Backend Data Interchange

AsaDB supports backend-owned import and export for MySQL SQL, PostgreSQL SQL, CSV, and XLSX. Portable interchange is separate from the authenticated `.asb` production-backup format:

use `.asb` for complete logical backup and verified disaster recovery;
use MySQL/PostgreSQL SQL, CSV, or XLSX to exchange selected tables and views with other tools.
The browser sends only format and selection controls. src/asadb_interchange.pl holds the normal backend execution lock for export and scans verified record pages directly, so a large export does not depend on rows currently loaded in AsAPanel.

Formats

Format	Export	Import behavior
MySQL SQL	Selected tables plus `CREATE VIEW` for selected views, indexes, and 256-row multi-value `INSERT` batches	Common dump controls are removed and portable DDL/DML enters the normal AsaDB parser and transaction path.
PostgreSQL SQL	PostgreSQL identifiers/types, selected view definitions, plus streaming `COPY ... FROM stdin` table data	`COPY` rows become bounded multi-value inserts; common schema qualifiers and portable type casts are normalized.
CSV	One selected relation produces `.csv`; multiple relations produce a ZIP containing one CSV per relation	Tables scan backend pages; views are evaluated by the normal backend executor. Plain CSV and AsaDB `-csv.zip` packages are accepted.
XLSX	A valid OOXML workbook with one sheet per selected relation	Tables scan backend pages; views are evaluated by the normal backend executor. ZIP paths/counts/sizes are validated; worksheets are parsed row by row; a 256-row sample infers column types without materializing the workbook.

CSV and XLSX imports support replace (DROP then CREATE) and append (`CREATE IF NOT EXISTS`, then insert). The target table field names the first CSV table or XLSX sheet; additional workbook sheets receive a sanitized sheet suffix.

Execution and safety

Uploaded interchange files enter the durable Reservoir spool with their format, source name, target, and write mode. The worker prepares bounded AsaDB SQL in Prolog, then uses the existing transactional streaming importer. Cancellation, rollback, capacity limits, persisted progress, idempotency, and single-writer execution therefore remain active.

XLSX input is rejected if it contains an unsafe member path, more than 4,096 ZIP entries, an entry larger than 256 MiB, or more than 512 MiB of declared uncompressed data. XML DTD and entity declarations are rejected before workbook parsing. These are defensive input limits, not recommended workbook sizes.

MySQL and PostgreSQL have vendor-specific extensions. The interchange layer covers the portable dump surface tested by AsaDB; unsupported procedural objects, extensions, custom types, and vendor-only clauses are reported by the normal parser rather than silently accepted.

HTTP contract

Authenticated portable export uses POST /api/export with:

database;
format=mysql|postgresql|csv|xlsx;
tables and data_tables as comma-separated backend table or view names;
include_schema, include_data, create_database, and drop_tables;
output=save|open. Raw imports use POST /api/reservoir/jobs. AsAPanel supplies X-AsaDB-Import-Format, X-AsaDB-Import-Name, X-AsaDB-Import-Table, and X-AsaDB-Import-Mode. Production `.asb` files continue to use POST /api/import_upload so backup integrity verification cannot be bypassed.

Validation

Run:

make test-interchange
make test-interchange-stress

The regression covers Unicode, NULL, quotes, multiline values, MySQL multi-row inserts, PostgreSQL COPY, CSV quoting/type inference, XLSX OOXML/ZIP validation, authenticated HTTP export, Reservoir import, and backend row verification. On the recorded PCLinuxOS 1.5.0 audit run, four exports plus CSV and XLSX preparation over 20,000 rows per format completed in 13.8 seconds. This is a development measurement, not a universal latency guarantee.