corda/docs/source/CLI-vs-IDE.rst
2016-11-29 16:52:49 +00:00

4.3 KiB

CLI vs IDE

We have tried to make every demo, example, tutorial and sample to be both usuable via the command line and also the IntelliJ IDE. Most developers will find writing, editing and debugging code more easily done via tools such as an IDE, but when code needs to be deployed to run as nodes, control must be done via the command line - no organisations allow their systems to be running via a developer environment.

IDE - IntelliJ

IntelliJ (the preferred IDE in R3) integrates well with gradle (our chosed build, deploy and CLI tool). IntelliJ understands gradle tasks and dependencies, automatically loading them in the background when a project is first loaded or the gradle project changes. Occasionally, however, you may need to refresh the gradle project manually - but this is hinted to you by the IDE. It's a good idea to do this before carrying on with other work (and in fact you may find it is essential to pick up new libraries etc).

There are some great resources about how to get started using IntelliJ. As opposed to trying to repeat them here, we advise you to go to the IntelliJ docs here.

Command Line

Windows vs Mac / Unix

Due to the nature of their respective command interfaces, gradle is typically ran in windows with the command gradle.bat (or gradlew.bat if using the wrapper) and in Mac / Unix environments it is ran via ./gradlew. For brevity, the simple windows syntax gradle is used for the majority of the documentation.

As well as including the most significant run and build configurations in the IDE, we also provide gradle tasks to build, install and run significant parts of Corda demos and tools. Gradle is highly extensible and we use it for downloading required resources, building components, installing those built components into shared areas, configuring the scripts that run nodes, starting up demonstration API calls amongst other things. It is exceptionally good at deriving dependency maps and therefore performing the preceeding tasks required in order to do the requested task. However, when confusing build errors manifest, then sometimes a gradle clean may be required in order to clear out any build areas that have an inconsistent state. The total build time from downloading / cloaning the repo to a complete build should be only a few minutes, obviously slightly longer if the unit tests are run.

Frequently Used Gradle Tasks

Note that the list of tasks can be ran for any gradle project can be displayed by running the task tasks. Also note that gradle is hierachical and therefore tasks in child directories can be run using a colon seperator - ie if you want to run the sample attachment-demo runSender you would use the command gradle samples:attachment-demo:runSender

The most frequent gradle tasks you will probably be running are build and install. The build command also executes the unit tests as well. If you want to build without this level of verification, then use the assemble command - but we do not recommend this. After running build, the install tasks copies over the built jars into the local maven repository which will then makes these available for either the sample code or use with the CorDapp template.

Debugging

Tasks and processes that are run directly via the IDE (including via the usage of the driver DSL) can be remotely debugged. We do not have java debugging currently enabled in the runnodes scripts generated by a process we refer to as 'cordformation' but we will be implementing that shortly.

Via the IDE

To debug: From the IDE, configure the debug connectivity option by the "Edit Configurations" and choosing "+" and then "Remote". The debug port start at 5005 and increments for each additional node that starts, the order given by the list in the main driver configuration (which is primarily listed in the main function of Main.kt for each sample. Look for the string Listening for transport dt_socket at address:5xxx in the log output to determine the exact port for that node. If the log messages are mixed from several nodes to the same console, then (as earlier stated), the port numbers increment in the order they are listed in the driver DSL configuration.