# Complete Schema Mermaid Diagram Generation

**Date**: 2025-11-24  
**Feature**: Extended LinkML's MermaidRenderer to generate complete schema diagrams

---

## Overview

LinkML's built-in `MermaidClassDiagramGenerator` generates **individual diagrams per class** (53 separate `.mmd` files). While this is useful for detailed class-by-class documentation, it doesn't provide a **holistic view of the entire schema**.

We've created a custom script that generates a **single comprehensive Mermaid diagram** showing:
- ✅ All 53 classes
- ✅ All 149 relationships (inheritance, associations)
- ✅ Key attributes for each class (limited to 10 per class for readability)
- ✅ Abstract classes marked with `<<abstract>>`
- ✅ Relationship cardinalities (`1`, `1..*`)

---

## Usage

### Generate Complete Diagram

```bash
cd /Users/kempersc/apps/glam
python3 scripts/generate_complete_mermaid_diagram.py
```

**Output**:
```
schemas/20251121/uml/mermaid/complete_schema_YYYYMMDD_HHMMSS.mmd
```

### View the Diagram

**Option 1: Mermaid Live Editor**
1. Open https://mermaid.live/
2. Copy contents of `complete_schema_YYYYMMDD_HHMMSS.mmd`
3. Paste into editor
4. Explore interactively (zoom, pan, export to SVG/PNG)

**Option 2: VS Code (with Mermaid extension)**
1. Install "Markdown Preview Mermaid Support" extension
2. Open `.mmd` file in VS Code
3. Right-click → "Open Preview"

**Option 3: GitHub (automatic rendering)**
1. Push `.mmd` file to GitHub repository
2. View file on GitHub (automatic Mermaid rendering)

---

## Generated Diagram Statistics

**Latest Generation** (2025-11-24 00:41:43):
- **Classes**: 53
- **Relationships**: 149
  - Inheritance (`<|--`): 42
  - Associations (`-->`): 107
- **Lines**: 697
- **File Size**: 31 KB

---

## Key Relationships Captured

### EncompassingBody Hierarchy
```mermaid
classDiagram
  class EncompassingBody <<abstract>>
  EncompassingBody <|-- UmbrellaOrganisation : inherits
  EncompassingBody <|-- NetworkOrganisation : inherits
  EncompassingBody <|-- Consortium : inherits
  
  Custodian --> EncompassingBody : encompassing_body
```

### CustodianType Hierarchy (19 types)
```mermaid
classDiagram
  class CustodianType
  CustodianType <|-- MuseumType
  CustodianType <|-- LibraryType
  CustodianType <|-- ArchiveOrganizationType
  CustodianType <|-- GalleryType
  CustodianType <|-- OfficialInstitutionType
  CustodianType <|-- ResearchOrganizationType
  CustodianType <|-- CommercialOrganizationType
  CustodianType <|-- UnspecifiedType
  CustodianType <|-- BioCustodianType
  CustodianType <|-- EducationalProviderType
  CustodianType <|-- CollectingSocietyType
  CustodianType <|-- FeaturePlaceType
  CustodianType <|-- IntangibleHeritageType
  CustodianType <|-- MixedCustodianType
  CustodianType <|-- PersonalCollectionType
  CustodianType <|-- HolySiteType
  CustodianType <|-- DigitalPlatformType
  CustodianType <|-- NonProfitType
  CustodianType <|-- TasteScentHeritageType
```

### Hub Architecture (Custodian as central reference)
```mermaid
classDiagram
  class Custodian
  CustodianObservation --> Custodian : identifies_custodian
  CustodianLegalStatus --> Custodian : refers_to_custodian
  CustodianName --> Custodian : refers_to_custodian
  CustodianPlace --> Custodian : refers_to_custodian
  CustodianCollection --> Custodian : refers_to_custodian
  OrganizationalStructure --> Custodian : refers_to_custodian
```

---

## Script Details

**Location**: `scripts/generate_complete_mermaid_diagram.py`

**Key Features**:
1. **Two-pass generation**:
   - Pass 1: Define all classes with attributes
   - Pass 2: Define all relationships (inheritance + associations)

2. **Smart filtering**:
   - Limits attributes to 10 per class (prevents diagram explosion)
   - Only includes class-to-class relationships (skips primitive types)

3. **Metadata embedded**:
   - Schema name and version
   - Generation timestamp
   - Class/relationship counts

4. **Timestamped output**:
   - Format: `complete_schema_YYYYMMDD_HHMMSS.mmd`
   - Allows multiple generations without conflicts

---

## Comparison: Per-Class vs Complete Diagrams

| Feature | Per-Class Diagrams (gen-yuml) | Complete Diagram (this script) |
|---------|-------------------------------|--------------------------------|
| **Files generated** | 53 (one per class) | 1 (entire schema) |
| **Total size** | 212 KB (53 files) | 31 KB (1 file) |
| **Use case** | Detailed class documentation | Schema overview and architecture |
| **Relationships shown** | Immediate class relationships | All schema relationships |
| **Holistic view** | ❌ Requires viewing 53 files | ✅ Single comprehensive view |
| **Detail level** | High (all slots, all docs) | Medium (10 slots per class) |
| **Maintainability** | Auto-generated by LinkML | Custom script (easy to extend) |

**Recommendation**: Use **both**:
- Per-class diagrams: Detailed reference for developers
- Complete diagram: Architecture overview for onboarding, presentations, ontology consultations

---

## Extending the Script

### Add More Attributes Per Class

Edit `scripts/generate_complete_mermaid_diagram.py` line 72:
```python
for slot_name in class_slots[:10]:  # Change 10 to 20, 30, etc.
```

### Include Enum Values

Add after line 86:
```python
# Add enums
for enum_name in sorted(schemaview.all_enums()):
    enum = schemaview.get_enum(enum_name)
    mermaid_lines.append(f"  class {enum_name} <<enumeration>>")
    for pv in enum.permissible_values.values():
        mermaid_lines.append(f"    {enum_name} : {pv.text}")
    mermaid_lines.append("")
```

### Filter to Specific Classes

Add after line 69:
```python
# Only include core classes
core_classes = ['Custodian', 'CustodianObservation', 'CustodianLegalStatus', 
                'CustodianName', 'CustodianPlace', 'EncompassingBody']
if class_name not in core_classes:
    continue
```

---

## Troubleshooting

### Diagram Too Large (Mermaid.live won't render)

**Solution 1**: Reduce attributes per class
```python
for slot_name in class_slots[:5]:  # Fewer attributes
```

**Solution 2**: Filter to core classes only (see "Extending" above)

**Solution 3**: Generate multiple diagrams by module
```python
# Filter by module
if 'CustodianType' in class_name:
    # CustodianType hierarchy diagram
```

### Missing Relationships

**Check**: Ensure slot `range` is a class name:
```python
if slot.range and slot.range in schemaview.all_classes():
    # Only creates relationship if range is a class
```

**Fix**: Some relationships may use `uriorcurie` (string reference) instead of class range. Script currently filters these out to avoid noise.

---

## Next Steps

### Generate SVG/PNG Exports

Install Mermaid CLI:
```bash
npm install -g @mermaid-js/mermaid-cli
```

Convert to SVG:
```bash
mmdc -i complete_schema_20251124_004143.mmd -o complete_schema.svg
```

### Add to Documentation Website

If building with Sphinx/MkDocs:
```markdown
## Schema Overview

```mermaid
{{ include('schemas/20251121/uml/mermaid/complete_schema_20251124_004143.mmd') }}
\```
```

### Automate Generation in CI/CD

Add to `.github/workflows/schema-docs.yml`:
```yaml
- name: Generate Complete Mermaid Diagram
  run: python3 scripts/generate_complete_mermaid_diagram.py
```

---

## References

- **LinkML Documentation**: https://linkml.io/linkml/generators/mermaid.html
- **Mermaid Class Diagrams**: https://mermaid.js.org/syntax/classDiagram.html
- **Mermaid Live Editor**: https://mermaid.live/
- **Script Location**: `scripts/generate_complete_mermaid_diagram.py`
- **Output Directory**: `schemas/20251121/uml/mermaid/`

---

## Changelog

**2025-11-24**: Initial implementation
- Created custom script extending LinkML's MermaidRenderer
- Generated complete schema diagram (53 classes, 149 relationships)
- Documented usage and extension patterns