ahmido
ff671d8d4a
## Why
Some Microsoft Graph / Intune identifiers are not UUIDs (e.g. scope tag id "0"). With `inventory_links.source_id` / `target_id` typed as `uuid`, PostgreSQL fails when inventory dependency extraction tries to persist those edges.
## What
- PostgreSQL migration changes `inventory_links.source_id` and `inventory_links.target_id` to `text`.
- Regression test ensures a non-UUID id ("0") can be persisted; on pgsql it also asserts the columns are `text`.
## Notes
- UUID identifiers continue to work (stored as strings).
- No UI/Filament changes.
## Testing
- `vendor/bin/sail artisan test --compact tests/Feature/Inventory/InventoryLinksNonUuidIdsTest.php`
Co-authored-by: Ahmed Darrazi <ahmeddarrazi@MacBookPro.fritz.box>
Reviewed-on: #96
23 lines
835 B
Markdown
23 lines
835 B
Markdown
# Spec 079: Inventory links support non-UUID IDs
|
|
|
|
**Date**: 2026-02-07
|
|
|
|
## Problem
|
|
|
|
Inventory dependency extraction writes edges into `inventory_links`. Some Microsoft Graph / Intune identifiers (notably scope tag IDs) can be non-UUID strings (e.g. `"0"`). The current schema defines `inventory_links.source_id` and `inventory_links.target_id` as UUID columns, causing PostgreSQL failures when non-UUID identifiers are inserted.
|
|
|
|
## Goal
|
|
|
|
Allow storing non-UUID identifiers in `inventory_links` without crashing inventory sync/extraction.
|
|
|
|
## Requirements
|
|
|
|
- `inventory_links.source_id` and `inventory_links.target_id` must accept arbitrary string identifiers.
|
|
- Existing UUID identifiers must continue to work.
|
|
- Behavior must be covered by tests.
|
|
|
|
## Non-goals
|
|
|
|
- No redesign of the dependency graph model.
|
|
- No UI/Filament changes.
|