ABIS is a desktop ERP / manufacturing-execution system for an aluminum
coil → sheet processing plant, built in PowerBuilder (internal application
name lion; codebase nicknamed "silverdome"). It covers quotation, order entry,
coil receiving, inventory, shop-floor production jobs, QA/mechanical testing and
certification, skid/scrap handling, warehouse, shipping/EDI, and accounting,
plus direct integration with plant-floor hardware (scales/gauges over serial and
PLCs over OPC).
This repository is undergoing a modernization effort (greenfield, strangler-fig).
- Phase 1 — discovery & roadmap under
docs/: analysis, reproducible tooling, the full recovered data model, and the plan.- Phase 2 — the API seam under
api/: an ASP.NET Core 8 REST API over the database (Oracle in production; a seeded SQLite fixture for dev/CI), with a fully-typed OpenAPI contract + generated TypeScript client.- Phase 3/4 — greenfield modules (in progress, far along): every business library in the legacy target is now rebuilt as a typed web module on that API — order entry, jobs, coil inventory & ownership transfer, skids/warehouse, sales, quotation, QA & coil-evaluation, recovery, accounting, EDI, OPC log, downtime, maintenance, scan, dies, parts, receiving (with coil-minting), shifts, shipping, sketches, security/authorization, the DAS & stacker shop-floor consoles, the production folder, and a production/customer/inventory/QA reporting suite. The legacy PowerBuilder source for these areas is vendored under
legacy/src/so each was built against the real tables/columns. What remains is the cutover:docs/REMAINING_WORK.mdtracks work that needs no input, anddocs/OPEN_QUESTIONS.mdwhat is waiting on a decision.
| Document | What's in it |
|---|---|
docs/REMAINING_WORK.md |
Start here to continue — the backlog: what is built, what is left, and what was investigated and deliberately not built. |
docs/OPEN_QUESTIONS.md |
Everything blocked on a plant decision, each with what it blocks and what happens under each answer. |
CONTRIBUTING.md |
Dev setup, the recipe for adding a module, and solved gotchas. |
docs/ARCHITECTURE.md |
As-is architecture: stack, 2-tier topology, the two app targets (lion + the da data-acquisition app), module map, integration surface, the Aug-2025 PB migration, and build-readiness issues. |
docs/DATA_MODEL.md |
Full database model (412 tables) recovered from a live Oracle data-dictionary dump of the DBO schema — real columns, PKs, FKs, indexes, and sequences. Supersedes the earlier partial DataWindow-inferred model. |
docs/OBJECT_INVENTORY.md |
Approximate object inventory (~1,345 DataWindows, ~588 windows, …) recovered from the compiled libraries — names only; read legacy/src/ for behaviour — per-library and grouped by domain. |
docs/MODERNIZATION_ROADMAP.md |
Strategic options, the greenfield/strangler-fig approach, and the phased plan with status. |
docs/data-model/BACKCHECK.md |
Verification of each built module's schema against the real DataWindow/Oracle columns. |
docs/ORACLE_VALIDATION.md |
Oracle data-path validation runbook + the live-only bug classes found and fixed. |
docs/INTEGRATIONS.md · docs/EDGE_SERVICE.md |
External integrations (EDI/serial/OPC) and the shop-floor edge service. |
docs/INSTALL.md |
One-command native install on Ubuntu — systemd service + nginx + HTTPS (Let's Encrypt). The Docker-free path. |
docs/DEPLOY_DEB.md |
Cheat sheet for rebuilding and redeploying the .deb on the running server (.110). |
docs/DEPLOY.md |
Run the API + greenfield UIs on a server (Docker Compose). |
legacy/src/README.md |
The vendored legacy PB source per area + which greenfield module each maps to. |
The docs are backed by deterministic extractors (Python 3, standard library only — no dependencies). Re-run them any time the sources change:
python3 tools/ingest_oracle_schema.py # live Oracle dump -> docs/data-model/oracle_schema.json + docs/DATA_MODEL.md
python3 tools/extract_schema.py . # DataWindow cross-check -> docs/data-model/schema.json
python3 tools/extract_inventory.py . # -> docs/inventory/objects.jsontools/ingest_oracle_schema.pyparses a live Oracle data-dictionary dump (docs/data-model/oracle_schema.txt, produced bytools/oracle_introspect.sql) into the full schema — this is the authoritative source forDATA_MODEL.md.tools/extract_schema.pyparses the exported DataWindow/query sources (*.srd,*.srq) fortable.columnbindings, types, and joins (now a partial cross-check; superseded by the live model above).tools/extract_inventory.pyscans the compiled*.pbllibraries (ANSI and UTF-16LE, since the live libraries are Unicode after the 2025 migration) for object names, grouped by PowerBuilder naming convention.
- The root
.pblfiles are compiled binaries; the original PowerScript event bodies aren't recovered. The greenfield modules are rebuilt from the exported text source (DataWindows/windows) vendored underlegacy/src/and the authoritative Oracle DDL — not from guesses. - The data model is complete:
docs/DATA_MODEL.mdis regenerated from a live Oracle data-dictionary dump (412 tables, PKs/FKs/indexes/sequences). - The legacy target builds as committed: all 50 libraries in
lion.pbt's LibList are present (the 7 previously-missing PFE/PFD libraries were located and committed). - The one piece that stays DB-side by design is the per-customer 861 EDI generation (Oracle PL/SQL functions) — the greenfield API exposes the trigger point and documents the wiring.
*.pbl PowerBuilder libraries (compiled, binary; legacy)
*.srd *.srq *.src Exported DataWindow / query / event sources (text)
abis.pbw lion.pbt PowerBuilder workspace / target (legacy)
lion_mig.log PowerBuilder migration report (2025-08-26)
api/ Greenfield ASP.NET Core API + typed web modules (clientapp/)
edge/ Shop-floor edge service (serial scales/gauges → HTTP)
deploy/ Native Ubuntu installer (systemd unit, install/uninstall, nginx)
legacy/src/ Vendored PB text source for the rebuilt areas (reference)
docs/ Modernization discovery, data model & roadmap
tools/ Reproducible analysis extractors