Thank you for your interest in contributing to Azure SDK for Java.
-
For reporting bugs, requesting features, or asking for support, please file an issue in the issues section of the project.
-
If you would like to become an active contributor to this project please follow the instructions provided in Microsoft Azure Projects Contribution Guidelines.
-
To make code changes, or contribute something new, please follow the GitHub Forks / Pull requests model: Fork the repo, make the change and propose it back by submitting a pull request.
-
Refer to the wiki to learn about how Azure SDK for java generates CheckStyle, SpotBugs, Jacoco, and JavaDoc reports.
-
There are two Maven projects in the repo. Refer to the wiki to learn about project structure for each.
- DO submit all code changes via pull requests (PRs) rather than through a direct commit. PRs will be reviewed and potentially merged by the repo maintainers after a peer review that includes at least one maintainer.
- DO NOT submit "work in progress" PRs. A PR should only be submitted when it is considered ready for review and subsequent merging by the contributor.
- DO give PRs short-but-descriptive names (e.g. "Improve code coverage for Azure.Core by 10%", not "Fix #1234")
- DO refer to any relevant issues, and include keywords that automatically close issues when the PR is merged.
- DO tag any users that should know about and/or review the change.
- DO ensure each commit successfully builds. The entire PR must pass all tests in the Continuous Integration (CI) system before it'll be merged.
- DO address PR feedback in an additional commit(s) rather than amending the existing commits, and only rebase/squash them when necessary. This makes it easier for reviewers to track changes.
- DO assume that "Squash and Merge" will be used to merge your commit unless you request otherwise in the PR.
- DO NOT fix merge conflicts using a merge commit. Prefer
git rebase
. - DO NOT mix independent, unrelated changes in one PR. Separate real product/test code changes from larger code formatting/dead code removal changes. Separate unrelated fixes into separate PRs, especially if they are in different assemblies.
-
DO use "Squash and Merge" by default for individual contributions unless requested by the PR author. Do so, even if the PR contains only one commit. It creates a simpler history than "Create a Merge Commit". Reasons that PR authors may request "Merge and Commit" may include (but are not limited to):
- The change is easier to understand as a series of focused commits. Each commit in the series must be buildable so as not to break
git bisect
. - Contributor is using an e-mail address other than the primary GitHub address and wants that preserved in the history. Contributor must be willing to squash the commits manually before acceptance.
- The change is easier to understand as a series of focused commits. Each commit in the series must be buildable so as not to break
- Install Java Development Kit 8 or 11
- add
JAVA_HOME
to environment variables
- add
- Install Maven
- add
MAVEN_HOME
to environment variables
- add
Note: If you ran into "long path" issue on
Windows
, enable paths longer than 260 characters by:
1.- Run this as Administrator on a command prompt:
REG ADD HKLM\SYSTEM\CurrentControlSet\Control\FileSystem /v LongPathsEnabled /t REG_DWORD /d 1
(might need to typeyes
to override key if it already exists)
2.- Set upgit
by running:
git config --system core.longpaths true
Refer to the build wiki for learning how to build Java SDKs and the unit testing wiki for guidelines on unit testing.
Live tests assume a live resource has been created and appropriate environment
variables have been set for the test process. To automate setting up live
resources we use created a script called New-TestResources.ps1
that deploys
resources for a given service.
To see what resources will be deployed for a live service, check the
test-resources.json
ARM template files in the service you wish to deploy for
testing, for example sdk\keyvault\test-resources.json
.
To deploy live resources for testing use the steps documented in Example 1 of New-TestResources.ps1
to set up a service principal and deploy live testing resources.
The script will provide instructions for setting environment variables before running live tests.
To run live tests against a service after deploying live resources:
mvn -f sdk/keyvault/pom.xml -Dmaven.wagon.http.pool=false --batch-mode --fail-at-end --settings eng/settings.xml test
Some live tests may have additional steps for setting up live testing resources. See the CONTRIBUTING.md file for the service you wish to test for additional information or instructions.
When building locally you might run into a Checkstyle such as the following:
Execution default of goal org.apache.maven.plugins:maven-checkstyle-plugin:3.1.0:check failed:
Plugin org.apache.maven.plugins:maven-checkstyle-plugin:3.1.0 or one of its dependencies could not be resolved:
Could not find artifact com.azure:sdk-build-tools:jar:1.0.0 in ossrh (https://oss.sonatype.org/content/repositories/snapshots/)
This is because the sdk-build-tools
project isn't released to Maven. To resolve this issue you'll need to copy the eng
folder locally then install sdk-build-tools
.
mvn clean install -f eng/code-quality-reports/pom.xml
All code in the Azure SDKs for Java repository must pass Checkstyle before being merged. The sdk-build-tools
is updated periodically, so if a new branch fails Checkstyle you'll need to reinstall.
Tooling has been introduced to centralize versioning and help ease the pain of updating artifact versions in POM and README files. Under the eng\versioning directory there exists version text files, one for client (version_client.txt) and one for data (version_data.txt). The format of the version files is as follows:
groupId:artifactId;dependency-version;current-version
The dependency-version should be set to the most recent released version and the current-version is set to the next version to be released. For example:
com.azure:azure-identity;1.0.0-beta.4;1.0.0-beta.5
Note: In the case of a new artifact both versions will be the same. In the case of a released artifact, the dependecny version should be the latest released version.
We no longer publish directly to Maven Central, any publishing is done through ESRP. Libraries published need to conform to one of the groupIds that we can publish to. Here is the list of accepted groupdIs:
- "com.microsoft"
- "com.azure.*"
- "com.azure"
- "com.microsoft.azure"
- "com.microsoft.azure.cognitiveservices"
- "com.microsoft.azure.functions"
- "com.microsoft.azure.kusto"
- "com.microsoft.azure.sdk.iot"
- "com.microsoft.azure.sdk.iot.provisioning"
- "com.microsoft.commondatamodel"
- "com.windowsazure"
- "com.microsoft.sqlserver"
- "com.microsoft.spring"
- "com.microsoft.servicefabric"
- "com.microsoft.rest"
- "io.clientcore"
Note: This list will be replaced with a link to a page created by ESRP providing a UI where GroupIds can be cheked and new ones requested if needed.
Libraries refer to things that are built and released as part of the Azure SDK. Libraries have a current version and a dependency version.
External Dependencies refer to dependencies for things that are not built and released as part of the Azure SDK regardless of the source. External Dependencies will only ever have a dependency version.
Current version, Dependency version, Unreleased Dependency version and Released Beta Dependency version
Current version - This is the version we should be using when defining a component in its POM file and also when dependent components are built within the same pipeline. The current version is the version currently in development.
Dependency version - This is the version we should be using when a given library is a dependency outside of a particular area. This should be the latest released version of the package whenever possible.
Unreleased Dependency version – Whenever possible, libraries should be using the latest released version for dependencies but there is the case where active development in one library is going to be needed by another library or libraries that are built in separate pipelines. These types of changes are specifically additive and not breaking. Once a library has GA’d, nothing short of breaking changes should ever force the dependency versions across the repo to an unreleased version. The reason for this is that it would prevent other libraries, that don’t need this change, from releasing. Unreleased dependcies of scope test will not prevent a library from being released.
Released Beta Dependency version – This is for when a library, which has already GA'd, is being released as a Beta version and we need to keep the dependency version to the latest GA. This particular tag will be used to allow other libraries to depend on the released Beta version. Libraries with released Beta dependencies can only be released as Beta, themselves, as a library cannot GA with Beta dependencies. An exception to the previous rule would be if the Beta dependency has a scope of test as this will not prevent a library from being released as GA.
An example of Current vs Dependency versions: com.azure:azure-storage-blob-batch
has dependencies on com.azure:azure-core
, com.azure:azure-core-http-netty
and com.azure:azure-storage-blob
. Because com.azure:azure-core
and com.azure:azure-core-http-netty
are both built outside of azure-storage pipeline we should be using the released or Dependency versions of these when they're dependencies of another library. Similarly, libraries built as part of the same pipeline, that have interdependencies, should be using the Current version. Since com.azure:azure-storage-blob-batch
and com.azure:azure-storage-blob
are both built part of the azure-batch pipeline when com.azure:azure-storage-blob
is declared as a dependency of com.azure:azure-storage-blob-batch
it should be the Current version.
An example of an Unreleased Dependency version: Additive, not breaking, API changes have been made to com.azure:azure-core
. com.azure:azure-storage-blob
has a dependency on com.azure:azure-core
and requires the additive API change that has not yet been released. An unreleased entry needs to be created in version_client.txt, under the unreleased section, with the following format: unreleased_<groupId>:<artifactId>;dependency-version
, in this example that would be unreleased_com.azure:azure-core;1.2.0
(this should match the 'current' version of core). The dependency update tags in the pom files that required this dependency would now reference {x-version-update;unreleased_com.azure:azure-core;dependency}
. Once the updated library has been released the unreleased dependency version should be removed and the POM file update tags should be referencing the released version.
All of the tooling lives under the eng\versioning directory.
- version_client.txt - Contains the Client library and versions
- version_data.txt - Contains Data library and versions
- external_dependencies.txt - Contains the external dependency versions
- update_versions.py - This is just a basic python script that will climb through the source tree and update POM and README files. The script utilizes tags within the files to do replacements and the tags are slightly different between the POM and README files.
- set_versions.py - This script should only be used by the build system when we start producing nightly ops builds.
In POM files this is done by inserting a specifically formatted comment on the same line as the version element.
<groupId>MyGroup</groupId>
<artifactId>MyArtifact</artifactId>
<version>1.0.0-beta.1</version> <!-- {x-version-update;MyGroup:MyArtifact;[current|dependency]} -->
The last element of the tag would be current or dependency depending on the criteria previously explained.
In README files this ends up being slightly different. Because the version tag is inside an XML element that we're explicitly telling a user to copy/paste into their product the comment tag really didn't make sense here. Instead there are tags before and after the XML element tags which effectively says "there's a version somewhere in between these two tags, when you find the line that matches replace it with the appropriate version of the group:artifact defined in the tag."
[//]: # ({x-version-update-start;MyGroup:MyArtifact;dependency})
```xml
<groupId>MyGroup</groupId>
<artifactId>MyArtifact</artifactId>
<version>1.0.0-beta.1</version>
```
[//]: # ({x-version-update-end})
Let's say we've GA'd and I need to tick up the version of azure-storage libraries how would I do it? Guidelines for incrementing versions after release can be found here.
- I'd open up eng\versioning\version_client.txt and update the current-versions of the libraries that are built and released as part of the azure storage pipeline. This list can be found in pom.service.xml under the sdk/storage directory. It's worth noting that any module entry starting with "../" are external module dependencies and not something that's released as part of the pipeline. Dependencies for library components outside of a given area would be downloading the appropriate dependency from Maven like we do for external dependencies.
- Execute the update_versions python script from the root of the enlistment. The exact syntax and commands will vary based upon what is being changed and some examples can be found in the use cases in the update_versions.py file.
- Review and submit a PR with the modified files.
- Management plane. Management is in the process of being moved to service pipeline builds. The versioning work needs to wait until that work is finished.
Making changes to an already GA'd library that require other libraries to depend on the unreleased version
This is where the unreleased_
dependency tags come into play. Using the Unreleased Dependency example above, where com.azure:azure-storage-blob
has a dependency on an unreleased com.azure:azure-core
:
- Make the additive changes to
com.azure:azure-core
- In version_client.txt add the entry for the unreleased azure core in the unreleased section at the bottom of the file. The entry would look like
unreleased_com.azure:azure-core;<version>
. Note: The version of the library referenced in the unreleased version tag should match the current version of that library. - In the pom.xml file for
com.azure:azure-storage-blob
, the dependency tag forcom.azure:azure-core
which was originally{x-version-update;com.azure:azure-core-test;dependency}
would now become{x-version-update;unreleased_com.azure:azure-core-test;dependency}
After the unreleased version ofcom.azure:azure-core
was released but beforecom.azure:azure-storage-blob
has been released. - In version_client.txt the the dependency version of
com.azure:azure-core
would become the released version and the "unreleased_" entry, at this time, would be removed. - In the pom.xml file for
com.azure:azure-storage-blob
, the dependency tag forcom.azure:azure-core
would get changed back to{x-version-update;com.azure:azure-core-test;dependency}
For general packaging versioning policies see Package Versioning and see Java for specific rules used in this repo.
While some Java projects use SNAPSHOT versions for nightly builds, we have opted not to use that convention because it has proven to be very unreliable in our scenarios. For example, if we use SNAPSHOT versions in our pom.xml files that usually ends up becoming viral throughout our entire repo and we want to more tightly control our versioning, especially our dependency versioning. On top of the viral nature, we have experienced a lot of network reliability issues when consuming SNAPSHOT versions from Maven central so we want to avoid this reliability issue in our build pipelines.
Given we don't use SNAPSHOT versions in our pom.xml files we generally have the version currently under development committed to the repo in the pom.xml file. This means if you are looking at our active development code for the version you will see a version that is not yet published to Maven central. If you want to try out our packages under development, you should look for our latest alpha build (see Dev Feed section below).
Every day our engineering system produces a set of packages for each component of the SDK. These can be used by other projects to test updated builds of our libraries prior to their release. The packages are published to an Azure Artifacts public feed hosted at the following URL:
For developers working within the repo, refer to the instructions above for updating versions numbers correctly. The parent POM for the Azure SDK already contains a repository reference to the daily feed and can download the packages.
For developers wishing to use the daily packages for other purposes, refer to the connect to feed instructions in Azure Artifacts.
Note: the daily package feed is considered volatile and taking dependencies on a daily package should be considered a temporary arrangement. We reserve the right to remove packages from this feed at any point in time.
Third party libraries should only be included in samples when necessary to demonstrate usage of an Azure SDK package; they should not be suggested or endorsed as alternatives to the Azure SDK.
When code samples take dependencies, readers should be able to use the material without significant license burden or research on terms. This goal requires restricting dependencies to certain types of open source or commercial licenses.
Samples may take the following categories of dependencies:
-
Open-source : Open source offerings that use an Open Source Initiative (OSI) approved license. Any component whose license isn't OSI-approved is considered a commercial offering. Prefer OSS projects that are members of any of the OSS foundations that Microsoft is part of. Prefer permissive licenses for libraries, like MIT and Apache 2. Copy-left licenses like GPL are acceptable for tools, and OSs. Kubernetes, Linux, and Newtonsoft.Json are examples of this license type. Links to open source components should be to where the source is hosted, including any applicable license, such as a GitHub repository (or similar).
-
Commercial: Commercial offerings that enable readers to learn from our content without unnecessary extra costs. Typically, the offering has some form of a community edition, or a free trial sufficient for its use in content. A commercial license may be a form of dual-license, or tiered license. Links to commercial components should be to the commercial site for the software, even if the source software is hosted publicly on GitHub (or similar).
-
Dual licensed: Commercial offerings that enable readers to choose either license based on their needs. For example, if the offering has an OSS and commercial license, readers can choose between them. MySql is an example of this license type.
-
Tiered licensed: Offerings that enable readers to use the license tier that corresponds to their characteristics. For example, tiers may be available for students, hobbyists, or companies with defined revenue thresholds. For offerings with tiered licenses, strive to limit our use in tutorials to the features available in the lowest tier. This policy enables the widest audience for the article. Docker, IdentityServer, ImageSharp, and Visual Studio are examples of this license type.
In general, we prefer taking dependencies on licensed components in the order of the listed categories. In cases where the category may not be well known, we'll document the category so that readers understand the choice that they're making by using that dependency.