Package {rdatagouv}

Title:	Tools to Download and Explore Datasets from Data.gouv.fr
Version:	0.1.0
Description:	Provides a client for the public API of data.gouv.fr, the French government's open data platform. It helps you find a dataset that matches your interests, judge whether it is usable, download it, and re-fetch the exact same table later in a reproducible way. You can search the catalog and filter by producer or theme (dg_find_datasets(), dg_find_organization(), dg_find_topics()), pull a dataset's tabular resources into tidy tibbles (dg_pull_dataset()), inspect the documented variables of its data schema (dg_schema()), and compute summary metrics such as size, number of columns and missing-value rate (dg_summary(), dg_summarise()). Each returned table carries a stable identifier (dg_table_id(), dg_refetch()) so it can be re-fetched later. Requests are built on top of 'httr2'.
License:	MIT + file LICENSE
URL:	https://astamm.github.io/rdatagouv/, https://github.com/astamm/rdatagouv
BugReports:	https://github.com/astamm/rdatagouv/issues
Encoding:	UTF-8
VignetteBuilder:	quarto
Config/Quarto/version:	>= 1.4
Imports:	cli, httr2, jsonlite, nanoparquet, readxl, tibble, vroom
Suggests:	dplyr, gt, knitr, pkgdown, quarto, testthat (≥ 3.0.0), withr, writexl
Config/testthat/edition:	3
Config/roxygen2/version:	8.1.0
NeedsCompilation:	no
Packaged:	2026-09-03 11:05:38 UTC; stamm-a
Author:	Magali Berland [aut], Pierre Gloaguen [aut], Arthur Leroy [aut], Mahendra Mariadassou [aut], Cédric Midoux [aut], Jean-François Rey [aut], Aymeric Stamm [aut, cre]
Maintainer:	Aymeric Stamm <aymeric.stamm@cnrs.fr>
Repository:	CRAN
Date/Publication:	2026-09-12 14:20:02 UTC

rdatagouv: Tools to download and explore datasets from the French open data platform

Description

This package provides a small client for the public API of data.gouv.fr, the French government's open data platform. It lets you list the published datasets, download their tabular resources and compute summary metrics (file weight, number of columns, missing-value rates, ...). Requests are built on top of the httr2 package.

Author(s)

Maintainer: Aymeric Stamm aymeric.stamm@cnrs.fr (ORCID)

Authors:

• Aymeric Stamm aymeric.stamm@cnrs.fr (ORCID)

• Magali Berland magali.berland@inrae.fr

• Pierre Gloaguen pierre.gloaguen@univ-ubs.fr

• Arthur Leroy arthur.leroy@inrae.fr

• Mahendra Mariadassou mahendra.mariadassou@inrae.fr

• Cédric Midoux cedric.midoux@inrae.fr

• Jean-François Rey jean-francois.rey@inrae.fr

See Also

Useful links:

• https://astamm.github.io/rdatagouv/

• https://github.com/astamm/rdatagouv

• Report bugs at https://github.com/astamm/rdatagouv/issues

Find datasets available on data.gouv.fr

Description

Collects the datasets published on the data.gouv.fr platform, searching the catalog via the v2 datasets/search endpoint (the same one the web interface uses). By default it returns the first n datasets; use q to search titles and descriptions server-side instead of enumerating the whole catalog.

Usage

dg_find_datasets(
  q = NULL,
  n = 1000,
  format = catalog_formats(),
  schema_only = FALSE,
  organization = NULL,
  geozone = NULL,
  access_type = NULL,
  license = NULL,
  tag = NULL,
  topic = NULL,
  granularity = NULL,
  last_update = NULL,
  producer_type = NULL,
  resources = FALSE
)

Arguments

Details

Fetching every dataset on the platform means paging through many thousands of records in many HTTP requests and is both slow and fragile, so the default is deliberately bounded. Set n = Inf to return as many matches as the API allows. Note that data.gouv caps a search at 10,000 matches, so an un-narrowed n = Inf crawl stops at that cap even though the platform holds more. For large or infinite n the crawl scales its page size up (to ~250) so a full 10,000-row crawl takes ~40 requests, not hundreds.

Because v2 search embeds rich per-dataset metadata inline, the returned tibble includes columns such as license, quality_score, views, access_type, frequency, temporal_start/temporal_end, archived and featured that help judge whether a dataset is worth pulling.

v2 search does NOT inline each dataset's resources (they are subsection pointers), so the exact resource-based columns n_resources, formats, has_table and has_schema can no longer be computed without one extra request per dataset (an N+1 crawl). By default these are NA; set resources = TRUE to opt into the per-dataset resource fetch and fill them exactly.

Value

A tibble::tibble() with one row per matching dataset. The title and id columns are always non-NA; the id column holds the stable, unique dataset identifier used to address a dataset with dg_pull_dataset(). When resources = TRUE, the columns also include n_resources (number of files/resources), formats (a list-column whose elements are the distinct file formats found among them), has_table (whether at least one resource is in a format this package can parse) and has_schema (whether at least one resource carries a pointer to a declared data schema, whose per-variable documentation is exposed by dg_schema()); these are NA when resources = FALSE (the default) unless schema_only = TRUE, which forces the fetch so has_schema is filled and the filter can run.

Examples

datasets <- dg_find_datasets(n = 20)
head(datasets)

# Search server-side instead of downloading the whole catalog.
cycle <- dg_find_datasets(q = "vélo", n = 10)

# Only datasets that carry at least one parquet resource; the v2 API matches
# multiple formats as a server-side union.
compact <- dg_find_datasets(format = "parquet", n = 10)

# Only datasets with a declared schema (documented variables). `schema_only`
# forces the per-dataset resource fetch itself (~30s for n = 1000), so
# `resources = TRUE` is optional here.
documented <- dg_find_datasets(schema_only = TRUE, n = 10)

# Narrow by producer and territory. A producer may be given by its 24-hex
# id or by its exact slug/name (resolved for you), and by geozone.
fr <- dg_find_datasets(organization = "sncf",
                       geozone = "country:fr", n = 10)

# Only datasets grouped under one topic (find its id with dg_find_topics()).
mob <- dg_find_topics(q = "mobilité")
dg_find_datasets(topic = mob$id[1], n = 10)

Search for organizations (data producers) on data.gouv.fr

Description

Searches the platform's producers via the v2 organizations/search endpoint and returns a tibble with one row per matching organization, including its stable 24-hex id. That id is exactly what you pass to the organization argument of dg_find_datasets() to restrict a catalog search to one producer — or, more conveniently, you can pass a producer's exact name or slug to dg_find_datasets() directly and it is resolved for you (see dg_find_datasets()).

Usage

dg_find_organization(q = NULL, n = 20)

Arguments

Details

Useful when you want to discover which producers exist, how a producer's name is spelled, or how many datasets it publishes — before narrowing a search.

Value

A tibble::tibble() with one row per matching organization and columns:

• id — the stable, unique 24-hex producer id (passable to the organization argument of dg_find_datasets()). Always non-NA.

• name — the producer's display name.

• slug — the URL-friendly slug.

• acronym — the acronym, or NA.

• description — the producer's description.

• datasets — number of datasets the producer currently publishes.

• badges — comma-joined badge kinds (e.g. "public-service, certified"), or NA.

• business_number_id — the French SIREN identifier when known, else NA.

Examples

# Who publishes rail/mobility data? (server-side ranked search)
orgs <- dg_find_organization(q = "SNCF")
orgs[, c("id", "name", "datasets")]

# Use the resolved id to narrow a catalog search to one producer.
data_gouv <- dg_find_organization(q = "data.gouv")
open_data <- dg_find_datasets(organization = data_gouv$id[1], n = 5)

Find topics (themes) on data.gouv.fr

Description

Searches the platform's themes via the v2 topics/search endpoint and returns a tibble with one row per matching topic, including its stable 24-hex id. That id is what you pass to the topic argument of dg_find_datasets() to restrict a catalog search to one theme. (data.gouv does not resolve topic names/slugs to ids inside dg_find_datasets(), so this finder is the way to discover a theme and get its id.)

Usage

dg_find_topics(q = NULL, n = 20, elements = FALSE)

Arguments

Details

Useful when you want to discover which curated themes exist and how many elements (datasets, reuses, dataservices...) they group, before narrowing a search — e.g. browse "Mobilité", "Environnement", "Énergie".

Value

A tibble::tibble() with one row per matching topic and columns:

• id — the stable, unique 24-hex topic id (passable to the topic argument of dg_find_datasets()). Always non-NA.

• name — the topic's display name.

• slug — the URL-friendly slug.

• description — the topic's description.

• tags — comma-joined tags, or NA.

• featured — whether the platform features this topic.

• n_elements — number of elements (datasets, reuses, dataservices, ...) grouped under the topic, from the API.

• n_datasets, n_dataservices, n_reuses — per-kind counts, only when elements = TRUE (else NA).

Examples

# Browse the curated themes.
dg_find_topics(n = 5)[, c("id", "name", "n_elements")]

# Narrow a catalog search to one theme once you have its id.
mob <- dg_find_topics(q = "mobilité")
dg_find_datasets(topic = mob$id[1], n = 10)

Glimpse the metadata of a dataset on data.gouv.fr

Description

Surfaces the dataset-level health and engagement metadata that the v2 API embeds inline but the v1 fetch path does not expose — the bridge between discover (dg_find_datasets()) and judge (dg_schema()'s column documentation). It reports the dataset's quality score and flags, its metrics (views, downloads, followers, ...) and its context (organization, license, frequency, temporal/spatial coverage, access type), helping a user decide whether a dataset is worth pulling.

Usage

dg_glimpse(id, table = NULL)

Arguments

Details

id composes naturally with the rest of the package: it may be a dataset id (24-hex), a composed table id or a pulled table (whose id attribute is read), so a dataset discovered or pulled elsewhere can be glimpsed directly.

Value

A list with the v2-inline metadata:

• quality: a list with score (0-1) and boolean flags (license, temporal_coverage, spatial, update_frequency, dataset_description_quality).

• metrics: views, resources_downloads, followers, discussions, reuses, dataservices.

• context: organization (name/slug/id), license, frequency, temporal_coverage, spatial/granularity, access_type, archived, featured.

• resources (only when table = TRUE): the list of resource objects.

Examples

g <- dg_glimpse("6a6be5976a05df136d48fb7a")
g$quality
g$metrics

Return the parsing problems of a downloaded table

Description

Returns the data frame of parsing issues that vroom encountered while reading a table downloaded with dg_pull_dataset() or dg_refetch(). These are vroom's "parsing issues" (rows that could not be converted to the inferred column type, e.g. a mostly-padded ISO date column holding a few non-padded values like 2021-7-01).

Usage

dg_problems(x)

Arguments

Details

The noisy per-cell warnings themselves are suppressed by default during a pull; use this accessor to inspect what happened. The problems live in the rdatagouv_problems attribute of the table. Returns NULL when the table carries no recorded problems (or when x is an ordinary data frame).

Value

A data frame with columns row, col, expected and actual (one row per parsing issue), or NULL if there were none.

Examples

dg_problems(iris)

Download a dataset from data.gouv.fr

Description

Downloads the first tabular resource of a dataset and parses it into a tibble::tibble() with format_tibble(). The dataset is identified by its id, which is the stable, unique identifier returned in the id column of dg_find_datasets(). For backwards compatibility, an exact title is also accepted and is resolved by searching the platform.

Usage

dg_pull_dataset(
  id,
  all_files = FALSE,
  remove_na = FALSE,
  col_types = NULL,
  use_tabular_types = TRUE
)

Arguments

Details

By default a single tibble is returned: the first resource that can actually be parsed as a table (for a multi-file ZIP, the first parseable file). The table's stable, unique address is attached as an id attribute, readable with dg_table_id() and accepted directly by dg_refetch() and dg_schema(). Set all_files = TRUE to instead receive one table per parseable file as a named list (useful for a ZIP holding several files).

Value

A tibble::tibble() (default) or, when all_files = TRUE and the resource is a multi-file ZIP, a named list of tibbles (one element per parseable file, named after it). Every table carries its stable, unique address as an id attribute — a URI of the form ⁠https://www.data.gouv.fr/datasets/<dataset_id>#<resource_id>⁠ (plus ⁠/<file>⁠ for a file inside a ZIP) — re-fetchable with dg_refetch() and readable with dg_table_id(). Any parsing issues vroom encountered are attached as an rdatagouv_problems attribute (a data frame), readable with dg_problems().

Examples

id <- "6397c0ff56d3963118a18345"
tbl <- dg_pull_dataset(id)
head(tbl)
dg_table_id(tbl)

Re-fetch a single parsed table by its stable address

Description

Downloads again the exact table addressed by a table URI, stored as an id attribute on the tables returned by dg_pull_dataset() and readable with dg_table_id(). The URI is built from the platform's own stable identifiers (dataset id + resource id, plus the file name inside a ZIP) and opens the dataset page in a browser, so this reproducibly returns the same table, independent of the human-readable list keys.

Usage

dg_refetch(x, remove_na = FALSE, col_types = NULL, use_tabular_types = TRUE)

Arguments

Value

A tibble::tibble() — the single re-fetched table (the id addresses one table, not a multi-file ZIP as a whole). The table's id is attached as an id attribute; parsing issues are attached as an rdatagouv_problems attribute, readable with dg_problems().

Examples

tbl <- dg_pull_dataset("6397c0ff56d3963118a18345")
again <- dg_refetch(tbl)

Documented schema of a parsed table's columns

Description

Returns the declared data schema of a table's columns: the per-variable fields recorded by the dataset producer. Because data.gouv attaches a schema only as a pointer (schema$name / schema$url), this resolves that pointer against schema.data.gouv.fr and returns the human-readable column documentation (name, title, description, type) that the schema carries — the information needed to judge whether a variable really means what a statistical exploration assumes.

Usage

dg_schema(x)

Arguments

Details

This is a supplement to dg_pull_dataset(): the table itself comes from the main API; the schema is read from the producer's declared data specification. Only resources that carry a schema pointer have documentation; resources without one return NULL with a message. Use dg_find_datasets() (column has_schema, or the schema_only argument) to target schema-documented tables in the first place.

Value

A tibble::tibble() with one row per column and the columns name, title, description, type and example (where the schema provides them; absent entries are NA), or NULL (with a message) if the resource has no declared schema. The schema's own title and name are attached as the attributes schema_title and schema_name.

Examples

tbl <- dg_pull_dataset("62c5961ff0013fb71d7278e3")
dg_schema(tbl)

Summarise several datasets

Description

Applies dg_summary() to a collection of tables and combines the resulting metrics into a single tibble. If datasets is NULL, the first n datasets returned by dg_find_datasets() are downloaded and summarised.

Usage

dg_summarise(datasets = NULL, n = 100)

Arguments

Value

A tibble::tibble() with one row per table and the columns described in dg_summary().

Examples

# Summarise in-memory tables (no network needed).
dg_summarise(datasets = list(iris = iris, mtcars = mtcars))


# Download and summarise the first datasets of the catalog.
dg_summarise()

Compute summary metrics for a dataset

Description

Computes key metrics describing a parsed dataset: its in-memory weight in kilobytes, the number of variables, the number of numeric and non-numeric variables, the number of rows and the proportion of missing values.

Usage

dg_summary(x, name = NULL)

Arguments

Value

A tibble::tibble() with a single row and the following columns: dataset, size_kb, n_vars, n_numeric, n_non_numeric, n_rows and prop_missing.

Examples

dg_summary(iris, name = "iris")

Read a table's stable address

Description

Returns the stable, unique address of a table downloaded with dg_pull_dataset() or dg_refetch(), which is stored as an id attribute on the table. The address is a URI of the form ⁠https://www.data.gouv.fr/datasets/<dataset_id>#<resource_id>⁠ (plus ⁠/<file>⁠ for a file inside a ZIP) and uniquely identifies a table on the platform, independent of the human-readable catalog titles. It can be passed directly to dg_refetch() or dg_schema() to re-fetch or document that exact table.

Usage

dg_table_id(x)

Arguments

Value

The composed table id, a string, or NULL if x carries no id attribute (e.g. an ordinary data frame).

Examples

tbl <- dg_table_id(iris)