Skip to content

evolution.md

Latest commit

 

History

History
92 lines (71 loc) · 3.54 KB

evolution.md

File metadata and controls

92 lines (71 loc) · 3.54 KB
Raw

XRM — Project Evolution

How a 3-hour hack became a production-ready framework in two weeks.

Timeline

Day 1 (Apr 20) — The 3-hour MVP

Started as a quick proof-of-concept: can you build a generic CRM-like platform where entities, fields, and relationships are all configurable at runtime?

  • Requirements document
  • Single-project Blazor Server app with EF Core + SQLite
  • Entity/field/relationship designers
  • Basic record CRUD with dynamic forms
  • Demo data seeded on startup

Result: A working prototype — define entities, add fields, create records. Ugly but functional.

Week 1 (Apr 20–27) — Foundation

Rapid iteration to make it real:

  • Service layer refactoring — extracted IEntityService, IRecordService, IRelationshipService
  • Child records — inline CRUD on parent detail pages with relationship links
  • Testing — 33 unit and API tests, CI workflow (GitHub Actions)
  • REST API — full CRUD endpoints for entities, fields, relationships, and records
  • Documentation — user guide with screenshots, architecture doc, MIT license
  • Reuse refactor — split into Xrm.Core (models/services) + Xrm.Blazor (RCL) + Xrm.Server (thin host)

Result: 71 tests, clean architecture, ready for consumption by domain projects.

Week 2 (Apr 28 – May 4) — Feature CRs & Production Hardening

Driven by a domain project consuming the framework:

  • CR-001 MultiChoice fields
  • CR-002 AutoNumber fields
  • CR-003/020 Computed fields (formulas + aggregates)
  • CR-005 Parent/Child relationship naming
  • CR-006 Domain metadata on entities
  • CR-007 Identity & authorization (domain-scoped RBAC)
  • CR-013 Record lifecycle hooks
  • CR-014 State machine transitions
  • CR-015 Cross-field validation rules
  • CR-016 Audit trail
  • CR-021 Locale-aware formatting
  • CR-022/023 Child section refresh + post-save warning hooks

Plus UX polish:

  • Collapsible nav groups, filter/sort on all lists, record navigation (prev/next)
  • User identity chip, access denied handling, role-based UI visibility
  • Entra ID integration guide for consumers

Result: 104 tests, ~8,000 lines of code, 73 commits, production-deployed.

Architecture evolution

Week 0:  [Single Blazor project, everything in one place]
            ↓
Week 1:  [Xrm.Core]  ←  [Xrm.Blazor RCL]  ←  [Xrm.Server host]
            ↓                    ↓
Week 2:  [Xrm.Core]  ←  [Xrm.Blazor RCL]  ←  [Any host project]
           services          UI components        (Your app, etc.)
           models            pages
           API controllers   layouts
           auth layer

Key design decisions

Decision Rationale
SQLite + EF Core Zero infrastructure, single-file DB, good enough for most deployments
JSON data storage (DataJson) Schema-flexible records without migrations per entity
RCL pattern UI reusable across multiple host apps without duplication
ICurrentUser abstraction Provider-agnostic auth — works with Entra, Google, anonymous
Domain-scoped roles Multi-tenant within one deployment, not per-database
Service-layer enforcement Auth checks can't be bypassed by navigating to URLs
AnonymousCurrentUser default Dev inner loop works without any auth setup

Stats

Metric Value
Time span 2 weeks (Apr 20 – May 4, 2026)
Commits 73
Code ~8,000 lines (C# + Razor)
Tests 104 (unit + API integration)
Docs 10 guides + architecture + backlog
Consumers 2 (domain app, XRM standalone)