update-pipeline

# Update Pipeline Converts tracked entity changes into database INSERT/UPDATE/DELETE commands during `SaveChanges()`. ## Flow `SaveChanges()` → `DetectChanges()` → `IDatabase.SaveChanges()` → `UpdateAdapter` creates `IUpdateEntry` list → `CommandBatchPreparer.BatchCommands()` → `ModificationCommand` per table row, composed of `ColumnModification` per column → `SharedTableEntryMap` is used to track entries mapped to the same row → Topological sort via Multigraph (FK dependency ordering) → Groups into `ModificationCommandBatch` (respects max batch size) → `UpdateSqlGenerator` generates SQL per batch → `BatchExecutor` executes all batches in a transaction → `StateManager.AcceptAllChanges()` ## Concurrency Concurrency tokens → WHERE conditions on UPDATE/DELETE. `AffectedCountModificationCommandBatch` checks affected rows. Throws `DbUpdateConcurrencyException` on mismatch. ## Shared-Row Value Access Column accessors may need values from a `SharedIdentityEntry` when one CLR entry replaces another mapped to the same row. Select the effective entry before resolving the column's property mapping: entries in a shared-identity pair can have different entity types, so a property from the original type must not be used against the replacement. Regression tests should exercise the resulting command dependency ordering, not just the accessor in isolation. ## Validation - `SaveChanges()` returns expected affected row count - Store-generated values propagate back to entities after INSERT/UPDATE - `DbUpdateConcurrencyException` thrown when expected for stale data