Skip to contents

Changelog

Source: NEWS.md

ONgeoR 0.4.0

  • Add retrieve_onmarg(), add_onmarg(), and onmarg_geographies() for the 2021 Ontario Marginalization Index (ON-Marg). add_onmarg() attaches the four dimension scores, and the quintiles where they are published, to an administrative boundary layer by that layer’s own key column; eight of the ten published geographies map onto a registered ONgeoR source, including Public Health Units on the pre-2025 34-unit geography. The workbook is fetched from Public Health Ontario at runtime, verified against a pinned SHA-256, and held in memory for the session only: its licence permits non-commercial use with attribution and forbids modifying the content, so ONgeoR neither bundles it, writes it to the retrieval cache, nor alters its values. readxl is a new optional (Suggests) dependency, needed only by these functions.

  • Add retrieve_census() for the 14 registered 2021 StatCan census boundary layers. Census queries now use the StatCan endpoint directly, filter to Ontario (PRUID = '35') on the server, return EPSG:4326 geometry, and accept an optional EPSG:4326 bounding box. Remove the impractically large dissemination-block registry entry.

  • Add retrieve_phu_pre2025() for the retained 34-boundary, 250 m simplified PHU snapshot that LIO no longer serves.

  • Export layer_id_col(), which resolves a retrieved layer’s id column from the source registry’s declared key fields. This is the supported way to join build_crosswalk() / build_intersection() output back onto the layer geometry it came from; downstream callers previously had to reach for the internal.

  • Add resolve_postal_points() for checksum-verified OPCC M1 postal-code to latitude/longitude resolution (282,409 address-derived nar_centroid matches, 17,373 coarser geonames matches, 14 codes with no coordinates), with optional as_sf = TRUE output as an EPSG:4326 POINT layer.

  • Add resolve_postal() for checksum-verified OPCC M5 postal-code to dissemination-area resolution, plus render_postal_reproducer_script(). normalize_postal_code() is exported; joins use a normalized, de-duplicated postal-code key, and render_postal_reproducer_script() gains all_links.

  • Add retrieve_monitoring_stations() for Ontario water and weather monitoring stations (LIO_Open08/30), with live retrieval automatically paginated across pages of 2,000 features, and a bundled offline subset retrieve_monitoring_stations_simple() (2,407 stations, a frozen 2023-06-23 GeoHub snapshot). The bundled layer is the first bundled point layer, so point-in-polygon examples now run offline.

  • Fix missing provenance on retrieve_monitoring_stations_simple(). It was the only bundled retriever returning no source_name / source_url / retrieved_at attributes, because its .rds was written by a plain saveRDS() while the others carry the attributes in the file. Every crosswalk built against the layer therefore reported to_source, source_url_to and retrieved_at as NA. The attributes are now attached at read time from the source registry, with retrieved_at set to the snapshot instant (2023-06-23T10:55:20Z) rather than the time of the call.

  • Fix the bundled HIVE grid: eight cells were invalid under planar GEOS while passing st_is_valid() on geographic coordinates (s2), so build_intersection() on HIVE Level 1/2 grids aborted with a TopologyException. The repaired cells now validate under GEOS and s2; all other cells are unchanged.

  • Point-in-polygon examples for link(), nearest(), map_nearest(), build_nearest_layers(), and build_nearest_pairs() now use the bundled station layer and execute during R CMD check.

  • The Shiny app moves to its own package, ONgeoRapp. run_app() is no longer exported from ONgeoR; install ONgeoRapp and call ONgeoRapp::run_app(). Nothing else changed — the app only ever used exported functions, so no linking, retrieval, or crosswalk behaviour is affected.

  • Suggests drops shiny, bslib, DT, shinytest2, chromote, future, later, and promises, which moved with the app. This also removes the standing “detritus in the temp directory” R CMD check NOTE, which came from the browser smoke test.

ONgeoR 0.3.0.9000

  • Add registry-driven LIO retrieval and 17 verified Ontario geography sources.
  • Fix generalized geometry silently dropping small polygons. maxAllowableOffset was 10, but the service returns EPSG:4326, making that a ~1,100 km tolerance: Toronto, Peel, Halton, Hamilton, Windsor-Essex, Niagara, York, Ottawa and Middlesex-London all came back with zero area, and PHU boundaries totalled 646,394 km2 instead of 983,322 km2. Now 1e-04 degrees (~11 m).
  • Repair geometry validity without deleting features. sf::st_make_valid() dispatches to s2 on geographic coordinates, which answered the degenerate rings by returning EMPTY, so 11 of 29 PHU boundaries arrived as empty GEOMETRYCOLLECTIONs and maps drew 18 units with no error. Retrieval now repairs on the planar GEOS path, rebuilds under s2 (a ring GEOS accepts can still abort st_intersects), keeps only polygonal parts of areal input, and errors rather than returning a layer with missing shapes.
  • Add a geometry-schema component to the cache key, so caches written before the above are not served afterwards.
  • retrieve_phu() now defaults to simplify = TRUE; full-resolution requests for that layer fail against the service.
  • Shiny app: rename the pickers to Source layer / Target layer, rename Link to Join, and move the pairing explanation from Preview to a Join confirmation that names both layers with their dimensions and the expected result shape. Join stays disabled until a preview of the current pair succeeds, and is now visibly greyed out in that state. The results download is mapping.csv.
  • Fix reproduce.R so it reproduces what the app produces: it used the CLI’s multi-pair path, which re-downloaded every layer and wrote 16 columns where the app writes 14.
  • Add build_intersection() for polygon-to-polygon linking: returns every overlapping pair with share_of_target (fraction of the target covered by the source) and share_of_source (fraction of the source falling inside the target, the apportionment weight for extensive variables). All source and target attributes are carried through with src_ and tgt_ prefixes.
  • Add build_nearest_pairs() for point-to-point linking: each target point is matched to its single nearest source point, with the same column schema as build_intersection() so the two are row-bindable.
  • Add summarise_by_target() to collapse a pairs table to one row per distinct target, with multi-valued fields as "; "-delimited strings and the dominant source identified by share_of_target.
  • Add build_link() as the no-choice entry point: it inspects the geometry types of the two layers and dispatches to build_nearest_pairs() for point-point, build_intersection() for polygon-polygon, or link() / build_crosswalk() for mixed types.
  • Duplicate id values in either layer now abort early in build_intersection() and build_nearest_pairs(), because summarise_by_target() keys on target_id and two distinct features sharing an id would silently collapse.
  • All changes are additive: build_crosswalk() and its five methods (within, intersects, point_on_surface, largest_overlap, weighted) are unchanged and remain available for direct callers.