statcatlib

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.

statcatlib is a Python library for building, managing, and querying statistical metadata catalogs.

Documentation: https://statcat-suite.github.io/statcatlib/

Inspired by DCAT and the broader landscape of metadata standards (SDMX, DDI), statcatlib provides a general-purpose engine for catalog storage, discovery, and retrieval — designed to be embedded in ETL pipelines, data platforms, and analytical tooling rather than used as a standalone application. It pairs naturally with domain-specific libraries like sdmxlib for standards-aware workflows.

import statcatlib as sc

What it does

statcatlib materializes remote catalog and structural metadata into a local catalog and exposes a fluent Python API for discovery queries — filtering by publisher, theme, dimension, code, free text, or provenance lineage.

A catalog is a directory of products. Each product is a chain of immutable editions, and each edition holds its datasets, its SDMX structures, its observations and its search index in one DuckDB database and its DuckLake.

The catalog is a materialized graph. Ingest pulls the full structural model into the edition's database, SQL JOINs are the graph traversal, and a query makes no network calls.

Installation

Requires Python 3.14+. Managed with uv.

uv add statcatlib
# or: pip install statcatlib

Published on PyPI as statcatlib (formerly pinaxlib, imported as pinax).

statcatlib is pre-1.0: until 1.0, a minor release may break the API.

Quick start

import statcatlib as sc
import statcatlib.query as q
import sdmxlib as sl


def german_datasets(view: sc.EditionView) -> sc.SearchResult[sc.AggregateDataset]:
    return (
        view.query(sc.AggregateDataset)
        .filter(q.has_code("geo", "DE"))
        .with_facets("themes", "frequency")
        .execute()
    )


# Create the catalog and start the product's working copy, its one unsealed edition
with sc.Catalog.create("my_catalog") as cat:
    wc = cat.product("estat").working_copy()
    with sl.RestRegistry(sl.Provider.ESTAT) as reg:
        reg.get(sl.Dataflow.urn_for(agency="ESTAT", id="NAMA_10_GDP"), resolve=True)
        sc.ingest_sdmx(wc, reg.registry)

    # wc.view() reads the working copy, writes included
    print(german_datasets(wc.view()).total, "datasets found")

    # Seal it into an immutable edition, with its search index, and record it live
    wc.seal()

# Open the catalog read-only and query the product's live edition, with no network calls
with sc.Catalog.open("my_catalog") as cat:
    results = german_datasets(cat.product("estat").view())
    print(results.total, "datasets found")
    print(results.facets["themes"])

Catalog.create(path) creates a catalog, and Catalog.create(path, exist_ok=True) reopens an existing one for writing. cat.product(id).working_copy() starts the product's one unsealed edition, a WorkingCopy. Write to it with put, the ingest_* functions and wc.registry, then seal() it, or discard() it. Catalog.open(path) opens a catalog read-only, and cat.product(id).view() returns an EditionView of the product's live edition, which every query below runs on. A view stays on its edition. After cat.refresh(), call product.view() again to read a newer one. statcatlib.testing.temporary_catalog() creates a throwaway catalog for tests and scripts. The on-disk layout is specified in docs/catalog-format.md.

Five dataset kinds

statcatlib uses a discriminated union of dataset types aligned with DCAT-AP and StatDCAT-AP:

# Generic datasets (CKAN, open portals)
sc.OpenDataset(identifier="co2-2024", title=i18n("CO2 Emissions 2024"), ...)

# Statistical tables with SDMX structure (dimensions, codelists)
sc.AggregateDataset(identifier="ESTAT:UNE_RT_M(1.0)", ..., sdmx_dataflow_urn="urn:...")

# Survey microdata with variable-level metadata
sc.MicrodataDataset(identifier="lfs-2023", ..., variables=[...])

# Spatial datasets with bounding box and CRS
sc.GeospatialDataset(identifier="boundaries-2024", ..., crs="EPSG:4326")

# Articles, reports, and analytical publications
sc.PublicationDataset(identifier="pub-71-607-x", ..., doi="10.25318/...", authors=[...])

All five share the same write and query API. Use sc.BaseDataset as the bound for generic code; sc.Dataset is a type alias for the full union.

Source connectors

SDMX (Eurostat, OECD, BIS, ABS, ...)

import sdmxlib as sl
import statcatlib as sc

with sc.Catalog.create("my_catalog", exist_ok=True) as cat:
    with cat.product("estat").working_copy() as wc:
        with sl.RestRegistry(sl.Provider.ESTAT) as reg:
            for df_id in ["NAMA_10_GDP", "UNE_RT_M", "PRIC_HPI_IDX"]:
                reg.get(sl.Dataflow.urn_for(agency="ESTAT", id=df_id), resolve=True)
            sc.ingest_sdmx(wc, reg.registry)

            # Write observations to the edition's DuckLake and bind them to the dataflow
            sc.ingest_data(wc, reg, "ESTAT:NAMA_10_GDP(latest)", measure_dim="na_item")
        wc.seal()

# Read them back through the edition's sdmxlib registry
with sc.Catalog.open("my_catalog") as cat:
    view = cat.product("estat").view()
    ds = view.get(sc.AggregateDataset, "ESTAT:NAMA_10_GDP(1.0)")
    obs = view.registry.obs(sl.Dataflow.parse_urn(ds.sdmx_dataflow_urn))
    wide = obs.pivot(on=ds.measure_dim)   # a polars.DataFrame

CKAN (open.canada.ca, data.gov, ...)

with sc.Catalog.create("my_catalog", exist_ok=True) as cat:
    with cat.product("open-canada").working_copy() as wc:
        sc.ingest_ckan(wc, "https://open.canada.ca/data", organization="statcan-statcan", limit=500)
        wc.seal()

Scope-based graph traversal

statcatlib exposes a lazy, scope-based API for navigating the catalog graph. Navigation builds scope objects without executing SQL; only terminal methods (.collect(), .count()) hit the database.

import statcatlib as sc

cat = sc.Catalog.open("my_catalog")
view = cat.product("estat").view()

# Navigate themes — no SQL until .collect()
concepts = view.themes["statcan"].collect()          # ItemList[Concept]
concept = view.themes["statcan"]["13"].collect()     # Concept

# Cross-entity navigation — .datasets returns a lazy QueryBuilder
datasets = view.themes["statcan"]["13"].datasets.collect()

# Enrich with sub-traversal expressions (like Polars' pl.col())
view.themes["statcan"].enrich(
    n=sc.each("datasets").count(),
    has_data=sc.each("datasets").exists(),
).collect()

# Codelist navigation — sc.urn builds URN strings for you
codes = view.codelist(sc.urn.codelist("SDMX", "CL_GEO")).collect()    # ItemList[Code]
code = view.codelist(sc.urn.codelist("SDMX", "CL_GEO"))["ON"].collect()  # Code

# CodelistsScope — parallel to ThemesScope, supports enrich
# Enriched output includes labels resolved via sdmx.localized_text
view.codelists.lang("en").enrich(n=sc.each("datasets").count()).collect()
# → [{"urn": "...", "label": "Geography", "n": 5}, ...]

# Filter codelists by label text (SQL-level, case-insensitive)
view.codelists.filter(text_contains="geo").enrich(n=sc.each("datasets").count()).collect()

view.codelist(urn).label("en")   # quick name lookup

# Enriched per-code output includes code name labels
view.codelist(urn).lang("en").enrich(n=sc.each("datasets").count()).collect()
# → [{"code_id": "ON", "label": "Ontario", "n": 5}, ...]

# Batch label resolution — single SQL query for many codes
view.codelist(geo_urn).batch_labels(["CA", "US", "DE"])
# → {"CA": "Canada", "US": "United States", "DE": "Germany"}

# Across multiple codelists at once
view.codelists.batch_labels([(geo_urn, "CA"), (freq_urn, "A")], lang="en")
# → {(geo_urn, "CA"): "Canada", (freq_urn, "A"): "Annual"}

# Dimension traversal
dims = view.dimensions(ds).collect()              # ItemList[DimensionInfo]
codelist = view.dimensions(ds)["GEO"].codelist    # CodelistScope (lazy)

Scope classes: ConceptSchemesScope, ConceptSchemeScope, ThemesScope, SchemeScope, ConceptScope, CodelistsScope, CodelistScope, CodeScope, DimensionsScope, DimensionScope.

Expression types: sc.each("edge") creates a context-free sub-traversal expression. Reusable across .enrich(), .filter(), and .sort_by().

Query API

Structured filters

import statcatlib.query as q

# Field filters (keyword arguments)
view.query(sc.AggregateDataset).filter(publisher="ESTAT", status="current").all()

# Composable filter objects
view.query(sc.AggregateDataset).filter(
    q.has_code("geo", "DE"),          # datasets with GEO=DE in their codelist
    q.has_dimensions(["geo", "freq"]), # datasets with both GEO and FREQ dimensions
).all()

# Hierarchy descent — any geography *under* Canada, strict (CA itself excluded).
# The applicable Hierarchy is resolved per dataflow; add include_self=True to
# match the root as well.
view.query(sc.AggregateDataset).filter(q.within("REF_AREA", "CA")).all()

# Distribution and service filters
view.query(sc.OpenDataset).filter(
    q.distribution(format="CSV"),
    q.has_service(endpoint_url="https://..."),
).all()

Cross-entity queries

# Agents that publish datasets with code GEO=CA
view.query(sc.Agent).filter(
    q.publishes(q.has_code("geo", "CA"), kind="aggregate")
).all()

# Data services serving aggregate datasets
view.query(sc.DataService).filter(
    q.serves(kind="aggregate")
).all()

MAP column filters

# Spatial coverage filter
view.query(sc.BaseDataset).filter(q.has_spatial("Canada")).all()

# Keyword filter
view.query(sc.BaseDataset).filter(q.has_keyword("employment")).all()

# Title/description search (case-insensitive ILIKE)
view.query(sc.BaseDataset).filter(q.title_contains("GDP")).all()
view.query(sc.BaseDataset).filter(q.description_contains("quarterly")).all()

# Sort by multilingual title
view.query(sc.BaseDataset).sort_by("title", lang="en").all()
view.query(sc.BaseDataset).sort_by("title", lang="en", desc=True).all()

Selective relationship loading

By default, querying a list of datasets loads all relationships (~17 queries). Use .include() to declare exactly which relationships to batch-load — the rest are set to an UNLOADED sentinel:

# Load publisher, themes and keywords only: 3 queries instead of ~17
results = (
    view.query(sc.BaseDataset)
    .filter(status="published")
    .include("publisher", "themes", "keywords")
    .sort_by("issued", desc=True)
    .limit(20)
    .all()
)

# Explicit full hydration shows the cost at the call site
results = view.query(sc.BaseDataset).filter(...).full().all()

# get() always loads everything — no include() needed
ds = view.get(sc.AggregateDataset, "ESTAT:NAMA_10_GDP(1.0)")

Unloaded fields raise sc.NotLoadedError on access. Use sc.is_unloaded(value) to check before accessing:

ds = view.query(sc.BaseDataset).include("publisher").first()
ds.publisher.name    # OK
ds.themes[0]         # raises NotLoadedError: 'themes' was not loaded

if not sc.is_unloaded(ds.themes):
    print(ds.themes)

Valid relationship names: publisher, contact_point, frequency, licence, themes, subject, dataset_type, keywords, spatial_coverage, distributions, conforms_to, quality_annotations, provenance, dimension_names, variables, feature_types, authors.

Lightweight projections

When you need just a few columns (e.g. autocomplete), projection modifiers bypass full object reconstruction — a single SQL query:

# Row projection — .select() + .rows() returns Row objects (dict subclass)
rows = (
    view.query(sc.BaseDataset)
    .filter(q.title_contains("GDP"))
    .sort_by("title", lang="en")
    .limit(3)
    .select("identifier", "title", lang="en")
    .rows()
)
# → [Row({"identifier": "GDP", "title": "GDP Growth"}), ...]
rows[0].identifier   # attribute-style access
rows[0]["title"]     # dict-style access — both work

# Flat value projection — .scalars() + .values() returns bare values
ids = view.query(sc.BaseDataset).filter(status="current").scalars("identifier").values()
# → ["EXR", "M1", "UNEMP", ...]

# Existence check — no object reconstruction
if view.query(sc.AggregateDataset).filter(q.has_code("geo", "DE")).exists():
    ...

Full-text search and facets

# BM25 search across titles, descriptions, keywords, themes, and dimensions
results = view.search("unemployment", limit=20)

# Combined search and filter
results = (
    view.query(sc.AggregateDataset)
    .filter(publisher="ESTAT")
    .search("labour force")
    .with_facets("themes", "frequency")
    .execute()
)
print(results.facets["themes"])   # {"Labour": 18, "Economy": 6, ...}

# Aggregation counts
counts = view.facets("publisher", "themes", "frequency", "status")

Lineage and provenance (PROV-O)

# Record derivation relationships between datasets, in a working copy
wc.add_lineage(
    "lfs-microdata",
    "14100287",
    "aggregated_from",
    activity_type="aggregation",
    activity_label="LFS monthly tabulation",
    confidence="asserted",
)

# Transitive upstream/downstream traversal on a view — returns QueryBuilder for chaining
ancestors = view.dataset("14100287").upstream(depth=5).collect()
dependents = view.dataset("CL_GEO").downstream(relationship="uses_classification").collect()

# Chain additional filters after traversal
current = view.dataset("14100287").upstream().filter(status="current").collect()

# Inspect lineage records
rows = view.dataset("14100287").lineage_records(role="target", relationship="aggregated_from")

Architecture

Each edition holds three layers in one DuckDB database and its DuckLake:

Discovery layer  — statcat schema     (dataset, agent, concept_scheme, concept, distribution, lineage)
Structural layer — sdmxlib tables     (dataflows, dsd_components, codes, codelists)
Observation layer — DuckLake tables  (actual time-series data)

statcatlib owns the discovery layer. sdmxlib owns the structural layer. A working copy writes both into its edition's database, and queries JOIN freely across both. ingest_data() writes each dataflow's observations to a DuckLake table through wc.registry.write_observations and registers an sdmxlib DataflowBinding, through which view.registry.obs(flow) reads them. docs/catalog-format.md specifies the files of a catalog and its editions.

Development

Run once per clone — installs a pre-push hook that runs the full gate, since hosted CI only builds the docs site:

just hooks-install

just ci                # full gate: index guard + lint + format + docstrings + prose + Any ratchet + pyrefly + unit and integration tests + strict docs build
just test              # unit tests
just test-integration  # integration tests (no network)
just test-live         # live tests against real SDMX endpoints
just lint              # ruff check
just typecheck         # pyrefly
just check-indexes     # only public PyPI is named anywhere in the repo
just docs              # local docs server

uv.lock is not committed: statcatlib is a library, so consumers resolve against pyproject.toml, and a lockfile would record whichever index resolved it. just check-indexes fails the build if a non-public package index is named in any tracked file.