sdmxlib 0.44.1

SDMX structural metadata library for Python

sdmxlib

Note: This library has been written extensively with AI assistance (Claude Code). Users who are not comfortable with AI-generated code should take that into account before adopting it.

SDMX structural metadata library for Python. Fetch and navigate Data Structure Definitions, Codelists, Dataflows, and related artefacts from any SDMX REST endpoint — with plain attribute access, no XML or URN chasing required.

Quick start

import sdmxlib as sl

with sl.RestRegistry(sl.Provider.BIS) as reg:
    dsd = reg.get(sl.DataStructure, agency="BIS", id="WS_CBPOL").resolve()

    for dim in dsd.dimensions:
        if dim.is_enumerated:
            cl = dim.representation()
            print(f"{dim.id}: {cl.id} ({len(cl.codes)} codes)")

FREQ: CL_FREQ (7 codes)
REF_AREA: CL_BIS_IF_REF_AREA (44 codes)
INSTR_ASSET: CL_INSTR_ASSET (12 codes)
BORROWER_CTY: CL_BIS_IF_REF_AREA (44 codes)

Installation

uv add sdmxlib
# or: pip install sdmxlib

What it does

• Connects to any SDMX 2.1 or 3.0 REST endpoint; built-in shortcuts for ECB, BIS, ESTAT, IMF, OECD, and the SDMX Global Registry via sl.Provider
• Three resolution modes: browse() for catalogue scanning, fetch() for unresolved artefacts, resolve() for the full reference graph (with optional constraint-trimming via partial=True/False)
• Returns fully resolved domain objects with plain attribute access — no XML, no URN chasing
• Filter collections with lambdas or the item() expression API
• Build SDMX artefacts locally and serialise to SDMX-ML or SDMX-JSON
• Registry -- a DuckDB-backed persistent store for offline access with lazy codelist + hierarchy loading, automatic Ref resolution, and direct SQL code maps; supports shared connections and any-language queries
• Uniform SKOS-aligned tree API (roots/leaves/narrower/broader/ *_transitive/depth_of plus bulk projections depths/broader_map/ ancestors_map/inherit_along_broader) on Codelist, Hierarchy, CategoryScheme, MetadataStructure — and their lazy SQL variants LazyCodelist/LazyHierarchy
• Reference metadata via MetadataStructure (MSD), Metadataflow, and MetadataSet — attach typed footnotes, methodology notes, or quality reports to any identifiable artefact. Definitions ride in StructureMessage; populated payloads ride in MetadataMessage (<mes:GenericMetadata> in SDMX-ML 3.0; metadata-schema envelope in SDMX-JSON 2.0). Both message kinds round-trip through Registry, JSON, and XML.
• Integrates with Polars for schema generation and data validation
• Fully typed — IDE completion and static analysis work throughout

For AI assistants

Two condensed reference files at the repo root are written for LLM tools (Claude Code, Cursor, Copilot, etc.):

• llms.txt — full library overview: install, registries, data fetching, reference metadata, and the domain-type cheat sheet
• llms-dsl.txt — focused query-DSL reference: sl.item / sl.code / sl.node / sl.dim predicates, Query / DataQuery / CodeQuery / HierarchyNodeQuery chains, and the uniform SKOS tree API

Point your assistant at these to keep generated code aligned with the library's actual surface.

Documentation

See the docs for the full cookbook, including:

• Connecting to registries and choosing a resolution mode
• Persisting artefacts in a local DuckDB store with Registry
• Fetching observation data with sl.dim() filters and pivoting to wide format
• Inspecting DSDs and building code maps for ETL pipelines
• Generating Polars schemas from SDMX Data Structure Definitions
• Browsing data catalogs and filtering dataflows
• Working with constraints, hierarchies, and serialisation

Development

just check           # lint + typecheck
just test            # unit tests
just test-integration  # live endpoint tests (requires network)
just ci              # full pre-commit gate