Register Sas Converted Units

Register Converted SAS SQL into the Code Unit Registry

© Snowflake Inc. This skill and its contents are the proprietary intellectual property of Snowflake Inc.

Purpose: The second of two registration steps. After register-sas-source-units has written source-side CUR entries, this attaches each converted .sql file (files.converted), confirms the objectType from the generated SQL, and stamps codeStatus.conversion — so scai test seed/validate can run against the SAS conversion in Snowflake self-consistency mode.

Boundary (do not cross): Writes/updates registry/<id>.json only. No SnowConvert dialect, no .NET CodeUnitRegistry engine, no MCP state machines, no AIM claims/waves. See ../INTEGRATION.md; contract in ../references/cur-schema.md.

When to use

• After a SAS conversion has produced .sql output and its conversion_state.json marks files complete.
• Invoked automatically by convert-sas-to-snowflake post-conversion, or run standalone.

Inputs

1. Project root — the SAS conversion <output_dir> (already registered by register-sas-source-units).
2. conversion_state.json (optional path) — defaults to <project-root>/conversion_state.json. Supplies, per file: output_file (the converted .sql), tier, and dependencies.

Workflow

1. Ensure source units exist. If registry/ is empty, run register-sas-source-units first (this pass updates existing entries; it does not create them).

2. Run the emitter:

   cd ../assess-sas-migration/tool
   python3 emit_cur.py converted \
     --project-root <output_dir> \
     --state <path/to/conversion_state.json>
   

   For each file with status == "complete" (and not a 3-PYSPARK tier), it:

   • locates output_file, copies it under snowflake/, md5-checksums it, and sets files.converted.path (root-relative);
   • reads the converted SQL and sets the authoritative objectType from CREATE [OR REPLACE] PROCEDURE|FUNCTION|TABLE|VIEW (falling back to the tier mapping if the pattern is absent);
   • fills target.database/target.schema from metadata.target_schema;
   • stamps codeStatus.conversion.status = "completed".

3. Report attached counts and which units are now procedure/function (seedable) vs table/view (registered but not proc-tested).

4. Run the harness (self-consistency): see ../convert-sas-to-snowflake/workflows/scai-test-selfconsistency.md. In short: scai test seed --where "source.platform = 'sas'" then scai test validate against Snowflake. Note: scai test capture from a source system is not applicable to SAS — the baseline is the converted code's own stable Snowflake run (or the convert skill's validation expected-outputs).

Notes

• Idempotent. Re-running re-attaches to the same UUIDv5 units.
• Units whose source was never registered are skipped without error (run Skill A first). PySpark (Tier 3) units remain source-only — they are not SQL objects.