OpenSearch/GRADLE.CHEATSHEET.asciidoc

62 lines
3.1 KiB
Plaintext

To compile `x-plugins`, you must clone the Elasticsearch repository into the same parent directory. For example:
[source,bash]
----
$ mkdir elastic
$ cd elastic
$ git clone git@github.com:elastic/elasticsearch.git
$ git clone git@github.com:elastic/x-plugins.git
$ git clone git@github.com:elastic/kibana.git <1>
----
<1> For anyone doing UI development, it's also useful to have Kibana at the same level.
Once cloned, any command should be executed from the **elasticsearch** directory. This ensures that the full dependency tree is available.
[source,bash]
----
$ cd elasticsearch
$ gradle clean test check <1>
----
<1> This will run the `clean` task, `test` task, and then the `check` task on _every_ project that has it. However, `check` requires that `test` be run, so it won't _rerun_ `test`. `clean` is unnecessary here, but people often use it anyway.
If this command were run in a different order, then it would still follow the same rules, but the behavior would change:
[source,bash]
----
$ gradle check test clean <1>
----
<1> It would run every task that `check` requires (e.g., `test` and `integTest`), skip `test` because it has already been run (indirectly by `check`), and then finally it would _wastefully_ delete every project output.
As a quick helper, below are the equivalent commands from `maven` to `gradle`. You can also run `gradle tasks` to see all tasks that are available to run.
[cols="3*", options="header"]
|====
| Maven | Gradle | Description
| `clean` | `clean` | Delete anything that exists already. You do _not_ generally need to run `clean` with Gradle for any task that _Gradle_ manages the inputs/outputs (in other words, it knows when it needs to rebuild versus reuse).
| `test` | `test` | Run all unit tests.
| `verify` | `check` | Run all tests, plus extra checks (e.g., `checkStyle`, `forbiddenApis`, etc.).
| `verify -Dskip.unit.tests` | `integTest` | Run only integration tests.
| `package -DskipTests` | `assemble` | Output is in `${project.projectDir}/build/distributions`
| `install -DskipTests` | `install` | Build jars and place them into the local _Maven_ repository (yes, even with Gradle).
This should be unnecessary with the unified build!
|====
The full task list, with a minor breakout as a graph of dependencies can be seen with:
[source,bash]
----
$ gradle tasks --all
----
Given that we currently have 80 projects, this can be extremely verbose.
With Gradle, you can easily target specific `projects` to run commands against, and it will build all necessary dependencies to make it happen. For example, if you make a change to a specific test in the `x-pack` subproject, then you can specifically invoke its `test` task.
[source,bash]
----
$ gradle :x-plugins:elasticsearch:x-pack:test -Dtests.class=*YourTests
----
This applies to any command that follows the Directed Acyclic Graph (DAG) for its dependencies. The above example would trigger Elasticsearch `core` to be built, as well as the test framework and any other dependencies that it may have.