ingestr v1.0.49 - Schema Evolution Fixes

ingestr v1.0.49 was published on June 29, 2026, with the main operator change in schema evolution for destination writes. The release stops BigQuery merge reruns from issuing an ALTER against integer primary key columns when an override narrows type width, and it also tightens where numeric width suppression applies. It is not marked as a prerelease.

The full release notes and downloads are on the GitHub release page.

The practical fix is around integer primary keys in BigQuery merge mode. The release notes say schema evolution now skips an ALTER for a narrower integer override and does not alter integer primary key columns on BigQuery merge reruns. For data loads that use primary keys to converge changed rows, that removes a class of noisy or risky schema drift during repeated syncs.

There is also an explicit mapping from int and integer overrides to int64. That matters because BigQuery integer storage is already INT64, while source systems and config files often use narrower names. The release keeps that mapping in the schema evolution path instead of letting width language leak into destination DDL.

For scheduled ELT, the visible outcome should be boring: fewer schema changes during reruns and fewer differences between the first load and later merge runs. The release notes do not claim a new BigQuery feature. They describe a guard around an existing schema evolution path.

Several commits narrow the scope of numeric width suppression by destination scheme. The notes call out BigQuery, Snowflake, and Trino gates, plus float width override skips on collapsing destinations. That points to a defensive change: only the destinations that need this behavior should get it.

This is the part platform operators should read as boundary cleanup, not a feature expansion. Type width rules are destination semantics. Applying a BigQuery shaped suppression to Snowflake or Trino can hide a real mismatch, while applying it nowhere can create churn in generated DDL. The new gating is about keeping those rules local to the destination adapter.

Watch the destination matrix during rollout. If an existing pipeline depended on a narrower override being pushed into destination DDL, this release may reduce that DDL activity instead of creating a visible schema change. That is usually the safer behavior for append and merge jobs, but it is still worth comparing generated plans against one representative table per destination.

The dependency side is smaller but still relevant for connector fleets. ingestr bumps go.mongodb.org/mongo-driver to v1.17.7 and updates the license audit lock for the same MongoDB driver version. That is a runtime dependency change for anyone pulling MongoDB data through ingestr, and an audit artifact change for teams that gate builds on dependency license state.

The remaining note drops and shortens comments around DestinationScheme. That is internal cleanup. It does not describe a user facing flag or config migration, but it does make sense beside the destination gate fixes because that field is now central to deciding where schema evolution rules apply.

If your deploy process vendors Go modules or mirrors license metadata, treat this as both code and policy input. The driver bump and lock refresh should be reviewed together. A deploy that updates the binary without the audit lock can fail in a strict supply chain workflow even when the connector itself builds.