I am trying to manage my dependencies of my play framework application. I am trying to remove/update some jar that exists in my classpath. Does that even possible?
It would be great to get an explanation of how does play framework manages classloaders. I understood that in order to obtain hot redeployment of code in development mode Play is restarting the application classloader and loads all classes again. Does it right?
Thanks a lot!
It's not really Play managing the hot reloading but rather the Play SBT plugin.
A Play application's main method is in play.core.server.NettyServer. The method receives a reloader of type SBTLink and each time a request hits the server, the application delegates the task of checking if there have been any code changes to the reloader.
If there has been a code change then the existing classloader is removed and replaced with a new one which includes the updated bytecode. The application is restarted but, of course, the JVM is not.
Play is not designed to provision dynamic swapping of Jars at runtime. For this, you would need to look at things like OSGi (which I would not personally recommend after a few years of working with it).
Related
Currently, with my Intern setup, I'm using an unbuilt Dojo build when running my Intern tests; like, for example, a test module loads app/ProductModuleA, and ProductModuleA references and loads dojo/request. I need to have the dojo/request.js file in the appropriate directory structure in order for the module to be resolved without errors and therefore the test to be able to run. Our product code does use a built dojo.js file and our previous DOH tests were able to use this, too, without any issues--I don't understand how that worked because I don't know anything really about building Dojo.
I know I've seen snippets in various Internet forums (like here) and the Intern User Guide that Intern supports source maps, which I guess suggests it's possible to use a built dojo.js file in conjunction with running Intern, but I haven't found anything at all in detail. Insights, or pointers to documentation or examples that so far I haven't been able to find?
One of the benefits of AMD is that you don't have to do anything special to your code when switching between a built and unbuilt Dojo. The first time you load a dependency using an unbuilt Dojo, the loader requests it over the network and then caches the result. Subsequent loads use the cached dependency. The loading process works the same with a built Dojo; the main difference is that all the modules built into the built Dojo are pre-cached. The loader doesn't have to request them over the network the first time because they start out in the module cache.
For Intern to use a built Dojo, you just need to make sure you're using the built Dojo as your loader during tests. You can do this by setting the useLoader option in your Intern config.
I tried what Jason suggested and it still didn't work--I was getting 404s for a Dojo_ROOT.js module, though nothing in tests or product files explicitly load that. I'm sure this is due to something unique in my product's build environment. That's okay, I will just use the Dojo source for now and return to this later.
I had this question ever since I learnt about Erlang and it's possibility for hot replacements of separate modules and even functions on a production server.
We are developing a project in Java for Glassfish 2.1. It's basically an .ear file consisting of a bunch of .war modules.
So, every time, I make a minor change in the code of one of the modules, in order to check that change I have to redeploy the whole ear on my development server. Is that really how it's supposed to work?
Basically we have several levels of replacement - are they even possible?
The war module. Happens during a big number of commits by several developers. Is it possible to replace just this module on the working server without redeploying the whole ear file.
The java classes (action handles and etc). Is it possible just to replace the bytecode of the classes that were changed?
The jsp pages. I think they are launched indepently by the special jsp interpreter in glassfish. So this one definitely should have the possibility of the hot reload of a single changed page.
I know I can perfectly replace the different parts of HTML design, such as css or js files, or image files. But what about the bytecode files described above?
As I am using eclipse and just set up a dropwizard server. On the command prompt I typed in java -jar target/hello-world-0.0.1-SNAPSHOT.jar server hello-world.yml and is running. Yet when ever I make a change to my eclipse file, like changing the yml file for example it doesn't update. I have to crtl+c and re-run what I typed in above. My question is, is there a faster way of testing so that it updates every time I change something or I just have to deal with the testing. Thanks.
Run from within the IDE
Different Java IDEs permit more efficient workflows. For example, in an IDE you can run up your application using a Runtime Configuration that executes your Service.main() method with parameters of server hello-world.yml. This will save you endless Maven builds.
Unfortunately, with Eclipse the hot swapping of code changes is often cumbersome, so I would recommend that you consider Intellij which is more reliable when it comes to hot swapping code. Even then hot swapping can be risky.
Sometimes a restart is unavoidable
That being said, in your situation hot swapping won't help. You are changing the startup configuration file which is only read at startup. You will have to restart to see the changes unless you create your own dynamic-refresh-on-file-hash-change mechanism (not advised).
One alternative is to put much of your configuration testing in unit tests and verify that your code is responding as expected.
Static assets give an optimal workflow (no restarts)
You may encounter a situation where you only want to change static assets (like JavaScript files) in which case Intellij will allow you to simply recompile on the fly and will copy the changed assets into the /target directory and have them immediately picked up by Dropwizard without a restart.
If you wanted to go one step further you could enlist the services of Grunt.js so that it continuously monitored the src/main/resources/assets (or similar) for changes and then automatically update your /target for you. Again, Intellij will autosave on focus change so this would lead to an optimal workflow where you change the asset, wait one second, refresh browser and see the immediate result.
I wrote a lengthy blog article covering Dropwizard and Ember Data a while ago if you want more details on this approach (and single page web application development in general).
I have developed an NPRuntime plugin, everything is ok, but it not initialized in Safari. I registered the plugin in MozillaPlugins registry key, it works fine for Chrome/FF/Opera, but Safari writes the the plugin is missing, although I found my plugin in the list of isntalled safari plugins. What am i doing wrong ?
That's real hard to say. Safari on windows isn't supported real well by Apple, so I don't bother supporting it much myself. If it's really important I'd recommend adding logging in all your NPP_ methods (and your NP_ functions too) and see what is getting called; find out if it even loads your plugin. Often if something doesn't go as the browser wants during startup it will act like it didn't find it at all.
You could also use Process Monitor to see if it is trying to load the file or not.
I want to work on intelliJ on my webapps and I don't know how to hot swap code while working.
For example while I was working in Eclipse when I edited jsp files Eclipse automatically, instantly swapped the file so when I refreshed the page my changes were there
When I change class code in Eclipse it worked a bit longer because he republished the app but did it automatically and instantly.
I saw that intelliJ in the runtime configuration has an option 'how class swap'. I did check it but nothing is happening. I had tried compile, make, save and everything else and nothing is happening. I had to reload the app and I had to do dis manually. Secondly intelliJ reloads EVERY application in my webapp directory. I have them a lot so it taking ages. how can I turn on hot swap?
Hotswap only works in debug mode. So you need to connect to your webserver through a debug configuration. Then, after compile either the project or at least the class with the modifications, IntelliJ tries to hotswap that class.
This only works for minor changes. For example, creating new methods on the fly is not possible using this way.
Hotswap works with exploded artifacts on Update action. If it doesn't work with your project, contact support for help and provide the project to reproduce it.