How to release

The guide to release PyIceberg.

The first step is to publish a release candidate (RC) and publish it to the public for testing and validation. Once the vote has passed on the RC, the RC turns into the new release.

Running a release candidate

Make sure that the version is correct in pyproject.toml and pyiceberg/__init__.py. Correct means that it reflects the version that you want to release.

Setting the tag

Make sure that you're on the right branch, and the latest branch:

For a Major/Minor release, make sure that you're on main, for patch versions the branch corresponding to the version that you want to patch, i.e. pyiceberg-0.6.x.

git checkout <branch>
git fetch --all
git reset --hard apache/<branch>

Set the tag on the last commit:

export RC=rc1
export VERSION=0.1.0${RC}
export VERSION_WITHOUT_RC=${VERSION/rc?/}
export VERSION_BRANCH=${VERSION_WITHOUT_RC//./-}
export GIT_TAG=pyiceberg-${VERSION}

git tag -s ${GIT_TAG} -m "PyIceberg ${VERSION}"
git push apache ${GIT_TAG}

export GIT_TAG_REF=$(git show-ref ${GIT_TAG})
export GIT_TAG_HASH=${GIT_TAG_REF:0:40}
export LAST_COMMIT_ID=$(git rev-list ${GIT_TAG} 2> /dev/null | head -n 1)

The -s option will sign the commit. If you don't have a key yet, you can find the instructions here. To install gpg on a M1 based Mac, a couple of additional steps are required: https://gist.github.com/phortuin/cf24b1cca3258720c71ad42977e1ba57

Upload to Apache SVN

Both the source distribution (sdist) and the binary distributions (wheels) need to be published for the RC. The wheels are convenient to avoid having people to install compilers locally. The downside is that each architecture requires its own wheel. use cibuildwheel runs in Github actions to create a wheel for each of the architectures.

Before committing the files to the Apache SVN artifact distribution SVN hashes need to be generated, and those need to be signed with gpg to make sure that they are authentic.

Go to Github Actions and run the Python release action. Set the version to main, since we cannot modify the source. Download the zip, and sign the files:

cd release-main/

for name in $(ls pyiceberg-*.whl pyiceberg-*.tar.gz)
do
    gpg --yes --armor --local-user fokko@apache.org --output "${name}.asc" --detach-sig "${name}"
    shasum -a 512 "${name}" > "${name}.sha512"
done

Now we can upload the files from the same directory:

export SVN_TMP_DIR=/tmp/iceberg-${VERSION_BRANCH}/
svn checkout https://dist.apache.org/repos/dist/dev/iceberg $SVN_TMP_DIR

export SVN_TMP_DIR_VERSIONED=${SVN_TMP_DIR}pyiceberg-$VERSION/
mkdir -p $SVN_TMP_DIR_VERSIONED
cp * $SVN_TMP_DIR_VERSIONED
svn add $SVN_TMP_DIR_VERSIONED
svn ci -m "PyIceberg ${VERSION}" ${SVN_TMP_DIR_VERSIONED}

Upload to PyPi

Go to Github Actions and run the Python release action. Set the version of the release candidate as the input: 0.1.0rc1. Download the zip and unzip it locally.

Next step is to upload them to pypi. Please keep in mind that this won't bump the version for everyone that hasn't pinned their version, since it is set to an RC pre-release and those are ignored.

twine upload -s release-0.1.0rc1/*

Final step is to generate the email to the dev mail list:

cat << EOF > release-announcement-email.txt
To: dev@iceberg.apache.org
Subject: [VOTE] Release Apache PyIceberg $VERSION
Hi Everyone,

I propose that we release the following RC as the official PyIceberg $VERSION_WITHOUT_RC release.

A summary of the high level features:

* <Add summary by hand>

The commit ID is $LAST_COMMIT_ID

* This corresponds to the tag: $GIT_TAG ($GIT_TAG_HASH)
* https://github.com/apache/iceberg-python/releases/tag/$GIT_TAG
* https://github.com/apache/iceberg-python/tree/$LAST_COMMIT_ID

The release tarball, signature, and checksums are here:

* https://dist.apache.org/repos/dist/dev/iceberg/pyiceberg-$VERSION/

You can find the KEYS file here:

* https://dist.apache.org/repos/dist/dev/iceberg/KEYS

Convenience binary artifacts are staged on pypi:

https://pypi.org/project/pyiceberg/$VERSION/

And can be installed using: pip3 install pyiceberg==$VERSION

Please download, verify, and test.

Please vote in the next 72 hours.
[ ] +1 Release this as PyIceberg $VERSION_WITHOUT_RC
[ ] +0
[ ] -1 Do not release this because...
EOF

cat release-announcement-email.txt

Vote has passed

Once the vote has been passed, you can close the vote thread by concluding it:

Thanks everyone for voting! The 72 hours have passed, and a minimum of 3 binding votes have been cast:

+1 Foo Bar (non-binding)
...
+1 Fokko Driesprong (binding)

The release candidate has been accepted as PyIceberg <VERSION>. Thanks everyone, when all artifacts are published the announcement will be sent out.

Kind regards,

Copy the artifacts to the release dist

svn checkout https://dist.apache.org/repos/dist/dev/iceberg /tmp/iceberg-dist-dev
svn checkout https://dist.apache.org/repos/dist/release/iceberg/ /tmp/iceberg-dist-release

mkdir -p /tmp/iceberg-dist-release/pyiceberg-<VERSION>
cp -r /tmp/iceberg-dist-dev/pyiceberg-<VERSION>rcN/* /tmp/iceberg-dist-release/pyiceberg-<VERSION>

svn add /tmp/iceberg-dist-release/
svn ci -m "PyIceberg <VERSION>" /tmp/iceberg-dist-release/

The latest version can be pushed to PyPi. Check out the Apache SVN and make sure to publish the right version with twine:

twine upload -s /tmp/iceberg-dist-release/pyiceberg-<VERSION>/*

Send out an announcement on the dev mail list:

To: dev@iceberg.apache.org
Subject: [ANNOUNCE] Apache PyIceberg release <VERSION>

I'm pleased to announce the release of Apache PyIceberg <VERSION>!

Apache Iceberg is an open table format for huge analytic datasets. Iceberg
delivers high query performance for tables with tens of petabytes of data,
along with atomic commits, concurrent writes, and SQL-compatible table
evolution.

This Python release can be downloaded from: https://pypi.org/project/pyiceberg/<VERSION>/

Thanks to everyone for contributing!

Release the docs

A committer triggers the Python Docs Github Actions through the UI by selecting the branch that just has been released. This will publish the new docs.