sql-machine-learning
The SQLFlow binary versions are inconsistent between Runnable image and the latest release of SQLFlow repo.
User usually use sqlflow/sqlflow:step as the base image to build the Runnable image. The base image contains the SQLFlow step binary and runtime library.
And then runnable authors publish the runnable image, the version of the SQLFlow binary and runtime library is v1.0 for example at that time. Users then use the Runnable in the SQLFlow TO RUN statement and deploy the SQL program as a daily production job. The job will always use the SQLFlow binary and runtime library of v1.0 in the runnable image.
At a later time, SQLFlow repo released v2.0. But the production job above always uses v1.0. It brings the version inconsistency and is hard to maintain.
Maybe the sqlflowserver version have very little chance to interference with the workflow step. The sqlflowserver is only generating the workflow and run, yet the version of TO RUN can manage the version by itself.
