You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Upstream sync: merge dataform 3.0.62 into sqlanvil main #52
Upstream dataform-co/dataform released 3.0.62 (2026-07-28); local upstream/main already has it fetched (f3335b2). Occasional-sync policy per CLAUDE.md: work happens on main, upstream merges are deliberate syncs.
This sync is also the checklists.io #7 dogfood candidate: encode the merge procedure as a runbook (fetch/diff review → merge → conflict resolution → bazel test //core/... //cli/... → integration specs → version/pins decision → release runbook hand-off), execute it via the MCP server with human gates on the merge push and any release, evidence per step.
DO NOT START until the currently-running task completes (owner's note, 2026-07-29).
Related: our upstream PR dataform-co/dataform#2234 (--actions wildcards) is not yet merged upstream — if it lands before this sync, the sync inherits it and our local wildcard implementation needs a dedupe check.
Done. Merged as 0647bed7 (upstream 3.0.62 + insert_overwrite per adapter), goldens regenerated in sqlanvil's dialect and proved against live BigQuery (729fe950, 40a94a09), shipped in 1.29.0. The 3.0.63 follow-up (prune targets filter, f60a7b10) shipped in 1.29.1.
Two notes for whoever picks up the next sync:
This was executed directly, not as a checklists.io runbook with MCP gates. That dogfood is still tracked in ChecklistsIO/checklists-io#7.
Support '*' wildcards in --actions selectors as documented (#2224) dataform-co/dataform#2234 (--actions wildcards) has not merged upstream, so nothing was inherited here. Since then the fork has taken the same globToRegExp simplification the upstream review asked for (d4b44493), so when #2234 eventually lands, expect that function to be near-identical on both sides — dedupe rather than merge it.
Activity
ihistand commentedon Aug 11, 2026
Done. Merged as
0647bed7(upstream 3.0.62 +insert_overwriteper adapter), goldens regenerated in sqlanvil's dialect and proved against live BigQuery (729fe950,40a94a09), shipped in 1.29.0. The 3.0.63 follow-up (prunetargets filter,f60a7b10) shipped in 1.29.1.Two notes for whoever picks up the next sync:
--actionswildcards) has not merged upstream, so nothing was inherited here. Since then the fork has taken the sameglobToRegExpsimplification the upstream review asked for (d4b44493), so when #2234 eventually lands, expect that function to be near-identical on both sides — dedupe rather than merge it.