Seed Dataset Generator Plugin

The Seed Dataset Generator is a built-in Schema Builder translator plugin that converts your active JSON-LD schema into realistic SQL INSERT statements — ready to load into DuckDB or PostgreSQL for analytics dashboards, integration testing, or Superset demos.

---

The Full Pipeline

JSON-LD Schema (active in Schema Builder)
    ↓  Mapper View  →  MappingRow[]
    ↓  sql-duckdb plugin  →  schema.sql     (CREATE TABLE + FK constraints)
    ↓  fake-data-sql plugin  →  inserts.sql  (N rows per table, faker.js)
    ↓  duckdb clinical.duckdb < schema.sql
    ↓  duckdb clinical.duckdb < inserts.sql
    ↓  Apache Superset  →  Live Dashboard

---

Quick Start

1. Open your schema and map nodes

1. Open a .jsonld file in Schema Builder.
2. Open the Mapper View (toolbar → map icon).
3. Nodes auto-load; review each row's mapTo (table/column name) and targetType.
4. Select all rows you want to include (checkbox column).

2. Generate the DDL schema

In Target Format, choose SQL DDL (DuckDB) and click Export → save as schema.sql.

3. Generate seed data

In Target Format, choose Seed Data — SQL INSERTs (faker.js) and click Export → save as inserts.sql.

4. Load into DuckDB

Using the Docker sidecar from the SureCentric stack:

# Start the stack
docker compose up -d postgres redis superset duckdb

# Apply schema
docker compose run --rm duckdb sh -c \
  "duckdb /data/clinical.duckdb < /generated/schema.sql"

# Load seed rows
docker compose run --rm duckdb sh -c \
  "duckdb /data/clinical.duckdb < /generated/inserts.sql"

Or with a local DuckDB binary:

duckdb clinical.duckdb < schema.sql
duckdb clinical.duckdb < inserts.sql

5. Connect Apache Superset

1. Open Superset at http://localhost:8088 (admin / admin).
2. + Database → DuckDB — connection string:

   duckdb:////data/clinical.duckdb

3. Add a dataset for each table.
4. Build charts: count by status, enrollment timeline, diagnosis distribution.

---

What Gets Generated

Row count

Each owl:Class table defaults to 50 rows. To override, enter a number in the notes field of the TABLE mapping row (e.g. 100 generates 100 rows).

IRI primary key

Every row gets a stable IRI primary key:

https://<sourceUri>/patient/1
https://<sourceUri>/patient/2
...

Column value heuristics

The plugin uses a two-tier value strategy:

1. Column name heuristics (clinical-domain aware)

Checked first — highest priority. Column names are matched against patterns:

Column name pattern	Generated value
patient_name, subject_name, full_name	faker.person.fullName()
first_name, given_name	faker.person.firstName()
last_name, family_name, surname	faker.person.lastName()
email	faker.internet.email()
phone, telephone	faker.phone.number()
site, center, hospital, facility, institution	faker.company.name()
country	faker.location.country()
country_code	faker.location.countryCode()
city	faker.location.city()
state, province, region	faker.location.state()
zip, postal	faker.location.zipCode()
address, street	faker.location.streetAddress()
description, comment, note, remark	faker.lorem.sentence()
title, label, name	faker.lorem.words(2)
status	One of: active, inactive, pending, completed, withdrawn
gender, sex	One of: M, F, U
race, ethnicity	Standard clinical race/ethnicity values
dose, dosage, amount	Float 0.5 – 1000
unit	One of: mg, mL, mg/kg, IU, mcg
code	6-char alphanumeric e.g. A7X3K9
age	Integer 18 – 90
weight	Float 40 – 150
height	Float 140 – 210
score, value, result, measure	Float 0 – 100

2. targetType fallback

If no name heuristic matches:

targetType	Generated value
INTEGER, INT, BIGINT	faker.number.int({min:1, max:9999})
FLOAT, DOUBLE, DECIMAL	faker.number.float({fractionDigits:2})
BOOLEAN	TRUE or FALSE
DATE	Past date up to 5 years ago: '2023-04-15'
DATETIME, TIMESTAMP	Past datetime: '2023-04-15 09:32:11.000'
UUID	faker.string.uuid()
URI, IRI, REFERENCE	faker.internet.url() — or FK reference (see below)
TEXT, STRING, VARCHAR (default)	faker.lorem.words(3)

Foreign key resolution

When a column is detected as a foreign key (via inferForeignKey() — same logic as the SQL DDL plugins), the generated value is a valid IRI from the referenced table, picked round-robin across its generated rows.

This preserves referential integrity — every FK column references a row that actually exists.

Example: if diagnosis has a patient column typed as URI or IRI, and patient is a known table, the patient column will contain values like:

'https://example.com/schema/patient/1'
'https://example.com/schema/patient/2'
...

---

Output Format

The generated inserts.sql looks like:

-- Auto-generated seed data from: https://example.com/eclinical
-- Schema Builder — Seed Dataset Generator (faker.js)
-- Generated: 2026-04-09T10:48:57Z
-- Rows per table: configured in notes field (default 50)

-- patient (50 rows)
INSERT INTO patient (iri, patient_name, site, status, age, gender, race) VALUES
  ('https://example.com/eclinical/patient/1', 'Margaret O''Brien', 'Acme Medical Center', 'active', 34, 'F', 'White'),
  ('https://example.com/eclinical/patient/2', 'James Thornton', 'Harborview Hospital', 'active', 51, 'M', 'Hispanic'),
  ...;

-- diagnosis (50 rows)
INSERT INTO diagnosis (iri, patient, diagnosis_code, description, onset_date) VALUES
  ('https://example.com/eclinical/diagnosis/1', 'https://example.com/eclinical/patient/1', 'D3K7A2', 'At the house.', '2024-01-22'),
  ...;

---

Reproducibility

The generator uses faker.seed(42) — every export of the same mapping produces identical data. This is useful for:

• Repeatable integration tests
• Consistent demo datasets committed to source control
• QA scripts that rely on known row content

To get different data, pass the output through sed 's/faker.seed(42)/faker.seed(999)/' or modify the plugin source.

---

Plugin Identity

Field	Value
id	com.sureclinical.seed-data-generator
pluginType	translator
formatId	seed-data-sql
version	1.0.0
author	SureClinical
Output extension	.inserts.sql

---

Advanced: Building a Superset Dashboard

After loading schema.sql + inserts.sql into DuckDB, here is a quick dashboard recipe:

Chart type	Dataset	Metric	Dimension
Big Number	patient	COUNT(*)	—
Bar Chart	diagnosis	COUNT(*)	diagnosis_code
Time-series	patient	COUNT(*)	onset_date (truncate to month)
Pie Chart	patient	COUNT(*)	status
Table	diagnosis	* columns	—

Save as dashboard "eClinical Model — Live" and share the URL as the integration demo artifact.

---

See Also

• Developer Reference — full Schema Builder architecture
• Plugin Authoring Guide — how to build your own translator plugin
• CARD System — provision seed data directly into Superset via CARD Profiles