Files
relspecgo/pkg/readers/dbml/README.md
T
warkanum 2f69205aa0 fix(dbml): resolve commented cross-file // Ref: lines
Commented refs are collected per file and resolved against the combined
model after all inputs are loaded (directory, --from-list, merge, jobs).
Matched refs become FKs and relationships; duplicates of existing FKs are
skipped; missing targets are skipped with a warning; column type
mismatches warn.

Also keep reused index names within a DBML table instead of overwriting,
give a second FK to the same table a distinct relationship name, and make
the pgsql writer match relationships to FKs by name first.
2026-09-23 18:48:07 +02:00

163 lines
4.5 KiB
Markdown

# DBML Reader
Reads Database Markup Language (DBML) files and extracts database schema information.
## Overview
The DBML Reader parses `.dbml` files that define database schemas using the DBML syntax (used by dbdiagram.io) and converts them into RelSpec's internal database model representation.
## Features
- Parses DBML syntax
- Extracts tables, columns, and relationships
- Supports DBML-specific features:
- Table groups and notes
- Enum definitions
- Indexes
- Foreign key relationships
## Usage
### Basic Example
```go
package main
import (
"fmt"
"git.warky.dev/wdevs/relspecgo/pkg/readers"
"git.warky.dev/wdevs/relspecgo/pkg/readers/dbml"
)
func main() {
options := &readers.ReaderOptions{
FilePath: "/path/to/schema.dbml",
}
reader := dbml.NewReader(options)
db, err := reader.ReadDatabase()
if err != nil {
panic(err)
}
fmt.Printf("Found %d schemas\n", len(db.Schemas))
}
```
### CLI Example
```bash
# Read DBML file and convert to JSON
relspec --input dbml --in-file schema.dbml --output json --out-file schema.json
# Convert DBML to GORM models
relspec --input dbml --in-file database.dbml --output gorm --out-file models.go
```
## Example DBML File
```dbml
Table users {
id bigserial [pk, increment]
username varchar(50) [not null, unique]
email varchar(100) [not null]
created_at timestamp [not null, default: `now()`]
Note: 'Users table'
}
Table posts {
id bigserial [pk]
user_id bigint [not null, ref: > users.id]
title varchar(200) [not null]
content text
indexes {
user_id
(user_id, created_at) [name: 'idx_user_posts']
}
}
Ref: posts.user_id > users.id [delete: cascade]
```
## DBML Features Supported
- Table definitions with columns
- Primary keys (`pk`)
- Not null constraints (`not null`)
- Unique constraints (`unique`)
- Default values (`default`)
- Inline references (`ref`)
- Standalone `Ref` blocks
- Commented cross-file refs (`// Ref:` — see below)
- Indexes and composite indexes
- Table notes and column notes
- Enums
- Dialect directives (`@postgres:` / `@sqlite:` — see below)
## Commented cross-file refs
`// Ref:` / `// ref:` lines (ignored by dbdiagram) become FKs + relationships once both ends are loaded.
| Rule | Behaviour |
|---|---|
| When | Single file / directory: end of read. `--from-list`, `merge`, jobs: after all inputs are combined |
| Match | `schema.table.column` on both sides, case-insensitive |
| Operators | `>`, `<`, `-` (parsed like `Ref:`) |
| Duplicate of an FK on the same columns | Skipped silently |
| Target not loaded | Kept pending; skipped with a warning on the final pass |
| Column type mismatch | Warning, FK still created (`serial`≈`integer`, `bigserial`≈`bigint`) |
| Pending state | `Database.Metadata["dbml.commented_refs"]` (`[]string`) |
API: `dbml.ResolveCommentedRefs(db, final bool) []string` returns warnings.
## Dialect directives
Lines of the form `@<namespace>[(<column>)]: <args>` embed database-specific
features that plain DBML cannot express (partitioning, `WITHOUT ROWID`,
tablespaces, index storage parameters, …). They are stored losslessly on the
relevant object's `Metadata` and round-trip unchanged through the DBML writer;
the PostgreSQL and SQLite writers translate the ones they understand to SQL.
```dbml
@postgres: search_path myapp
Table myapp.events {
id bigint [pk]
created_at timestamp [not null]
@postgres(id): identity always
@postgres: partition by RANGE (created_at)
@sqlite: without rowid
indexes {
(created_at) [name: 'idx_events_created']
@postgres: with (fillfactor=90)
}
}
```
| Position | Attaches to |
|----------|-------------|
| Before the first `Table {` | database |
| Table body, no `(target)` | that table |
| Table body, `(col)` target | column `col` (error if unknown) |
| Inside `indexes { }` | the most recently listed index entry |
`args` is preserved verbatim; the **key** (lowercased first token) drives
duplicate detection. Repeated directives are kept in order; catalog "singleton"
keys error on a second occurrence at the same location. All errors are
line-numbered.
`ReaderOptions.StrictDirectives` (CLI `--strict-directives`) turns an unknown
namespace or key into an error instead of preserving it silently.
See [`docs/DBML_DIRECTIVES.md`](../../../docs/DBML_DIRECTIVES.md) for the full
grammar and the supported-directive matrix.
## Notes
- DBML is designed for database documentation and diagramming
- Schema name defaults to `public`
- Relationship cardinality is preserved