onto the previous value. I am thinking of someything like this:

• If feature flag is in cache and not older than timeout specified - use that.
• If cache expired, try remote.
• If remote timeout (needs to be configurable), trip circuit breaker and try loading from local persistance store.
• If remote answers, put into cache and persist to local storage
• If all fails, use default value

All actions performed within ff4j can be stored in a EventRepository (event are already published asynchronously). https://github.com/clun/ff4j/blob/master/ff4j-core/src/main/java/org/ff4j/audit/repository/EventRepository.java

The implement must be provided for EventRepository for REDIS in the same way as FeatureStore.

Become a standard for all cache manager

Write Chapter upon JDBC feature Store

I have the following org.ff4j.strategy.el.ExpressionFlipStrategy in ff4j.xml file which I use for initializing FF4j instance.

<feature uid="kernel-featrue1" enable="false" description="description" >
        <flipstrategy class="org.ff4j.strategy.el.ExpressionFlipStrategy">
                <param name="expression"><![CDATA[Feature1 & !Feature2]]></param>
        </flipstrategy>
</feature>

But when I use FF4j.exportFeatures() method, it produces the following XML element which is invalid.

<flipstrategy class="org.ff4j.strategy.el.ExpressionFlipStrategy" >
     <param name="expression" value="Feature1 & !Feature2" />
</flipstrategy>

Those can be used for flipping strategies or as configurations

org.ff4j.test.store.FeatureStoreMongoDBCore1Test refers to org.ff4j.store.FeatureStoreMongoDB;

However org.ff4j.store.FeatureStoreMongoDB is not in the 1.2 distribution or in the source. Can you please address this?

Currently, the constructor for FeatureCacheProviderRedis accepts 2 parameters but uses defaults while creating new Jedis object.

public FeatureCacheProviderRedis(String host, int port) {
    jedis = new Jedis(redisHost, redisport);
}

Kindly update the constructor to use the parameters.

The method below uses a prefix to search for cache keys.

@Override
public Set<String> listCachedFeatureNames() {
    return jedis.keys(PREFIX_KEY + "*");
}

I believe the presence of the prefix "PREFIX_KEY" may be an error. The "put" method does not use the prefix to cache the feature. Hence this method returns empty set unless it is backed by a Redis feature store (which is not correct, as in that case it return the actual feature rather than the cached feature)

The script to create the tables needed to store the features via JDBC has a problem. The column "FEAT_UID" has different sizes, but they should be the same:

-- Roles to store ACL, FK to main table
CREATE TABLE FF4J_ROLES (
"FEAT_UID" VARCHAR(50) REFERENCES FF4J_FEATURES("FEAT_UID"),
...
);


The column should VARCHAR(100) for both:

-- Roles to store ACL, FK to main table
CREATE TABLE FF4J_ROLES (
"FEAT_UID" VARCHAR(100) REFERENCES FF4J_FEATURES("FEAT_UID"),
...
);


Hello,

I am interested to use feature toggles on an Android project. Do you have any experience on this? Is it possible?

Cheers.

ff4j-1.4.0.xsd contains two attribute elements with name value. When running
xjc ff4j-1.4.0.xsd
xjc issues the following error:
Property "Value" is already defined

While it's possible using xml bindings file and map value to valueAttribute, it's a bad solution because further along the way when you unmarshal an ff4j xml with JAXB you'll get:
Multiple fields representing property "value"

I suggest changing the value attribute names in the xsd to something like valueAttribute

Need to change:

    public FF4j disableGroup(String groupName) {
        getFeatureStore().enableGroup(groupName);
        getEventPublisher().publish(groupName, EventType.DISABLE_GROUP);
        return this;

To

    public FF4j disableGroup(String groupName) {
        getFeatureStore().disableGroup(groupName);
        getEventPublisher().publish(groupName, EventType.DISABLE_GROUP);
        return this;

The aop capabilities is only working through ComponentScan and annotating bean with @component, @service or @repository.

Hi,

I'm using FeatureStoreMongodb and I'm wondering if if would be a good idea to be able to pass a filename to initialize with features configuration if features do not exist in database.

Currently, I use a workaround like this :

    @Bean
    public FeatureStore featureStore() {
        MongoCollection ff4j = jongo.getCollection("ff4j");
        FeatureStoreMongoDB storeMongoDB = new FeatureStoreMongoDB(ff4j.getDBCollection());

        InMemoryFeatureStore localStore = new InMemoryFeatureStore("ff4j.xml");
        Map<String, Feature> features = localStore.readAll();

        features.keySet()
                .stream()
                .filter(featureName -> !storeMongoDB.exist(featureName))
                .forEach(featureName -> storeMongoDB.create(features.get(featureName)));

        return storeMongoDB;
    }

I divert the original use of InMemoryFeatureStore just to use its parsers. Then I check if features exists in the mongodb store. And if not, I create the feature.
That's a solution to initialize the application with some features or to add features lately.

Maybe another way of writing that would be

    @Bean
    public FeatureStore featureStore() {
        MongoCollection ff4j = jongo.getCollection("ff4j");
        // all the previous logic managed by FeatureStoreMongoDB ?
        return new FeatureStoreMongoDB(ff4j.getDBCollection(), "ff4j.xml");
    }

I'm trying to deploy full stack console on a Tomcat server. I was able to deploy an embedded console on the Tomcat server. But what I need is to deploy the full stack console. I couldn't find any documentation on this.

Any help on this is much appreciated.

• Produce JavaDoc with last look and feel
• Generate maven site and upload on FF4J.org
• Javadoc on Github

ConsoleServlet should response with Javascript content-type when provided JavaScript scripts

I am testing with ff4j version 1.3.5

When i want to edit a feature i click on the edit button and it looks like this:

When i select any Group, Strategy or Permission, then i come to this page.

When i click the back button, no change is stored.

My browser is Mozilla Firefox 43.0.4

When I try to export the features, I get funny result.
I get the xml file and the html file concateneted in one file.

For example :

<?xml version="1.0" encoding="UTF-8" ?>
<head/><features xmlns="http://www.ff4j.org/schema/ff4j" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:schemaLocation="http://www.ff4j.org/schema/ff4j http://ff4j.org/schema/ff4j-1.2.0.xsd">>
<feature uid="choose_lang" description="Capability to choose language" enable="true">
<flipstrategy class="com.hopwork.ff4j.strategy.AdminOrAccountFilterStrategy">
<param name="accounts" value="some values"/>
</flipstrategy>
</feature>
<feature uid="moderation" description="Enable moderation" enable="true">
</feature>
</features>
<!DOCTYPE config>
<html lang="en">
<head>
<meta http-equiv="content-type" content="text/html; charset=utf-8"/>
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<meta name="apple-mobile-web-app-capable" content="yes">
<title>FF4J - Embedded Administration Console</title>
<link href="/admin/ff4j?rsc=css" rel="stylesheet">
<script type="text/javascript" charset="utf-8" src="/admin/ff4j?rsc=js"></script>
</head>
....

If I extract the xml part it's ok, the file is valid.

Will be useful for my project to use with Thymeleaf : http://www.thymeleaf.org/ecosystem.html
Thanks

See: getFeatureStore().enableGroup(groupName) in the following method. Issue in ff4j-core-1.3.1

/**
* Disable group.
*
* @param groupName
* target groupeName
* @return current instance
*/
public FF4j disableGroup(String groupName) {
getFeatureStore().enableGroup(groupName);
getEventPublisher().publish(groupName, EventType.DISABLE_GROUP);
return this;
}

Due to Spring 4.x Java Config the @Autowired property could lead to conflit in configuration beans, remove the annotation and let people inject the datasource on their own.

The thread which calls FF4j.check(String) get renamed to "FF4J Monitoring worker processing xyz". This is caused by the constructor for org.ff4j.audit.EventWorker. I am checking a flag in a servlet filter and my tomcat threads are being renamed.

Is there a way to read the reference documentation online without downloading a pdf?
I haven`t found a way, but i prefer to read it online instead of opening a pdf.

So maybe also add the documentation on http://ff4j.org/?

I could like to completely disable the auditing in ff4j in certain scenarios. Unfortunately EventPublisher is not an interface, so I cannot implement a NoOpEventPublish, but with the new setter for EventPublisher, I could set the EventPublisher to something like:

        new EventPublisher() {
            @Override
            public void publish(final Event e) {
                // do nothing
            }
        };

It would be nice if org.ff4j.FF4j had a setAuditEnabled(boolean) or something similiar which would cause it to not call getEventPublisher().publish(..) in the first place.

What do you think?

ConsoleServlet generates exception on renderPage method when a custom strategy is configured as Spring bean.
Attached is the screenshot.

Here is my config in Spring xml:

<bean id="ff4j" class="org.ff4j.FF4j" >
    <property name="store" ref="ff4j.store.inmemory" />
</bean>
<bean id="ff4j.store.inmemory" class="org.ff4j.store.InMemoryFeatureStore" >
    <constructor-arg type="java.util.Map">
        <map key-type="java.lang.String" value-type="org.ff4j.core.Feature">
            <entry key="myFeature" value-ref="myFeatureRef/>
        </map>
    </constructor-arg>
</bean>
<bean id="myFeatureRef class="org.ff4j.core.Feature">
    <constructor-arg name="uid" value="myFeature"/>
    <property name="enable" value="true"/>
    <property name="description" value="Feature description"/>
    <property name="flippingStrategy" ref="myReleaseStrategy"/>
    <property name="group" value="myFeatureGroup"/>
</bean>
<bean id="myReleaseStrategy" class="com.mycompany.feature.ReleaseDateFlipStrategy">
    <constructor-arg type="java.lang.String" value="${effectiveDate}"/>
</bean>

Everything in place to manage properties with :

• A PropertyPlaceholderConfigurer
• CRUD over properties
• Dedicated tab on console

If I edit a feature which is currently enabled through the embedded console, the feature status if flipped to false after the update. This is because the console servlet calls:

org.ff4j.web.embedded.ConsoleOperations.updateFeatureDescription(FF4j, HttpServletRequest)

which does not read the feature from the store but instead uses:

Feature fp = new Feature(featureId, false);

This means the object passed to the feature store for update always has its status set to false. Is this expected behaviour?

Today metrics relative to feature can go to JDBC or Inmemory databases with custom schema.

We wloud like to leverage on metrics and its numerous exporters. As a consequence will will publish the monitoring information into a metrics registry.

Hi,

I'm running a benchmark on our application and I found a place that looks weird at the first place in FeatureTagEnable :

executionContext.putString("LOCALE", pageContext.getRequest().getLocalName());

This is the full stack trace :

http-nio-8087-exec-173@13802" daemon prio=5 tid=0x14e nid=NA runnable
  java.lang.Thread.State: RUNNABLE
      at java.net.Inet6AddressImpl.getHostByAddr(Inet6AddressImpl.java:-1)
      at java.net.InetAddress$2.getHostByAddr(InetAddress.java:932)
      at java.net.InetAddress.getHostFromNameService(InetAddress.java:617)
      at java.net.InetAddress.getHostName(InetAddress.java:559)
      at java.net.InetAddress.getHostName(InetAddress.java:531)
      at org.apache.coyote.http11.Http11NioProcessor.actionInternal(Http11NioProcessor.java:341)
      at org.apache.coyote.http11.AbstractHttp11Processor.action(AbstractHttp11Processor.java:919)
      at org.apache.coyote.Request.action(Request.java:379)
      at org.apache.catalina.connector.Request.getLocalName(Request.java:1277)
      at org.apache.catalina.connector.RequestFacade.getLocalName(RequestFacade.java:989)
      at javax.servlet.ServletRequestWrapper.getLocalName(ServletRequestWrapper.java:326)
      at javax.servlet.ServletRequestWrapper.getLocalName(ServletRequestWrapper.java:326)
      at javax.servlet.ServletRequestWrapper.getLocalName(ServletRequestWrapper.java:326)
      at javax.servlet.ServletRequestWrapper.getLocalName(ServletRequestWrapper.java:326)
      at javax.servlet.ServletRequestWrapper.getLocalName(ServletRequestWrapper.java:326)
      at javax.servlet.ServletRequestWrapper.getLocalName(ServletRequestWrapper.java:326)
      at javax.servlet.ServletRequestWrapper.getLocalName(ServletRequestWrapper.java:326)
      at javax.servlet.ServletRequestWrapper.getLocalName(ServletRequestWrapper.java:326)
      at org.ff4j.web.taglib.FeatureTagEnable.eval(FeatureTagEnable.java:52)
      at org.ff4j.web.taglib.AbstractFeatureTag.doStartTag(AbstractFeatureTag.java:86)

Each time this is called, there is a Inet6AddressImpl.getHostByAddr called and it's quite costly.

What is the point of having the host name ?

When adding a feature using FeatureStoreMongodb, the produced json in database is :

> db.ff4j.find().pretty()
{
        "_id" : "choose_lang",
        "enable" : false,
        "description" : "Capability to choose language",
        "groupname" : null,
        "strategy" : "com.hopwork.ff4j.strategy.AdminOrAccountFilterStrategy",
        "expression" : "some values",
        "roles" : [ ]
}

However, it fails in FeatureDBObjectMapper on line 92 :

        if (dbObject.containsField(ROLES)) {
            for (Object role : (BasicBSONList) dbObject.get(ROLES)) {
                authorisation.add(role.toString());
            }
        }

=> 
ClassCastException: org.bson.LazyBSONList cannot be cast to org.bson.types.BasicBSONList

If I unset this variable in mongoshell, everything works correctly.