fhir4ds.sqlquery

The SQL-on-FHIR v2 Analytics Layer: FHIR Library resources that conform to the SQLQuery and SQLView profiles. A SQLQuery carries typed, FHIR-declared parameters plus SQL content, so reusable analytics queries can travel as FHIR resources — stored in repositories, referenced by canonical URL, and executed anywhere the FHIR4DS engine runs.

1. At a Glance

from fhir4ds.sqlquery import (
    parse_library,      # Library dict/JSON -> SQLQuery or SQLView
    parse_sqlquery,     # Strict parse against the SQLQuery profile
    parse_sqlview,      # Strict parse against the SQLView profile
    SQLQueryRunner,     # Execute against a DuckDB connection
)

A SQLQuery is a FHIR Library whose content entry holds SQL (base64 data, contentType application/sql). Queries may declare parameters — referenced in the SQL as $name — and reference other queries or ViewDefinitions through relatedArtifact canonical URLs. Each depends-on label becomes a virtual table in the SQL: the runner resolves the dependency graph, materializes each dependency as a view, and executes the query against them.

A SQLView is the parameter-free variant (parameter 0..0) — a reusable named query that other queries can depend on.

2. Function Reference

parse_library

parse_library(input_) -> SQLQuery | SQLView

Parse a Library dict or JSON string into its profiled form. The profile is detected from meta.profile; unknown profiles fall back to SQLQuery.

Parameter	Type	Description
input_	dict | str	A Library object or its JSON text.

parse_sqlquery

parse_sqlquery(input_) -> SQLQuery

Strict parse against the SQLQuery profile — validates the canonical, required fields (status, type, content dialect), and parameter declarations (primitive FHIR types only).

parse_sqlview

parse_sqlview(input_) -> SQLView

Strict parse against the SQLView profile; rejects any parameter declaration.

---

3. Class Reference

SQLQueryRunner

SQLQueryRunner(connection, resolver, *, source_table="resources")

Executes SQLQuery / SQLView resources against a DuckDB connection. Each dependency is materialized with CREATE OR REPLACE VIEW, so re-running the same query is idempotent.

Parameter	Type	Description
connection	duckdb.DuckDBPyConnection	Caller-supplied connection.
resolver	DependencyResolver	Callable mapping a canonical URL to the referenced resource — a parsed ViewDefinition/SQLQuery/SQLView, a ViewDefinition dict, or a Library dict. May raise to signal "not found".
source_table	str	The physical table that ViewDefinition dependencies read from (default "resources").

Methods

• execute(library, parameters=None) -> list[tuple]: Execute a parsed SQLQuery/SQLView (or a Library dict). parameters is a name → value dict, coerced to the declared FHIR types; required parameters must be present. Raises typed errors for unresolvable relatedArtifact (SQLQueryMaterializationError), dependency cycles (SQLQueryCycleError), parameter type mismatches or missing parameters (SQLQueryTypeError), and unsupported content dialects (UnsupportedDialectError).

---

4. Example

import base64
import fhir4ds
from fhir4ds.sqlquery import parse_sqlquery, SQLQueryRunner, SQLQUERY_PROFILE_CANONICAL

def b64(text):
    return base64.b64encode(text.encode()).decode()

vd_canonical = "https://example.org/ViewDefinition/PatientIds"
vd_dict = {
    # SQL-on-FHIR v2 typed resourceType (no "name" on the root select).
    "url": vd_canonical,
    "resource": "Patient",
    "select": [{"column": [{"path": "id", "name": "id", "type": "id"}]}],
}

query = {
    "resourceType": "Library",
    "url": "https://example.org/SQL/Seniors",
    "meta": {"profile": [SQLQUERY_PROFILE_CANONICAL]},
    "status": "active",
    "type": {"coding": [{"code": "sql-query"}]},
    # The label "pt" becomes a virtual table, resolved via the canonical.
    "relatedArtifact": [
        {"type": "depends-on", "label": "pt", "resource": vd_canonical},
    ],
    "parameter": [{"name": "min_year", "type": "integer", "use": "in"}],
    "content": [{
        "contentType": "application/sql",
        "data": b64("SELECT id FROM pt WHERE id LIKE $min_year || '%'"),
    }],
}

con = fhir4ds.create_connection()
runner = SQLQueryRunner(
    connection=con,
    resolver=lambda canonical: vd_dict if canonical == vd_canonical else None,
)
rows = runner.execute(parse_sqlquery(query), parameters={"min_year": 1970})