* feat(import): widen issueops batch-create surface from *sql.Tx to DBTX (bd-o58fz)
Mechanical signature widening: CreateIssuesInTxWithResult and its transitive
closure (batch context, ID generation, orphan check, conditional upsert,
labels/comments/dependencies persistence, child counters, blocked recompute)
now take the existing DBTX interface instead of *sql.Tx. *sql.Tx satisfies
DBTX, so every classic caller compiles unchanged and no SQL changes; the
domain/db Runner (proxied unit-of-work) satisfies the same interface, which
lets the proxied import port call the SAME batch-upsert code instead of
mirroring its semantics in parallel SQL.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NvU6bUda8oQjvG3zoa4uyX
* feat(import): add the Importer capability — one-transaction bulk import behind the uow provider (bd-o58fz)
The public issueops.Importer role lands one whole import batch in ONE unit
of work with ONE history entry: issue rows through the same batch-upsert
engine the classic stores run (issueops.CreateIssuesInTxWithResult over the
DBTX seam widened in the previous commit), memory records, and the optional
issue_prefix reconciliation. Engine callbacks are folded into the result
inside the attempt so RunTxResult's serialization retries cannot double-count.
Accessor follows the BatchCloserSource capability pattern on doltSQLProvider.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_019aBRPPEe7c9LiiM6NABJJg
* feat(import): port bd import to proxied-server mode through the Importer capability (bd-o58fz)
Removes the proxied refusal. The parse loop, dedup filter, dry-run
classification, stale pre-filter, result assembly and reporting are now
shared seam-based code (importIssueLookup / importTitleSearcher), so both
modes classify and report through the same functions; the proxied write is
ONE ImportBatch call — rows, memories and the issue_prefix reconciliation in
one transaction with one history entry, per the one-commit-per-invocation
doctrine. The stale guard keeps its classic two-half shape (pre-filter read
+ RejectStaleUpserts inside the write transaction).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_019aBRPPEe7c9LiiM6NABJJg
* test(import): pin the Importer role contract on the uow backend (bd-o58fz)
One batch = one history entry with direct content assertions (labels,
comments, dependency edges, memory value, commit message); idempotent
re-import; the in-transaction stale guard and --allow-stale restore; empty
batch commits nothing; prefix-sync-only commits under the sync message;
empty actor refused.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_019aBRPPEe7c9LiiM6NABJJg
* test(import): proxied-server import journey (bd-o58fz)
TestProxiedServerImport: round-trip with direct content assertions (labels,
comment text, blocks edge, memory value) and the one-commit-per-invocation
oracle; idempotent re-import commits nothing; stdin '-' semantics and the
bd-axluy redirect guard; stale skip / --allow-stale restore / newer-row
upsert with the updated-issues report; --dry-run writes nothing; --dedup;
import as the sanctioned upsert door for an occupied ID.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_019aBRPPEe7c9LiiM6NABJJg
* lint(import): annotate three gosec G705 stderr-progress lines (bd-o58fz)
Lion's #5355 gate review: import.go renderImportOutcome dedup/stale lines and
the import_shared.go chunk-progress line were missing the G705 nolint the
sibling renderImportDryRun lines already carry. Approve-on-green per #5346
convention.
* lint(import): annotate the fourth G705 progress line gosec surfaced on re-run (bd-o58fz)
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>