A typical roadmap is something which gives a chosen possible path with alternate routes, options and milestones. But these alternatives and options stop at the map level and do not move into product roadmaps in the industries. It is very common to see product roadmaps published with milestones and when to expect them but all of them are linear. It is not a roadmap, it is just the road ahead which is more or less static. Responding to change is possible only when you have a roadmap, else every problem that comes in the way puts us in a hardship as we were not prepared to respond with alternatives.
How do we come up with a roadmap for software projects? Just like maps the requirements have to be visually triaged. We need to collect them at granular level such that it can be captured recorded in one or two sentences that can be written in bold on an index card. These granular requirements are popularly called as stories which represent an action of an end user or the ability of the system.
Once we collect most of the possible actions and abilities then we lay them all together on a large table or put it up on a wall to see them all at one glance. This helps us to triage quickly and discard duplicate ones and add some more if they were found to be missing. Once we have come up with an initial list then we collect the cards as per the priorities and arrange them linearly. If a requirement can be done in parallel then a parallel stream is introduced. In this manner we will be able to have a visual data on what are the dependencies and how much can be done in parallel.
We should play around the sequence and parallel streams to come up with multiple options for milestones and these multiple options together constitute a roadmap. Every few months it is better to revisit to the roadmap as software projects are always a moving target. This is nothing but the release planning meeting mentioned at XP website. We have to be aware of the options and not just rely on the first release plan that we have created, if we rely only on the first draft then it is just the road ahead; we will achieve all the milestones as per the plan but it may mostly lead to a not so desired location.
TestPackage provides a clean way to package your functional tests and run them as a jar file with sharding and package level run options. All that we need to do is created a maven/gradle subproject for functional tests and have the testpackage’s dependency and manifest file entry to get an executable jar package.
Once the jar is created all we need to do is run ‘java -jar the-test-package.jar package.that.contain.tests’. This helps a lot in continuous delivery where the functional test artifact can be created at the build stage and promoted to run at the functional test stage without having to access the source code and rebuild later.
In one of the spring web applications we decided to use Log4J2 with SLF4J; when we deployed the web app in tomcat server along with the required jars we stumbled on the issue multiple binding for SLF4J and spring dependencies not getting satisfied. We had used Gradle and there were transitive dependencies which included log4j-12xx.jar and slf4j-log4j.jar as some of the dependencies in the project were using log4j version 1 for logging. Here is how we were able to start logging using log4j2.
Exclude the log4j-12xx.jar and slf4j-log modules from the gradle build
org.slf4j:log4j-over-slf4j|route all the log4j v1 to slf4j
org.slf4j:slf4j-api|slf4j api
org.apache.logging.log4j:log4j-slf4j-impl|log4j2 bridge
org.apache.logging.log4j:log4j-core|log4j2 core
org.apache.logging.log4j:log4j-api|log4j2 api
Add the log4j2.xml configurations to the classpath or as specified at log4j manual