Skip to content

SQLFlow Tagging and Releasing

weiguoz edited this page Nov 18, 2019 · 2 revisions

SQLFlow is still under heavy development, the Extended SQL grammar and usage are evolving rapidly. We need a regular tagging and releasing mechanism to solve the below problems:

  1. It's hard to trace issues if the user doesn't know what version is he using. e.g. the error message in issue https://github.com/sql-machine-learning/sqlflow/issues/1166 is already removed in develop branch.
  2. Let users upgrade their deployment to a newer version to fix known issues and adding support of features. Directly upgrade to develop branch may be risky since the develop branch may contain new bugs.

Periodically Releasing

I propose to increase the SQLFlow version every one or two months before we release v1.0. The release process should be:

  1. Make sure the develop branch has passed the CI.
  2. Make sure the code coverage is acceptable.
  3. Create a new releasing branch named release-0.x.
  4. If there are bugs to fix for this release, merge the bug fix pull requests to develop and then cherry-pick the commit into the release branch.
  5. Once all bugs are fixed, push a tag named v0.x.x to the current release branch.
  6. If there are more bugs to fix, repeat steps 4 and 5.