Two versions of {duckspatial} have shipped since the last update: v1.1.2 and v1.2.0. Both are now on CRAN, so let’s go through the highlights together.
## Packages setup
# pak::pak("duckspatial") # upgrade the package to the latest version
library(dplyr)
library(duckspatial)
library(ggplot2)
library(mapgl)
library(sf)
## Load some data
argentina_ddbs <- ddbs_open_dataset(
system.file("spatial/argentina.geojson", package = "duckspatial")
)A batch of new unary functions covers common needs that weren’t in the package yet.
This function flips the vertex order of a geometry.
ddbs_reverse(argentina_ddbs)# A duckspatial lazy spatial table
# ● CRS: EPSG:4326
# ● Geometry column: geom
# ● Geometry type: POLYGON
# ● Bounding box: xmin: -73.525 ymin: -52.398 xmax: -53.624 ymax: -21.818
# Data backed by DuckDB (dbplyr lazy evaluation)
# Use ddbs_collect() or st_as_sf() to materialize to sf
#
# Source: table<temp_table_eda21e35_a697_4ac7_aff0_b34404274264> [?? x 8]
# Database: DuckDB 1.5.4 [cidre@Linux 7.0.0-22-generic:R 4.6.0/:memory:]
OGC_FID CNTR_ID NAME_ENGL ISO3_CODE CNTR_NAME FID date geom
<dbl> <chr> <chr> <chr> <chr> <chr> <date> <wk_wkb>
1 0 AR Argentina ARG Argentina AR 2021-01-01 <POLYGON ((-62…It returns the canonical form of the geometry, sorting the geometry in a deterministic way. This is useful before comparing geometries for equality.
ddbs_normalize(argentina_ddbs)# A duckspatial lazy spatial table
# ● CRS: EPSG:4326
# ● Geometry column: geom
# ● Geometry type: POLYGON
# ● Bounding box: xmin: -73.525 ymin: -52.398 xmax: -53.624 ymax: -21.818
# Data backed by DuckDB (dbplyr lazy evaluation)
# Use ddbs_collect() or st_as_sf() to materialize to sf
#
# Source: table<temp_table_6c5d7a2b_7430_4d40_8114_c5de5ac10f25> [?? x 8]
# Database: DuckDB 1.5.4 [cidre@Linux 7.0.0-22-generic:R 4.6.0/:memory:]
OGC_FID CNTR_ID NAME_ENGL ISO3_CODE CNTR_NAME FID date geom
<dbl> <chr> <chr> <chr> <chr> <chr> <date> <wk_wkb>
1 0 AR Argentina ARG Argentina AR 2021-01-01 <POLYGON ((-73…Collects every vertex of a geometry into a MULTIPOINT.
argentina_vertices_sf <- ddbs_vertices(argentina_ddbs, mode = "sf")
maplibre_view(argentina_vertices_sf)Snaps coordinates to a grid to shrink geometry size. The next example reduces the precision of argentina boundary to 1 degree (e.g. 62.4328 degrees would shrink to 62 degrees).
argentina_precision_sf <- ddbs_reduce_precision(argentina_ddbs, precision = 1)
maplibre_view(ddbs_collect(argentina_precision_sf))A full set of extent helpers landed too: ddbs_xmax(), ddbs_xmin(), ddbs_ymax(), ddbs_ymin(), plus the Z/M equivalents. By default they work by_feature = TRUE adding a new column per feature (useful for maintaining workflows within duckdb, and apply filters):
# A duckspatial lazy spatial table
# ● CRS: EPSG:4326
# ● Geometry column: geom
# ● Geometry type: POLYGON
# ● Bounding box: xmin: -73.525 ymin: -52.398 xmax: -53.624 ymax: -21.818
# Data backed by DuckDB (dbplyr lazy evaluation)
# Use ddbs_collect() or st_as_sf() to materialize to sf
#
# Source: table<temp_table_3c4a40a8_b6f4_48d1_9096_fc118e38818f> [?? x 4]
# Database: DuckDB 1.5.4 [cidre@Linux 7.0.0-22-generic:R 4.6.0/:memory:]
NAME_ENGL xmin xmax geom
<chr> <dbl> <dbl> <wk_wkb>
1 Argentina -73.5 -53.6 <POLYGON ((-62.6452 -22.25047, -62.25779 -22.53346, -61…However, we can get the global minimum or maximum:
ddbs_xmax(argentina_ddbs, by_feature = FALSE)[1] -53.62409And ddbs_dimension() returns the topological dimension of each geometry (0 = point, 1 = line, 2 = polygon):
ddbs_dimension(argentina_ddbs)# A duckspatial lazy spatial table
# ● CRS: EPSG:4326
# ● Geometry column: geom
# ● Geometry type: POLYGON
# ● Bounding box: xmin: -73.525 ymin: -52.398 xmax: -53.624 ymax: -21.818
# Data backed by DuckDB (dbplyr lazy evaluation)
# Use ddbs_collect() or st_as_sf() to materialize to sf
#
# Source: table<temp_table_aa7a5276_5b15_4455_80e2_a86c17310ae3> [?? x 9]
# Database: DuckDB 1.5.4 [cidre@Linux 7.0.0-22-generic:R 4.6.0/:memory:]
OGC_FID CNTR_ID NAME_ENGL ISO3_CODE CNTR_NAME FID date dimension geom
<dbl> <chr> <chr> <chr> <chr> <chr> <date> <int> <wk_>
1 0 AR Argentina ARG Argentina AR 2021-01-01 2 <POL…Rounding out the polygon-side additions, ddbs_get_ninterior_rings() counts holes in a POLYGON.
ddbs_point() creates POINT geometries directly from coordinate vectors, with 2D/3D/4D support and CRS assignment:
cities_ddbs <- ddbs_point(
x = c(-58.3816, -64.1811),
y = c(-34.6037, -31.4201),
another_column = c("A", "B"),
crs = 4326
)
cities_ddbs# A duckspatial lazy spatial table
# ● CRS: EPSG:4326
# ● Geometry column: geometry
# ● Geometry type: POINT
# ● Bounding box: xmin: -64.181 ymin: -34.604 xmax: -58.382 ymax: -31.42
# Data backed by DuckDB (dbplyr lazy evaluation)
# Use ddbs_collect() or st_as_sf() to materialize to sf
#
# Source: table<temp_table_a7614852_e444_4ae3_adf4_77eb75c593ae> [?? x 2]
# Database: DuckDB 1.5.4 [cidre@Linux 7.0.0-22-generic:R 4.6.0/:memory:]
another_column geometry
<chr> <wk_wkb>
1 A <POINT (-58.3816 -34.6037)>
2 B <POINT (-64.1811 -31.4201)>For line work, ddbs_shortest_line() returns the connecting LINESTRING between the closest points of two geometries, and ddbs_line_locate_point() gives the fractional position (0–1) of the closest point on a line to a reference point. ddbs_line_node() is the batch version: it nodes a set of lines, splitting them at every crossing into a fully noded MULTILINESTRING.
ddbs_azimuth() computes the clockwise bearing between two sets of points, in radians by default or degrees via unit = "degrees".
On the aggregation side, ddbs_intersection_agg() joins ddbs_union_agg() as the intersection counterpart — computes the common area of a set of geometries, optionally grouped by column.
## create three overlapping polygons
polys <- sf::st_as_sf(
data.frame(grp = c("a", "a", "b", "b")),
geometry = sf::st_sfc(
sf::st_polygon(list(matrix(c(0,0, 3,0, 3,3, 0,3, 0,0), ncol = 2, byrow = TRUE))),
sf::st_polygon(list(matrix(c(1,1, 4,1, 4,4, 1,4, 1,1), ncol = 2, byrow = TRUE))),
sf::st_polygon(list(matrix(c(2,2, 5,2, 5,5, 2,5, 2,2), ncol = 2, byrow = TRUE))),
sf::st_polygon(list(matrix(c(3,3, 6,3, 6,6, 3,6, 3,3), ncol = 2, byrow = TRUE)))
),
crs = 4326
)
## intersect all geometries into their common area
polys_inters <- ddbs_intersection_agg(polys, mode = "sf")
## intersect within groups
polys_inters_grp <- ddbs_intersection_agg(polys, by = "grp", mode = "sf")
## plot them
ggplot() +
geom_sf(data = polys, aes(fill = grp), alpha = 0.3) +
geom_sf(data = polys_inters_grp, aes(fill = grp), alpha = 0.95) +
geom_sf(data = polys_inters, size = 4) +
geom_text(aes(x = 3, y = 3, label = "(3,3)"), nudge_x = 0.2, nudge_y = 0.1) +
scale_x_continuous(breaks = 0:6) +
scale_y_continuous(breaks = 0:6) +
theme_minimal() +
labs(fill = "grp")
Note that we have generated:
polys_inters (the black point): intersects all four polygons together (ignoring grp). Since each square is shifted diagonally by 1 unit and only 3×3 in size, the only area common to all four is that single point at (3,3).
polys_inters_grp: the intersection computed within each group (by = "grp"). The two transparent red squares (grp = "a") overlap in the red square; the two transparent teal squares (grp = "b") overlap in the teal square.
ddbs_as_geojson(), ddbs_as_wkb(), and friends already let you serialize geometries out. Now there’s a matching set of parsers to bring them back in: ddbs_geom_from_text(), ddbs_geom_from_wkb(), ddbs_geom_from_hexwkb(), ddbs_geom_from_hexewkb(), and ddbs_geom_from_geojson().
ddbs_as_geojson() itself got more useful: it now carries all non-geometry columns as feature properties instead of dropping them, and returns a single FeatureCollection by default — matching geojsonsf::sf_geojson(). Pass feature_collection = FALSE if you want one Feature per row instead.
Two new functions bring Mapbox Vector Tile support into {duckspatial}: ddbs_as_mvt_geom() transforms geometries into MVT coordinate space, clipping to a tile’s bounding box. ddbs_write_mbtiles() goes further and generates a full tile pyramid from a spatial dataset, writing it straight to an MBTiles file — ready to serve or convert to PMTiles.
ddbs_extension_info() prints a glimpse() of a DuckDB extension’s row from duckdb_extensions() — installed/loaded status, version, install path — for the spatial extension by default:
conn <- ddbs_create_conn()
ddbs_extension_info(conn)Rows: 1
Columns: 9
$ extension_name <chr> "spatial"
$ loaded <lgl> TRUE
$ installed <lgl> TRUE
$ install_path <chr> "/home/cidre/R/x86_64-pc-linux-gnu-library/4.6/duckd…
$ description <chr> "Geospatial extension that adds support for working …
$ aliases <list> <>
$ extension_version <chr> "28db190"
$ install_mode <chr> "REPOSITORY"
$ installed_from <chr> "core"This is not restricted to spatial, you can use it with any other extensio:
ddbs_extension_info(conn, "h3")Rows: 1
Columns: 9
$ extension_name <chr> "h3"
$ loaded <lgl> FALSE
$ installed <lgl> TRUE
$ install_path <chr> "/home/cidre/R/x86_64-pc-linux-gnu-library/4.6/duckd…
$ description <chr> ""
$ aliases <list> <>
$ extension_version <chr> "be9ee91"
$ install_mode <chr> "REPOSITORY"
$ installed_from <chr> "community"ddbs_install() gains a repos argument to pull an extension from a specific DuckDB repository ("core", "core_nightly", "community"). Leaving it NULL keeps the previous fallback behavior (core, then community).
ddbs_install() had a broken “already on the latest version” check — it referenced a column DuckDB doesn’t provide and compared install_mode with the wrong case, so it silently never triggered. Removed.ddbs_union_agg() gains a mem argument: set mem = TRUE to use ST_MemUnion_Agg() instead of ST_Union_Agg() — slower, but lighter on memory for large unions.That’s about 20 new functions across the two releases, plus some quality-of-life fixes for extension installation and GeoJSON export. As always, full changelogs are on GitHub, and the package is on CRAN.
Feedback and issues are very welcome — open one here if you run into anything.
Analyses were conducted using the R Statistical language (version 4.6.0; R Core Team, 2026) on Ubuntu 26.04 LTS, using the packages duckspatial (version 1.2.0; Cidre González A et al., 2026), sf (version 1.1.1; Pebesma E, Bivand R, 2023), mapgl (version 0.5.0; Walker K, 2026), ggplot2 (version 4.0.3; Wickham H, 2016) and dplyr (version 1.2.1.9000; Wickham H et al., 2026).