relialearnr vs worldbank

worldbank provides R access to the World Bank's public data: World Development Indicators, the Poverty and Inequality Platform, project records, and the Finances One datasets. It settled its type contract early — always a data.frame, never a conditional tibble — and has since layered on opt-in request caching, multi-indicator queries, and query conveniences like most-recent-values and gap filling. The most recent addition sidesteps the API entirely, pulling the full WDI archive as a zip.

Two threads run through the log. The first is query ergonomics: multiple indicators per call, mrv and gapfill parameters, regex search across the indicator catalog, a shorter wb_data() name that has since become the primary entry point. The second is coverage of things the standard API handles poorly — bulk download reaches footnote and series-time metadata the endpoints never expose, and PIP nowcasts and project records extend past the indicator tables most users start with. The maintainer runs the same infrastructure across their other data packages, and the caching design here is identical to what bbk and treasury received.

Expect the remaining rough edges of the World Bank's own API — inconsistent empty responses, metadata only available in bulk files — to keep driving releases, rather than a push into new data providers.