@rdbms-erd/core

@rdbms-erd/core is the headless model/DDL package for rdbms-erd. It provides document types, validation/serialization, logical->physical type defaults, DDL generation, diagnostics, and host-extensible database metadata.

Installation

npm i @rdbms-erd/core

Quick Start

import { createEmptyDesign, createColumn, generateDdl } from "@rdbms-erd/core";

const doc = createEmptyDesign("postgres");
doc.model.tables.push({
    id: "table-users",
    logicalName: "Users",
    physicalName: "users",
    columns: [
        createColumn("postgres", {
            id: "col-users-id",
            logicalName: "ID",
            physicalName: "id",
            logicalType: "NUMBER",
            nullable: false,
            isPrimaryKey: true,
        }),
    ],
});

console.log(generateDdl(doc));

Core Concepts

• DesignDocument: ER JSON source of truth
• RdbmsDialect: dialect id stored in document
• LogicalDataType: normalized logical type ids
• DialectMetaJson: host-overridable per-dialect metadata
• DdlGeneratorHook: optional per-dialect SQL generation hook

Relationship Model Notes

RelationshipModel supports canvas-specific rendering state used by the designer:

• cardinality?: "1:1" | "1:N"
• canvasLineHidden?: boolean
• linePivotRatio?: number (middle vertical segment ratio)
• sourceLineY?: number (source edge absolute Y inside table)
• sourceLineRatio?: number (legacy fallback; kept for compatibility)

Host-Extensible DB Metadata

1) Override/append dialect metadata with JSON

Use hostMetas in options. Merge policy:

• same id: override builtin
• new id: append

import { resolveDialectMetas } from "@rdbms-erd/core";

const metas = resolveDialectMetas({
    hostMetas: [
        {
            id: "acme",
            label: "AcmeDB",
            supportsSchema: true,
            logicalTypes: [
                { id: "TEXT", defaultPhysicalType: "STRING(255)" },
                { id: "NUMBER", defaultPhysicalType: "INT64" },
            ],
            ddlStyle: { quote: "double", boolLiteral: "oneZero" },
        },
    ],
});

2) Optional per-dialect DDL function hook

If a dialect DDL is complex, provide a function in hostDdlGenerators. If omitted, style-based builtin SQL generation is used.

import {
    generateDdlForSelection,
    type DdlGenerateInput,
} from "@rdbms-erd/core";

const sql = generateDdlForSelection(doc, ["table-users"], {
    hostDdlGenerators: {
        acme: (input: DdlGenerateInput) => {
            // input.scope.kind: "all" | "selected"
            return { sql: "-- custom acme ddl" };
        },
    },
    fallbackOnHookError: true,
}).sql;

Main APIs

• Document lifecycle:
  • createEmptyDesign
  • serializeDesign, parseDesign, validateDesignDocument, roundTripDesign
• Type defaults / mapping:
  • defaultPhysicalType (logical → default physical for a dialect)
  • inferLogicalTypeFromPhysical (physical string → logical, dialect-aware)
  • convertPhysicalTypeByLogicalType (reshape physical for nextDialect while keeping logicalType)
  • createColumn
  • applyLogicalTypeChange
  • convertDesignDialect
• DDL:
  • generateDdl
  • generateDdlForSelection
  • generateIndexDdl
  • generateDdlWithDiagnostics
  • generateIndexDdlWithDiagnostics
• Metadata:
  • resolveDialectMetas
  • mergeDialectMetas
  • createDefaultDbMetaAdapter (legacy/advanced adapter path)

Notes

• DDL function hooks are runtime values and are not serialized in DesignDocument.
• DesignDocument stores only dialect id and model data.

License

MIT