orm: add bulk insert/update support with CASE WHEN batch updates, MySQL time conversion, and checker/codegen validation

[Jengro777]

fix #26993

Summary

Add bulk (batch) INSERT and UPDATE support to the V ORM, enabling efficient multi-row operations through a single SQL statement instead of individual per-row round trips.

Both the compiler-level sql { } syntax and the function-call API (QueryBuilder) are supported.

Changes

vlib/orm/orm.v — Core batch SQL generation

• QueryData: add batch_rows and batch_key fields for batch operations
• orm_stmt_gen(.insert): generate multi-value VALUES (?, ?), (?, ?) tuples when batch_rows > 0
• orm_stmt_gen(.update): generate CASE "key" WHEN ? THEN ? ... ELSE "col" END expressions for batch updates when batch_rows > 0
• prepare_insert_query_data: handle batch data layout (interleaved row-by-row), filter types/kinds/auto_fields arrays correctly, and auto-skip serial fields only when all batch rows have zero values
• clone_query_data: propagate parentheses, batch_rows, batch_key

vlib/orm/orm_func.v — Function-call API

• insert_many (QueryBuilder): changed from N individual conn.insert() calls to a single batch INSERT with interleaved data and batch_rows set
• update_many (new standalone function): batch UPDATE using CASE WHEN with a key field, accepting field_names to select which columns to update and key_field as the matching column; WHERE clause uses IN

vlib/orm/orm_fn_test.v — SQL generation unit tests

• test_orm_stmt_gen_bulk_insert: verify multi-value tuple generation for default, pg, and mysql dialects
• test_orm_stmt_gen_bulk_update: verify CASE WHEN batch UPDATE generation for all three dialects
• test_orm_stmt_gen_insert_default_values_mysql: verify batch () VALUES (), () for MySQL with auto-fields

vlib/orm/orm_func_test.v — Integration tests

• test_orm_func_update_many: end-to-end SQLite test for update_many with 2-row batch, single-row batch, and empty-values error

vlib/v/ast/ast.v — AST node fields

• SqlStmtLine: add is_array_insert, is_array_update, array_update_var, array_update_key

vlib/v/checker/orm.v — Compile-time validation

• Detect array-typed objects in sql { insert array into T } and sql { update T set ... = array.field where key == array.key }
• Validate element types match the table, reject pointer element types, reject non-primitive/non-enum/non-time fields in bulk mode
• Reject array upsert (not yet supported)

vlib/v/gen/c/orm.v — C codegen for bulk operations

• write_orm_bulk_insert: for non-serial fields, build a single QueryData with batch_rows and call conn.insert() once; for tables with serial fields, fall back to per-row inserts so each row gets its own auto-generated ID
• write_orm_bulk_update: build per-column, per-row key+value data for CASE WHEN, with OR-connected WHERE conditions

vlib/db/mysql/orm.c.v — MySQL backend fix

• convert_query_data_to_primitives: use i % data.fields.len to correctly map batch data elements back to field names for time.Time conversion
• Construct QueryData with all fields propagated (including batch_rows, batch_key, parentheses, auto_fields)

vlib/v/tests/orm_bulk_insert_update_test.v — End-to-end compiler tests

• test_orm_bulk_insert_and_update: insert 2 rows, verify, update both with CASE WHEN, verify
• test_orm_bulk_insert_preserves_all_default_rows: insert 3 default rows, verify all get distinct IDs
• test_orm_bulk_insert_with_mixed_serial_values_keeps_defaults: mix zero and explicit ID values in batch
• test_orm_bulk_update_with_renamed_column: batch update with @[sql: 'display_name_text'] renamed column

vlib/v/checker/tests/orm_bulk_pointer_array_error.vv — Checker error test

• Verify proper error messages when using pointer arrays ([&T]) for bulk insert/update

Testing

• All 37 ORM unit/integration tests pass
• The new orm_bulk_insert_update_test.v (4 test functions) passes
• vlib/v/compiler_errors_test.v: 1565 passed, 5 skipped
• vlib/v/gen/c/coutput_test.v: 95/95 .out files match, 117/117 .c.must_have patterns match
• SQLite (in-memory) integration test for update_many API

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: d7ba71c209

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[chatgpt-codex-connector]

Preserve per-row default handling in insert_many

insert_many now unconditionally batches by setting batch_rows, which changes behavior for tables with serial/default columns: rows that used to omit the auto field on a per-row basis are now forced through one batched shape, so mixed explicit/zero IDs no longer preserve defaults and can write literal zero/default-breaking values. This regresses semantics from the previous per-row loop and should fall back to row-by-row inserts when auto/default fields are present (as the compiler SQL bulk path already does).

Useful? React with 👍 / 👎.

[chatgpt-codex-connector]

Generate all rows for empty-field batch inserts

In the batch insert path, when all insert fields are skipped and batch_rows > 1, sqlite/pg/h2 still emit a single DEFAULT VALUES statement, so only one row is inserted instead of the requested batch size. Any caller that batches records containing only default/serial columns will silently lose rows; the SQL generation needs to account for row_count in this branch (or execute per-row statements).

Useful? React with 👍 / 👎.

[Jengro777]

Failing doesn't seem to have anything to do with PR