Phase 4: Statement/Session separation (future consideration)

[apstndb]

Summary

Investigate whether separating statements and session into sub-packages is worthwhile.

Challenge

• Execute methods take *Session directly
• Session has ~30+ methods, making a SessionExecutor interface unwieldy
• Interface bloat risk may outweigh the organizational benefit

Possible approaches

1. Interface approach: Define narrow SessionExecutor interface — risk of interface being too large
2. File naming convention: Keep in same package, use consistent file prefixes for logical grouping
3. Partial extraction: Extract only statement types/definitions, keep execution in main package

Decision

This phase may not be necessary if file naming conventions provide sufficient organization. Evaluate after Phase 2-3 are complete.

Evaluation (post Phase 2-3)

Phase 2-3 completed:

• internal/mycli/ reduced from ~97 to 80 files
• 5 sub-packages extracted: decoder/, filesafety/, format/, metrics/, streamio/
• sysvar/ extraction abandoned — unexported struct fields in CustomVar/EnumVar used in ~20 struct literals in var_registry.go make extraction impractical without exporting fields or adding constructors

Assessment

File naming conventions provide sufficient organization for now. The ~30+ methods on Session make interface extraction high-effort/low-value. Keeping this open as a future consideration if the package grows further or natural interface boundaries emerge.

Remaining extraction candidates (not blocking)

• sysvar/ — blocked by unexported field access pattern; could revisit if var_registry.go is refactored
• Result struct decomposition — not urgent, could be done incrementally via embedded structs

Prerequisites

• Phase 3: Extract formatters via interface boundary #498 (Phase 3)

Parent: #495