coding paradigm: line by line

Maven Assembly Error on Ubuntu 11.10 (oneiric)

2012-01-20T17:01:00.002+02:00

I don't get it. I always expect OSS tools/utils to work correctly on Linux (especially on ubuntu). However, in the real life it's often not the case at all.
See https://jira.codehaus.org/browse/MASSEMBLY-588 for example (in two words maven assembly fails with some weird exception on Ubuntu while listing folder contents). Weird thing is - it works on Win7 no problem. During this particular project development (I use win7 as development box, this probably is the case) I have seen ~5 cases of weird maven behavior, all of them happend to be on Linux. Not judging, just saying.

Android multi-profile maven build

2011-12-21T12:14:00.002+02:00

Building the same Android app in different maven-profiles (backend connection urls ...) solved by copying projects pom.xml into different sub-folders and pointing them to andoid-application-original-source folder. The same (unduplicated) code, duplicated (adjusted) pom files. Any other recommendations?

Blog host is updated

2011-07-04T12:38:00.002+03:00

My pers blog moved out to http://leonardinius.galeoconsulting.com/ RSS update might be needed

jira-rest-cli-1.1-SNAPSHOT work in progress

2011-04-12T02:29:00.006+03:00

At the moment working on jira-rest-cli-1.1 release.

What is done already?

console client improved (groovy): understands a whole set of command line options. See client help message below:

usage: rest-cli-groovy -h <host> -u <user> -w <password> [options]
 -c,--context <context>               application context (e.g.: jira)
 -d,--drop-session <cli-session-id>   will terminate cli session
 -f,--file <file>                     will read file and evaluate it's
                                      contents. use - for stdin
 -h,--host <host>                     server hostname
 -help                                prints this message
 -l,--list-sessions                   list cli session ids
 -n,--new-session                     will create new session and exit
                                      immediately
 -p,--port <port>                     server port. defaults to [80]
 -proto,--protocol <protocol>         http/https protocol; could be
                                      derived from port. defaults to
                                      [http]
 -s,--session <cli-session-id>        cli session id to connect to
 -u,--user <user>                     admin user name to connect with
 -w,--password <password>             password to authenticate with

class loader problem fixed, now it's possible to use import and reference Jira classes from the scripts. See example below:

import com.atlassian.jira.project.ProjectManager;
import com.atlassian.jira.ComponentManager;

ProjectManager manager = ComponentManager.getInstance().getProjectManager();
manager.projectObjects.collect { it.name }

Work on automated configuration deployment (mini-framework started). If you are interested - please track https://github.com/leonardinius/jira-rest-cli/issues/1 and submit comments/use-cases.

Plans

Implement automated configuration deployment scripts, pre-alfa stage
Port (license, reasoning??) Script Runner bundled scripts? Make them rest-callable.
Get JRuby client to the state of groovy client.
Documentation and tests.

Release announcement: jira-rest-cli-1.0 (jsr223 powered interactive jira scripting console)

2011-03-29T01:47:00.027+03:00

Check it out

logo

jira-rest-cli is JIRA plugin which provides a possibility to use your favorite programming language to script and interact with JIRA server realtime. The plugin provide following working modes:

JIRA web-executor interface - allows to execute script input from Jira admin interface, no continuos working session support and no working context preservation between invocations.
JIRA web-cli interface - allows to create and manage working scripting sessions from Jira admin interface, connect to them and execute script code in the scripting session context - state is preserved between invocations.
The are sample console clients available (Ruby, Groovy) which works similar to interactive language shells (irb, groovysh).

At the moment following programming languages are supported:

JavaScript (Rhino) shipped with Oracle JDK - default
Groovy 1.7.9 - separate
JRuby 1.5.6 - separate.

Languages are implemented as standalone plug-able components, installed separately, except for Rhino available by default.

The target is to come as close to Firebug / IRB / Groovysh as possible :)

Project page: https://github.com/leonardinius/jira-rest-cli
README: README.md
Link to plugin exchange: at the moment not mature enough. Needs more documentation, specification and testing...

Screenshots

Web-executor mode (non-interactive)

Web-cli mode, session management

Web-cli mode, script execution (jRuby)

Remote console-cli mode, script execution (jRuby)

Project goal / scratch the itch
The problem I periodically get with the JIRA is the necessity of creating and configuring multiple different JIRA instance. In some cases this is pilot version, prototype extension or some stage/perf tests related configuration. Let's say it as it is - it's extremely difficult to setup thing in JIRA from scratch: projects, issue types, fields and all the schemas; there are just so mane of them out there.

So, I tried to automate certain things, at least at the local development environment. And I've found it extremely difficult to do so. So, I had this crazy idea - why couldn't I have some kind of JIRA API Firebug? Some environment I could actually script my environmental changes, experiment with the stuff and optionally save/reuse scripts later on. And yes: it would be cool if I would be able to use familiar to me ruby and maybe try out some different languages.

Basically it's the core idea behind my pet jira-rest-cli project.
Isn't something already available out there?
Before doing something I did my homework and looked for other available options:

Python CLI for JIRA - basically cli interface to JIRA SOAP interface;
Jira Scripting Suite - provides a convenient way to put custom conditions, validators and post-functions into workflow in a form of Jython scripts..
Script Runner - provide ability to script (JSR-223 capable) workflow validators, conditions, etc..

What's wrong with? Actually - nothing. They are great pieces of software and they excellently do they should do - extend JIRA functionality and provide possibility to easily extend workflows (Jira Scripting Suite, Script Runner) w/o need to restart Jira server, or provide access to built-in remote access (Python CLI for JIRA).

The thing I've missed here - is to play with Jira API in the realtime, see what's inside of teddy bear looks like (latvian: kas lācītim vēderā) and the ability to use the same approach to automate certain operations.

How could I use it?
When working with it I have several use-cases in mind:

Use it as console-tool to script and automate certain configuration changes (local development; staging etc development rollout)
Use it as a tool to play with JIRA system API at realtime (local development needs)

I would really appreciate if you will think out other use-cases and will report them back to me. So do it :)

How to start?

git clone git@github.com:leonardinius/jira-rest-cli.git
cd jira-rest-cli/
git submodule init
cd jira-rest-cli-parent/
atlas-mvn clean install

To start play with the REST Cli - you need to install jira-rest-cli-runner plugin, which is a main entry point and Rhino language provider.
If you want to try out JRuby or Groovy language support - then you should install jira-rest-cli-jruby or jira-rest-cli-groovy accordingly.

OR you could get all this artifacts here

jira-rest-cli-runner - Script runner, web-console, session mgmt admin interface, Rhino language support.
jira-rest-cli-jruby - JRuby support
jira-rest-cli-groovy - Groovy support
cli-examples - browse on GitHub, fetch locally and play around yourselves :)

NB: on my local dev environment I install all the plugins using atlas-cli since I launch JIRA using atlas-run / atlas-debug commands.

Ok, it's useful. How could I help?
Cool. First of all you are in the right place with the right attitude :)
The areas I'm struggling with on my personal Roadmap:

High priority: Documentation - both creating one and understanding what actually is missing. Even GitHub issues for this are welcome.
Medium priority: Upload plugin to Atlassian Plugin Exchange
Medium priority: Improving console-cli modes sample applications (JRuby, Groovy) - cleaning up, since I'm not expert in those languages; adding command line options.
Low priority: improving web-interface (more Ajax-like) etc...

PS: project is useful (at least for me) at the current stage and IMO documentation what is the only crucial piece missing.

Thin badge:

OpenJDK 7 - Developer Preview on Mac OS X

2011-03-12T12:23:00.003+02:00

All the glory to Oracle now ready with JDK 7 Developer Preview article. Thanks guys for sharing this:
<blockquote>

There is no Mac OS X version yet, but this link should get the Apple aficionados happy.

</blockquote>

Will try to setup and see how it goes. Looking forward to enjoy out fork/join framework :)
Will keep you posted.

Maven2 project auto-release [@GitHub] example

2011-02-25T01:31:00.052+02:00

What's about?
So, I finally got my head around maven-release-plugin maven plugin. And, yes, it's awesome.
I have not tried any kind of different/difficult scenarios yet, just plain basic ones.

So, what I wanted to achieve in the first place?

G1 - maven artifact auto-versioning on build success
1.1-SNAPSHOT on master branch/TRUNK gets promoted to 1.1 maven release, at the same time master/TRUNK becomes 1.2-SNAPSHOT
G2 - some kind of tagging/branching on build success
G1 support on CVS level via tags/branches
G3 - maven artifact deployment (from CVS repository to maven repository)
G4 - ability to launch it from CI tool

Visually that could be depicted in the following way:

Branching [source]

All the goals accomplished. Yada-ya-da!

Ok, enough reading. Gimme the code

See test-maven-release Github project

How-to?

G1, G2 - are accomplished by release:prepare goal.
Input needed: scm info
Input example:

<scm>
  <connection>
     scm:git:git://github.com/leonardinius/test-maven-release.git
  </connection>
  <developerConnection>
     scm:git:ssh://git@github.com/leonardinius/test-maven-release.git
  </developerConnection>
  <url>
     https://github.com/leonardinius/test-maven-release
  </url>
</scm>

Execution sample:


prusax:test-maven-release leonidmaslov$ mvn release:clean release:prepare
[INFO] Scanning for projects...
[INFO] Searching repository for plugin with prefix: 'release'.
[INFO] ------------------------------------------------------------------------
[INFO] Building test-maven-release
[INFO]    task-segment: [release:clean, release:prepare] (aggregator-style)
[INFO] ------------------------------------------------------------------------
[INFO] [release:clean]
[INFO] Cleaning up after release...
[INFO] [release:prepare]
[INFO] Verifying that there are no local modifications...
[INFO] Executing: /bin/sh -c cd /Users/leonidmaslov/Documents/workspace/test-maven-release && git status
[INFO] Working directory: /Users/leonidmaslov/Documents/workspace/test-maven-release
[INFO] Checking dependencies and plugins for snapshots ...
What is the release version for "test-maven-release"? (leonardinius-test:test-maven-release) 1.0: : 
What is SCM release tag or label for "test-maven-release"? (leonardinius-test:test-maven-release) test-maven-release-1.0: : myTag
What is the new development version for "test-maven-release"? (leonardinius-test:test-maven-release) 1.1-SNAPSHOT: : 
[INFO] Transforming 'test-maven-release'...
[INFO] Not generating release POMs
[INFO] Executing goals 'clean verify'...
[WARNING] Maven will be executed in interactive mode, but no input stream has been configured for this MavenInvoker instance.
[INFO] [INFO] Scanning for projects...
[INFO] [INFO] ------------------------------------------------------------------------
[INFO] [INFO] Building test-maven-release
[INFO] [INFO]    task-segment: [clean, verify]
[INFO] [INFO] ------------------------------------------------------------------------
[INFO] [INFO] [clean:clean]
[INFO] [INFO] [resources:resources]
[INFO] [WARNING] Using platform encoding (MacRoman actually) to copy filtered resources, i.e. build is platform dependent!
[INFO] [INFO] skip non existing resourceDirectory /Users/leonidmaslov/Documents/workspace/test-maven-release/src/main/resources
[INFO] [INFO] [compiler:compile]
[INFO] [INFO] Compiling 1 source file to /Users/leonidmaslov/Documents/workspace/test-maven-release/target/classes
[INFO] [INFO] [resources:testResources]
[INFO] [WARNING] Using platform encoding (MacRoman actually) to copy filtered resources, i.e. build is platform dependent!
[INFO] [INFO] Copying 1 resource
[INFO] [INFO] [compiler:testCompile]
[INFO] [INFO] Compiling 1 source file to /Users/leonidmaslov/Documents/workspace/test-maven-release/target/test-classes
[INFO] [INFO] [surefire:test]
[INFO] [INFO] Surefire report directory: /Users/leonidmaslov/Documents/workspace/test-maven-release/target/surefire-reports
[INFO] 
[INFO] -------------------------------------------------------
[INFO]  T E S T S
[INFO] -------------------------------------------------------
[INFO] Running me.doautomate.ExampleServletTest
[INFO] Tests run: 1, Failures: 0, Errors: 0, Skipped: 0, Time elapsed: 0.024 sec
[INFO] 
[INFO] Results :
[INFO] 
[INFO] Tests run: 1, Failures: 0, Errors: 0, Skipped: 0
[INFO] 
[INFO] [INFO] [jar:jar]
[INFO] [INFO] Building jar: /Users/leonidmaslov/Documents/workspace/test-maven-release/target/test-maven-release-1.0.jar
[INFO] [INFO] ------------------------------------------------------------------------
[INFO] [INFO] BUILD SUCCESSFUL
[INFO] [INFO] ------------------------------------------------------------------------
[INFO] [INFO] Total time: 3 seconds
[INFO] [INFO] Finished at: Thu Feb 24 22:02:49 CET 2011
[INFO] [INFO] Final Memory: 22M/81M
[INFO] [INFO] ------------------------------------------------------------------------
[INFO] Checking in modified POMs...
[INFO] Executing: /bin/sh -c cd /Users/leonidmaslov/Documents/workspace/test-maven-release && git add pom.xml
[INFO] Working directory: /Users/leonidmaslov/Documents/workspace/test-maven-release
[INFO] Executing: /bin/sh -c cd /Users/leonidmaslov/Documents/workspace/test-maven-release && git status
[INFO] Working directory: /Users/leonidmaslov/Documents/workspace/test-maven-release
[INFO] Executing: /bin/sh -c cd /Users/leonidmaslov/Documents/workspace/test-maven-release && git commit --verbose -F /var/folders/IA/IA2V6NfrE84m+Fcb-LJguU+++TI/-Tmp-/maven-scm-811668125.commit pom.xml
[INFO] Working directory: /Users/leonidmaslov/Documents/workspace/test-maven-release
[INFO] Executing: /bin/sh -c cd /Users/leonidmaslov/Documents/workspace/test-maven-release && git push
[INFO] Working directory: /Users/leonidmaslov/Documents/workspace/test-maven-release
[INFO] Tagging release with the label myTag...
[INFO] Executing: /bin/sh -c cd /Users/leonidmaslov/Documents/workspace/test-maven-release && git tag -F /var/folders/IA/IA2V6NfrE84m+Fcb-LJguU+++TI/-Tmp-/maven-scm-1676785108.commit myTag
[INFO] Working directory: /Users/leonidmaslov/Documents/workspace/test-maven-release
[INFO] Executing: /bin/sh -c cd /Users/leonidmaslov/Documents/workspace/test-maven-release && git push origin myTag
[INFO] Working directory: /Users/leonidmaslov/Documents/workspace/test-maven-release
[INFO] Executing: /bin/sh -c cd /Users/leonidmaslov/Documents/workspace/test-maven-release && git ls-files
[INFO] Working directory: /Users/leonidmaslov/Documents/workspace/test-maven-release
[INFO] Transforming 'test-maven-release'...
[INFO] Not removing release POMs
[INFO] Checking in modified POMs...
[INFO] Executing: /bin/sh -c cd /Users/leonidmaslov/Documents/workspace/test-maven-release && git add pom.xml
[INFO] Working directory: /Users/leonidmaslov/Documents/workspace/test-maven-release
[INFO] Executing: /bin/sh -c cd /Users/leonidmaslov/Documents/workspace/test-maven-release && git status
[INFO] Working directory: /Users/leonidmaslov/Documents/workspace/test-maven-release
[INFO] Executing: /bin/sh -c cd /Users/leonidmaslov/Documents/workspace/test-maven-release && git commit --verbose -F /var/folders/IA/IA2V6NfrE84m+Fcb-LJguU+++TI/-Tmp-/maven-scm-1157159798.commit pom.xml
[INFO] Working directory: /Users/leonidmaslov/Documents/workspace/test-maven-release
[INFO] Executing: /bin/sh -c cd /Users/leonidmaslov/Documents/workspace/test-maven-release && git push
[INFO] Working directory: /Users/leonidmaslov/Documents/workspace/test-maven-release
[INFO] Release preparation complete.
[INFO] ------------------------------------------------------------------------
[INFO] BUILD SUCCESSFUL
[INFO] ------------------------------------------------------------------------
[INFO] Total time: 1 minute 
[INFO] Finished at: Thu Feb 24 22:02:59 CET 2011
[INFO] Final Memory: 14M/81M
[INFO] ------------------------------------------------------------------------
prusax:test-maven-release leonidmaslov$

G3 - is accomplished by release:perform plugin. The plugin checkouts the tag/branch from the CVS system, builds it and deploys to maven repository.
NB: relies on release:prepare being called first (needs release.properties file).
Input needed: scm info, distribution management
Input example:

<scm>
    <connection>
      scm:git:git://github.com/leonardinius/test-maven-release.git
    </connection>
    <developerConnection>
      scm:git:ssh://git@github.com/leonardinius/test-maven-release.git
    </developerConnection>
    <url>https://github.com/leonardinius/test-maven-release</url>
</scm>

<distributionManagement>
    <site>
        <id>site-id</id>
        <name>site-name</name>
        <url>scp://host/path/to/repository/</url>
    </site>
    <repository>
        <id>repo-main</id>
        <name>Repository Name</name>
        <url>file://tmp/</url>
    </repository>
    <!-- use the following if you ARE using a snapshot version. -->
    <snapshotRepository>
        <id>repo-snapshot</id>
        <name>Repository Name</name>
        <url>file://tmp/</url>
    </snapshotRepository>
</distributionManagement>

G4 is accomplished by launching maven in the batch mode.
Example:
```
mvn clean --batch-mode  release:prepare release:perform
```

Conclusion
Adding automated maven project versioning is not so hard and could be easily automated.

#oldstuff - 3. Mini seminar 2 - Mylyn

2011-02-21T23:25:00.005+02:00

Check out this SlideShare Presentation:

3. Mini seminar 2 - Mylyn

View more presentations from Leonid Maslov.

This presentation was part of the mini-seminars periodical sessions I've prepared and led in my spare time while being employed at Exigen Services. Kudos, guys!
Since these presentations were spare time hobby - I've decided to share them :)

#oldstuff - 2. Design patterns. part #2

2011-02-21T23:21:00.007+02:00

Check out this SlideShare Presentation:

2. Design patterns. part #2

View more presentations from Leonid Maslov.

This presentation was part of the mini-seminars periodical sessions I've prepared and led in my spare time while being employed at Exigen Services. Kudos, guys!

Since these presentations were spare time hobby - I've decided to share them :)

#oldstuff - 1. Mini seminar intro

2011-02-21T23:20:00.004+02:00

Check out this SlideShare Presentation:

1. Mini seminar intro

View more presentations from Leonid Maslov.

This presentation was part of the mini-seminars periodical sessions I've prepared and led in my spare time while being employed at Exigen Services. Kudos, guys!

Since these presentations were spare time hobby - I've decided to share them :)

Atlassian Plugin SDK - how to provide optional selenium tests

2011-02-21T00:20:00.007+02:00

You probably are familiar with one of the most popular web UI testing tool - Selenium and it's most famous tool Selenium IDE.

I don't want to read all this, give me the code

You could find all the code you need on github.

Sorry for not embedding all the code in the article. That way the article becomes clumsy unreadable monster. If anyone could recommend me some collapsable/expandable solution - I would appreciate it a lot.

Why do I need to bother with yet-another-testing-tool?
The main advantages or features of the tools are (imo):

real UI behavior testing - tests are run inside the real browser and will fire/use real end-user key, mouse, etc events
ability to export the test in whole bunch of programming languages (java, python, ruby etc ...)
ability to run tests on multiple target web-drivers (firefox, internet explorer, etc ..)
Selenium IDE (mentioned earlier) is really nice tool which allows to create simple tests cases very fast, in few minutes or even seconds, literally
basically the only way to automate difficult Ajax/RIA component testing on different target environments; other httpUnit like approaches are not good enough

How could I automate that?
Selenium has child project - Selenium Remote Control which could be used.
Quote and image from Selenium Remote Control site:

Selenium RC comes in two parts.
A server which automatically launches and kills browsers, and acts as a HTTP proxy for web requests from them.

Client libraries for your favorite computer language.

Lucky to us, selenium has maven artifacts available to be used to plug-in into Atlassian PDK build procedure.
Goals:

make this optional (enabled by default) option
selenium-rc server side should be forked before integration tests will be executed
selenium-rc client side tests should be executed as part of the integration test suite
selenium-rc should be revoked upon integration-test completion

Solution:

is achieved by default maven profile (activated by !property)
is achieved by maven - specifying selenium-rc start-server goal to be run on pre-integration-test phase
to achieve this please refer to Atlassian article - Writing Integration Tests. Basically that means - you should put your code inside src/test/java/it/ folder and specifying proper locatest.properties file.
is achieved by maven - specifying that selenium-rc stop-server goal should be executed on post-integration-test phase.

To get pom.xml file snippet - see this pom.xml gist

How to check:
execute atlas-integration-test command from the project home folder.

Recommendations:
Atlassian FuncTestCase (Jira in this case. There should be one for other Atlassian products as well) contains a lot of useful utilities and setup methods, which personally I've found very useful. So, the idea is to reuse it as much as possible. That could be done in you create your own abstract base MySeleniumTestCase which would contain FuncTestCase instance as testHelper member field and would make it available via getHelper() or other similar way.
Still, you should use your own loginAs method on top of Selenium, cause helper object will have different (not in-browser) http session.

Benefits:

ability to reuse utility methods (such as create new custom fields, restore from xml dump etc ..)
ability to configure your Selenium test against jira integration test environment - using helper to obtain JiraEvironmentData's host, port, context path etc..

Unlucky I'm not able to share that piece of code, so you should to create it on your own. Still, it should be pretty straightforward, since all you need to do - is do declare member field and delegate to is in your test setUp/tearDown, as well as using it to configure your Selenium test host, port etc ..

Problems with Continuous Integration
Ok, so we think we have automated our process so far and it runs locally. On my machine it launches internet explorer and executes the test. Awesome. The next thing is putting the test on our CI, which should be pretty straightforward, right? Wrong!
Problems with CI:

CI slave might be Linux terminal machine, no UI and therefore no real launch browser capabilities.
There is a bug with previous pom.xml file. Yes, the profile is optional, however Atlassian PDK framework execute all the it/ tests regardless what type of tests are they, therefore all the Selenium tests will fail.

The former one is solvable by using Selenium-RC Headless X11 with XVFB maven goal. The solution is to execute the goal on Linux machine and skip one on a Windows box. That could be achieved by additional profile setting xvfb skip flag to true and activated on windows platform only.

The later one could be solved by specifying AMPS plugin testGroups. Using that we could specify different sets of independent functional tests, located in different locations (it/functests* and it/selenium/*). We could use configuredTestGroupsToRun property as well, but I decided not to.

So, as a result we have completely independent sets of integration/functional tests, with ability to launch them all-together or pick just one of them.

The code

Please see updated_pom.xml file.

HowTo: Reusing JIRA gadgets on the issue operations

2011-01-10T20:26:00.000+02:00

Hi everybody,

Problem description
Recently I had to implement the following use case:

Customer has very strict defined issue creation policy and wants to control the every bit of issue creation (things like enhanced priority etc support).
The main focus is on usability and rich user interface support.
Basically it's multi-step procedure.

After reviewing customer requirements it was decided to go with custom Dashboard Open Social Gadgets.

So the initial solution mockup looked like that:

The wizard-gadget itself was tricky to implement as well, but's an another story.

For the sake of this article let's consider that we already have out plugin implemented with the REST services to create / update, read issue data.

Transforming gadget to it an Issue operation

At some complicated updates are required from the user interface.We want to reuse the wizard created previously. The way I decided to proceed is to implement issue operation. So, that does it mean for end user? See the mockup below:

But! Jira does not provide this kind of functionality. So, we need to implement this on out own.

First of all we will create our own issue operation (useful information web-fragments, updating jira plugins for jira 4.1).

1. See the atlassian-plugin.xml snippet for that

So, we have implemented new Issue Operation which will open web-work action page we have not defined yet so far. Let's do so.

2. See action code.

3. As you may see, MyWebAction depends on GadgetRequestContextFactory and GadgetViewFactory which are OSGi components, so you should import them as component dependencies.

4. Let's add add web-action definition to attlasian-plugin.xml

5. Lets specify /templates/mywebaction/render.vm template.

And we are finally done. Note xx.yy.zz.arguments.issueId javascript assignment. It's done to pass in arguments to gadget instance.

6. The following code could be used to get the overall idea:

Implementing issue operations the Jira way - using AJS popup dialogs

At this stage you have working and complete issue-gadget-operation. Cool? Yeah!

anyway, purists would argue that it's not inline with Jira Look&Feel. At the moment it does open as separate action, which is not inline with the rest of issue operation actions.

Basically we want our action too look like that (in the popup dialog as all the other actions):

Is it possible? Yes, it is.

To make our wizard to appear as popup dialog box we need to include some javascript to browse issue page. The easiest way to do so is via servlet filter.

1. See atlassian-plugin.xml snippet for defining the servlet filter and javascript resource.

2. See the filter definition itself

3. And the AJS javascript magic itself

And finally, all the pieces are tight together and you have Jira-AJS-POPUP style issue operation. And the cool part is - it reuses gadget instance created previously.

Hopefully you will find this stuff useful.

JIRA workflows -Postfunction exceptions, database commits and Pretty workflow exception

2010-12-25T01:57:00.003+02:00

Hi again,
For all of you who happens to be catholic - Happy Xmas futurama style.

So, back to coding again.

I assume most of you are familiar with JIRA post functions. So, would like to share my 0.02$ on the subject though. Boring stuff.

OS Workflow

The less known things about post functions are:

Workflow, steps and history are stored in different database tables than issue data. So Jira uses database commits to enforce data consistency.
Post functions might throw WorkflowException exception and workflow manager will rollback transaction in this case. Actually if the other exception will be thrown the transaction won't be commit (see Ofbiz Workflow et cetera), but imo it's implementation specific nuances and I wouldn't recommend that
In case you use MySQl - please ensure you have configured it properly - to use transactional database engine. If you haven't - let's say you do Resolve transition from status Opened -> to Closed and exception occurs. The workflow engine will think (no transaction rollback) the issue is closed, while the rest of the JIRA (issue view etc, issue.getStatus()) will think the issue is still Opened. Happened with me :).

The pain

I've found myself to put some conditioning into my custom wf functions, like if smth has ben selected in the transition screen view - the function should fail. The same could be achieved my using wf validators and for sure! would be more correct solution - it's the type of component that should do the validation thing. However - think about it - I have 1 thing to do during transition - it's one post function. And I have 3-8 validations for it, like is it configured properly, are all fields entered from the UI etc. It would be more flexible to do it using wf postfunction and 3-5 validations, but it's so much more work (at least configuration and maintenance). So I decided to stick with 1 postfunction approach.

The way to implement this would be to throw exceptionn during the wf transition. In the regular case I would throw smth like IllegalArgumentException, but for reasons mentioned earlier I decided to convert it to general WorkflowException. The thing with it is - in case it does contain exception cause (and I like to preserve stacktrace as long I can) - it messes the exception getMessage() result and user on-screen output is smth like "rootCause: Please specify ...", which is ... wrong.

So, the approach i decided to go with is:



//... postfunction class declaration
{

public void execute(java.util.Map transientVars, java.util.Map args, com.opensymphony.module.propertyset.PropertySet ps) throws com.opensymphony.workflow.WorkflowException {
try
{
 //..
 doStuff();
}
catch(XYZIssueException e){ throw new WorkflowException(e); }
//...
catch(IllegalArgumentException e) { throw new PrettyMessageWorkflowException(e); }
}

// ... rest the postfunction
}

// And the pretty lady herself
class PrettyMessageWorkflowException extends WorkflowException
{
// ------------------------------ FIELDS ------------------------------

 private final String prettyMessage;

// --------------------------- CONSTRUCTORS ---------------------------

 public PrettyMessageWorkflowException()
 {
     super();
     this.prettyMessage = null;
 }

 public PrettyMessageWorkflowException(String message)
 {
     super(message);
     this.prettyMessage = message;
 }

 public PrettyMessageWorkflowException(Throwable rootCause)
 {
     super(rootCause);
     this.prettyMessage = null;
 }

 public PrettyMessageWorkflowException(String message, Throwable rootCause)
 {
     super(message, rootCause);
     this.prettyMessage = message;
 }

// -------------------------- OTHER METHODS --------------------------

 @Override
 public String getMessage()
 {
     if (this.prettyMessage == null || this.prettyMessage.isEmpty())
     {
         return super.getMessage();
     }

     return this.prettyMessage;
 }
}

It saves a day as usually :)

Atlassian SAL Plugin Settings

2010-12-21T00:19:00.000+02:00

Has just discovered SAL PluginSettings entry point. No more painful plugin configuration storage for me. Good-bye propertyset-storage-toolkit, you was so young, so brave, so dumb :)

However, if you have more sophisticated needs than PluginSettings API, smth more than get, put, remove and you want all the power PropertySet could provide you - like lookup, more sophisticated data types, more sophisticated scopes (not like Global and Project-level only, but anything you could map to id, prefix and entityName) then you still might take a look into propertyset-storage-toolkit.

Note: please you SAL PluginSettings until you are really forced not to.

Jira - let's code update, edit issue operations (CRUD)

2010-12-19T02:10:00.001+02:00

If you are familiar with JIRA and it's API then you might agree, that some easy operations become a little difficult when it comes to coding (all the workflows, overall JIRA stack understanding, configuration items, integration with other systems etc).

The same is especially true when it comes to creating / updating JIRAs issues, which definitely is backbone for issue tracking systems. Some might argue that this type of things should be done by JIRA itself only. And I couldn't agree more. However JIRA tend to become issue-tracking puzzle-building machinery, where JIRA itself is LEGO puzzle only, so plenty of crazy stuff "you better shouldn't do this" or "you are not supposed to do that" is very popular in the JIRA plugin development world.

So, back to issue creation/update problem. This is kind of task I code almost in every plugin. And anyway, it bugs me every time :) It's difficult for me to remember all the nuances and it's context dependent as well. Lets see that JIRA development form says about this - create, update. Seems already solved problem, isn't it? However it takes me a plenty time to solve it out if I don't have out-the box solution in my local repository.

So, this time I decided to share a little code with you, and just a reminder to myself as well (I tend to forget things all the time).

And here comes the implementation itself:

atlas-cli IntelliJ support

2010-12-13T01:33:00.005+02:00

I assume most of you, Atlassian Plugin SDK consumers are well familiar with atlas-cli command.

After reading Down's Browns great article titled Maven without all the slowness - now with IDEA support decided to give it a try. I wish it would be so straightforward :) See the bugreport. It seems the cli:idea artifact is broken.

So, of course, you could get the sources and compile it by your own. However I have found what JBoss repository have compiled Intellij Artifact itself, get it here. Copy the jar file in the plugin directory (in my case it's /Users/leonidmaslov/Library/Application Support/IntelliJIdea10CE/) and viola, you are done.

Let'see it in action, configuration first:

And now 3 more build commands/shortcuts are available

Enjoy :)

HOW-TO: propertyset-storage-toolkit - Jira configuration data storage made easy

2010-12-12T02:15:00.000+02:00

Q: What is your favorite type of job? Atlassian Plugin SDK? A: Sure!
Q: How exactly you like to code? A: As little as possibe!

Here it comes - propertyset-storage-toolkit with as bare minimal API set as it could be.

Let's see what the standard project for me as Jira plugin developer is? The answer probably would be - to make this thing A, that thing B and yeah, almost forgot - to make this configurable, there definitely should be a switch et cetera and this should be administration UI dialog.

And yeah, there are a lots of data persistence options available, but they all too complic-API-cated. So after scratching my head several times I remembered one piece of code I've written before while being employed by some company and decided to re-implement the idea from the scratch on my own and to use in my current project assignment.

So, to illustrate how EASY is to do plugin configuration from now on, see the code snippet below


private final StorageService storageService; // IoC by framework via Constructor Dependency Injection
 ...

 // data fields
 private boolean adminOnly;
 private String helloText;
 private String welcomeText;

 private StorageFacade getStorage()
 {
     return storageService.actionStorage(this);
 }


 private void loadConfig(StorageFacade storage)
 {
     setAdminOnly(storage.getBoolean(ADMIN_ONLY));
     setHelloText(StringUtils.defaultString(storage.getString(HELLO_TEXT), 
         "Hello, "));
     setWelcomeText(StringUtils
         .defaultString(storage.getString(WELCOME_TEXT), "World"));
 }

 private void saveConfig(StorageFacade storage)
 {
     storage.setBoolean(ADMIN_ONLY, isAdminOnly());
     storage.setString(HELLO_TEXT, getHelloText());
     storage.setString(WELCOME_TEXT, getWelcomeText());
 }

And voila, it's as easy as this, no strings attached.

Please take a look into README file for more details.

And, please, do not shy to contact me in case of any problems, issues, thoughts or comments. Your feedback is always appreciated.

See ya ;)

Announcement: propertyset-storage-toolkit plugin has been released

2010-12-12T01:52:00.000+02:00

Hi folks,

I assume many of you are Jira developers and I think I will express the overall feeling that configuration data storage in the Jira is a tricky moment many of you spent a lot of time on.

propertyset-storage-toolkit - is plugin to rescue from boilerplate configuration data storage coding. The plugin itself is more like library for configuration data persistence and retrieval (and a sample application along to start rolling).

Hopefully you will find it as useful as I do. The reason for me of creating one is - to help newcomers come on board the Jira Plugin Development process.

This said, here theere are list of feature available:

Database persistence using PropertySet for underlying data storage
Has no additional JIRA dependencies and provides enhanced PropertySet API sub-set
Could be used both as JAR dependency bundled with your plugin OR as separate plugin, depending one your current needs.
Provides a rich set of out-of box list of Scope objects. (Scope defines the sandbox of the PropertySet storage instance). At the moment following configuration options are supported

Named configuration scope (Example: My Gadget1 Configuration)
Action configuration scope (Example: your plugin may contain Administration action, which may require to persist and access administrator input).
Project configuration scope (Example: you may associate configuration information with project object)
Issue configuration scope (Example: you may associate configuration information with issue object)
The list of scopes is not limited and you may come-up with your own Scope
Apache 2 license

If you are interested to use, provide a feedback, review or any type of comments - you are always welcome to contribute.
Readme file is your place to start.
Hopefully with the help of you guys, will end-up with something really useful.

PS> Really enjoyable and productive weekend at the end of the day

Jira - How to Include JSS and CSS resources on every page

2010-12-07T02:33:00.003+02:00

Hi all,

Disclaimer 1: The post and approach is 99% based on Jira Behaviors Plugin source code.

Disclaimer 2: The approach described here should be considered a hack/workaround and does not inline with Atlassian Jira Plugin structture and maintainability.

Update: See the comments section. I would recommend better approach using Servlet Filter.

You may find yourself requiring to either show/hide some part of Jira UI elements, or extend JIRA built-in edit/search capabilities. If it's custom field we are talking about - you might consider to implement Jira Custom Field (See Quick Custom Field Types Primer on render options and capabilities).

For some other cases - like Jamie Eclin awesome Behaviors Plugin, or like in my example - hiding some always visible JIRA UI built-in elements - you might consider different approach to implement. Your options basically are:

Hack and modify velocity/JSP file whatever does render the element in context. It even might be system level plugin - like system Issue operations. In some cases it might require JIRA instance restart and even re-compilation and re-deployment. And thus we will increase maintainability technical debt a lot.
We could manually configure Jira Announcement Banner to include required resources on every page. Thus it will be one-time configuration option and it will be our responsibility to do so and maintain that piece of JSS/CSS snippet. In case our product consists of several plugins - we will endup with long list of configuration items + 1 more unrelated item - which is kind of implementation details only item. So, not so good afterwards (good enough for internal stuff, not so good for giving away).

So, after reconsidering those options - I've chosen the second one + basically the Jamie Eclin approach on Behaviors Plugin. The main reasons to do so were:

Not really a complete hack - could be implemented as valid Atlassian Plugin v2 plugin
Could be automated - no additional configuration needed
Source code for Behaviors Plugin is provided :)
See pt 3.

Implementation

1. Component
Simple interface with the only method - void setup(); Serves only as an entry point for resource injection during Jira startup procedure.

2. ComponentImpl
The actual component implementation. The purpose is to check whether Jira announcement banner already contains required JSS/CSS and it doesn't the inject them to the banner contents.


public class ComponentImpl implements Component, Startable {
    private static final String HIDE_CREATE_ISSUE_MARKER = "XYZ Plugin Marker - Do not modify!";
    private static final String CREATE_ISSUE_MARKER_START = "<!-- --Start--[" + HIDE_CREATE_ISSUE_MARKER + "]--Start--";
    private static final String CREATE_ISSUE_MARKER_END = "<!-- --End--[" + HIDE_CREATE_ISSUE_MARKER + "]--End-- -->";

    private final ApplicationProperties applicationProperties;

    public ComponentImpl(ApplicationProperties applicationProperties) {
        this.applicationProperties = applicationProperties;
    }

    @Override
    public void setup() throws RuntimeException {
        String alertHeader = applicationProperties.getDefaultBackedText(APKeys.JIRA_ALERT_HEADER);
        String injectedAlertHeader = installHideCreateIssue(alertHeader);
        if (!StringUtils.equals(alertHeader, injectedAlertHeader)) {
            applicationProperties.setText(APKeys.JIRA_ALERT_HEADER, injectedAlertHeader);
            //applicationProperties.setString(APKeys.JIRA_ALERT_HEADER_VISIBILITY, EditAnnouncementBanner.PUBLIC_BANNER);
        }
    }

    @Override
    public void start() throws Exception {
        setup();
    }

    
    private boolean isIncludesCreateIssueMarker(String input) {
        return StringUtils.contains(input, CREATE_ISSUE_MARKER_START)
                && StringUtils.contains(input, CREATE_ISSUE_MARKER_END);
    }

    String getInjection(String comment, String cssDefinitions, String javascript) {
        StringBuilder injection = new StringBuilder(CREATE_ISSUE_MARKER_START).append(comment).append(" -->");

        if (StringUtils.isNotBlank(cssDefinitions)) {
            injection = injection.append("\n<style type=\"text/css\">\n")
                    .append(prepareCleanInject(cssDefinitions))
                    .append("\n</style>");
        }

        if (StringUtils.isNotBlank(javascript)) {
            injection = injection.append("\n<script type=\"text/javascript\">\n")
                    .append(prepareCleanInject(javascript))
                    .append("\n</script>");
        }

        injection = injection.append(CREATE_ISSUE_MARKER_END);

        return injection.toString();
    }

    public String installHideCreateIssue(String alertHeader) throws RuntimeException {
        if (!isIncludesCreateIssueMarker(alertHeader)) {
            alertHeader = new StringBuilder()
                    // PROOF OF concept Items - it really should be done with Permission Schemes, not this hacky way
                    .append(getInjection("", "div#create-issue, div#createItem, ul#opsbar-edit li a#editIssue { display : none; }", ""))

                    .append("\n")
                    .append(StringUtils.defaultString(alertHeader)).toString();
        }

        return alertHeader;
    }
}

3. atlassian-plugin.xml snippet
Component and ComponentImpl registration snippet


<component key="myJSSCSSInjectionSnippet" name="My CSS/JSS injection snippet"
    class="com.xyz.ComponentImpl">
    <interface>com.xyz.Component</interface>
</component>

and Voile, we are done :)

My source code highlight updates - blogger.com

2010-12-07T01:53:00.006+02:00

Hi everyone,

After updating my template I noticed that I have JS source code highlighting broken.
What do do - setup once again :)

So,m after a little research I found one I liked the most - Highlight.js - See Demo Page.

So, step by step instruction on putting this in Your blog:

Go To Design
Goto Edit HTML
</head> tag

Place the code just before the closing </head> tag

<script src='http://yandex.st/highlightjs/5.16/highlight.min.js'/>
<link href='http://yandex.st/highlightjs/5.16/styles/school_book.min.css'
           rel='stylesheet'/>

<script language='javascript'>
hljs.tabReplace = &quot;    &quot;;
hljs.initHighlightingOnLoad();
</script>

Save the template - the result you mas see above. Note: put you code in <pre><code class="html"> tags (or whatsever is your language name)

Atlassian Plugin SDK / maven proxy issues

2010-11-10T14:15:00.020+02:00

Recent problem: maven repositories on https behind proxy.


<?xml version="1.0"?>
<proxy>
    <id>http.XXX.global</id>
    <active>true</active>
    <protocol>http</protocol>
    <host>XXXt</host>
    <port>8080</port>
    <nonproxyhosts>localhost|127.0.*.*|191.168.*.*</nonproxyhosts>
</proxy>   
 <proxy>
<id>https.XXX.global</id>
<active>true</active>
<protocol>https</protocol>
<host>XXXt</host>
<port>8080</port>
<nonproxyhosts>localhost|127.0.*.*|191.168.*.*</nonproxyhosts>
</proxy>

and it failed until I switched http proxy of (set active to false)

Unit tests, What they should be like? (2)

2009-10-16T02:27:00.016+03:00

My humble opinion only

Every test should be story (scenario) test.

It could either be User story (user logins in, does A, soes B), Engineering story (the application developer will create component A, pass in reference for component B, will invoke #xxx(), #yyy() in turn and will await for something). It doesn't mean, it couldn't test just like one function, ore just one invocation. It's about check separation to different coherent scenarios.

Tests should not calculate.

Test logics is logics too. It easily becomes clumsy, buggy and all the stuff like that. It should be supported too.

a) If one (wrong!) application logics value equals to some other (wrong too!) value it doesn't mean the test is true (false positive) .

b) the efforts to support tests should be minimized

c) tests should be readable. assertEquals("concrete-value", calculate()) is the way more readable than assertEquals(calculateA(calculateB()), calculateC(calculateD())) .

Tests should be as readable as they could be.

In my opinion automated tests could be considered as separate maintenance project. Test code is not throwaway code, otherwise it could be 1 time successfully executed and safely removed.

Test code is the most fragile part of the application, but it needs hands-on all the time. The developers tend to write unreadable test code and when, obviously, don't want to spend lot's of time maintaining it. As a result, tests stories degrade over the time and become useless at some point (no one knows for sure that the hell does this story test check...). So, the more readable the test would be, the more time it would take to degrade to such as state.

Monogamy

Testcase, test story (continue this sequence on your own) should stick to one particular component, one particular User/Usage Use Case (UUC) and should not influence or gets influnced by any other subsystems, logics, components, whatever. The test should create Well Known, Hardcode, As Small as Possible sexy Universe and marry and have children with her.

a) Well-known (see Don't calculate; Readable)

b) Hardcode (see Don't calculate; Readable)

c) Universe (all the space, objects as needed).

As the result, no affairs are allowed. If TestA testCase fails, only A should be blamed. No chances of getting children with B should be allowed. If test scenario (story) testScenarioA fails, there shouldn't be option for testScenarionB to be blamed. Doing so, we will easily find the root cause and understand the possible impact on both the application and business behind it.

Tests should fail

See Readable. If the test will not ever fail, that is the reason for keeping it? It will never fail anyway (run once and purge from your eyes away). The test should test

a) will probably fail someday (or we do change this component a lot and rely on it tightly)

b) something having high value (business probably) we want to have a eye on it. No one likes receiving urgent emails, night calls or working Sundays' eves.

Test "stories" should be as minimal as possible, without any any additional assumptions (temporally true).

a) The smaller the "story" (the code) is the less maintenance effort will be required.

b) Test developer should not add any non-business/non specified assumptions into unit tests. Consider this, POJO_A contains #toString() method delegating to Object#toString(). The test developer might considering to add additional assumption that #toString () would return a string with '@' character. Since this is not a requirement at all, at some point we might considering to provide some meaningful (business point of view) implementation. Thus we will not only to write new xUnit tests, but will also spend some effort due to previous assumptions maintenance (those in this particular cases could be omitted).

If for some reason the test story / test setup is unreasonable big (put your threshold here) you might consider refactoring, or starting from some point - complete rewrite of this particular application logics/or test part. It's like a doorbell, jing, the application code will rot soon.

Will reread this twice and continue the listing.. Someday :)

Automated Tests, Unit tests, Pros (1)

2009-10-16T02:07:00.008+03:00

My humble opinion only

Improves design.

At least +1 client, decoupling from the Application.main([]) method.

Regression tests, cost of running them VS cost of manual retesting.
Difficult to test by mortal hand areas

(multi-threading issues, timing, DB connectivity loses). As extra point, the cost of such hand-made tests are extremely high and won't disappear on 2nd/3rd/... run. _Not my opinion, but I share this point of view_

Easier to explore internal (unclear from user point's of view) Grey Areas or Unpredictability Corners (more commonly known as old plain Boundary Values).

It's not always clear from User Story or User Interface that the boundary values would be, or other subsystem boundary values influence on this particular component. Or the boundary value application will eventually hit is very-hard to prepare by mortal hand. For example tester should repeat Long.MAX_VALUE operations to hit this one.

Anti-Plag API design

2008-12-16T09:25:00.004+02:00

Trying to get some hours of spare time to make a API design to:

index documents
remove document from the index
Query document

Requirements&Considerations:

API should be domain problem agnostic
It's not worth trying to left Lucene behind the scene (I mean hide it).

!!!! XFT is another TaxonomyOverLucene framework with structural document processing features ( See demo). Worth to investigate it a little.

I will post API extracts to make some kind of the public plasce to judge it :)))))))))))

Antiplagiarism mockup session continued

2008-12-03T03:38:00.001+02:00

Mockup session continued. Mockups were published here.

Preview:

From Drop Box

Design was done in the wonderful Balsamiq Studo, which is completely free for OSS projects as mentioned here. This is really a superior tool to draw mockups. Thank Balsamiq Studio.

Mockup files: