fullscale / evo Goto Github PK

Clean up logging messages (Don't log variables that are known to be long strings such as resource code, etc. Investigate using logback vs. log4j. See how we can capture elasticsearch specific logs such as the slow query log.

When a new controller is created in the IDE, the boilerplate code is generated in dialogs.js. This was done because when a new resource is created, a blank tabe is created by default.

What should happen is we should be doing a realtime GET on that resources and loading the actual content from the index. The ideal place to create this boilerplate code would be in the AppsController (createResourceInDir()) but the tabview code (javascript) needs to be refactored to support dynamic fetching of newly created resources.

Improve support for user libs and configuration files.

When a session times out and the originating browser session attempts to access a page we get a Null Pointer exception.

Considering the "convention over configuration" model of the JavaScript controllers, it would be helpful if we created a little boilerplate code when a new controller is created.

The controller name has to be the same name as the file itself so by creating this bit a boilerplate code, we make that a bit more intuitive for the user. Otherwise, we present them with a blank file and it's not completely obvious.

When a file has a long name it wraps in the IDE file explorer making it look like shit. It should be wider or possible make it adjustable (hiding any overflow)

Mapping each application to its own index can consume a lot of resources. Applications, and their associated resources, should be stored inside a single index.

This should be trivial to do by joining the app name with the content-type and storing that as a single type within elasticsearch.

Example.

// external url
/test/css/main.css

// maps internally to (where 'app' is a common index for applications)
/app/test_css/main.css

When running in unicast mode, the master node attempts to start but doesnt have a node to connect to so it get's stuck in a weird loop. Investigate how to start hazelcast correctly in unicast mode.

The following should trigger the bug:

./bin/cloud9 -f -Dc9.unicast.hosts=127.0.0.1

Investigate using hazelcast spring managed beans in hazelcast service and resources.

In order to prevent the overwriting of code files, the IDE should check the version of a document before saving it. If the version being submitted is older than the version in the index, the save should not succeed.

When you add an image to an application in the Web IDE, it's opened in a new tab but not displayed. This should be using real-time GETs but for some reason, it's not working.

Sessions should not expire.

The IDE uses dynamic loading of resources in the file explorer. If you try to create a new resource from the file menu and the root application has not been expanded (i.e., clicked) then the resource is not opened in the new tab view.

The resources does get created on the server but when the new tab calls fresh(), it fails because the node hasn't yet been created. The workaround is to simply expand the root node (i.e., app) in the file explorer which loads all the nodes.

It would be good to minify js/css files before caching them. Looks like we can use yui-compressor which is written in Java and leverages Rhino.

YUICompressor
Blog Post
Example

Add more tests. Integration, unit, and performance.