r-freestiler

freestiler: Vector Tile Generator

Overview

freestiler converts spatial data into PMTiles vector tilesets. Rust-powered tool for creating .pmtiles from sf objects, files, or DuckDB SQL.

References

Read references/API.md before writing code.

• references/API.md - Complete function reference with all parameters
• references/getting-started.md - Usage patterns and parameter examples
• references/zoom-strategy.md - Zoom level guidance by geometry type
• references/workflows.md - Complete workflows and integration patterns
• references/mapping.md - Visualization and server configuration

When to Use

Use when:

• Dataset >10k features (too large for direct mapgl)
• Need static hosting
• Dataset exceeds memory

Don't use:

• <10k features - pass sf to mapgl directly
• Quick exploration - use maplibre_view()
• Existing tile pipeline works

Quick Reference

Function	Use Case	Key Params
freestile()	Tile sf objects	input, output, layer_name, min_zoom, max_zoom, base_zoom (opt), drop_rate (opt), tile_format (opt)
freestile_query()	Tile from SQL (streaming)	query, output, layer_name, min_zoom, max_zoom, streaming = "always"
freestile_file()	Tile without loading	input, output, layer_name, min_zoom, max_zoom, engine
view_tiles()	Quick visualization	input, layer_type, color
serve_tiles()	Start local server	path, port

Quick Start

library(freestiler)
library(sf)

# Basic workflow - counties use max_zoom = 10
nc <- st_read(system.file("shape/nc.shp", package = "sf"))
freestile(input = nc, output = "nc.pmtiles", layer_name = "counties", max_zoom = 10)
view_tiles("nc.pmtiles")  # Auto-starts server + opens map

# Large datasets: streaming
freestile_query(query = "SELECT * FROM read_parquet('huge.parquet')",
                output = "points.pmtiles", layer_name = "locations",
                streaming = "always")  # Critical for 10M+ points

# Direct file (no loading to memory)
freestile_file(input = "data.gpkg", output = "output.pmtiles",
               layer_name = "features")

Visualization

view_tiles() (Easiest): view_tiles("nc.pmtiles")

Manual Server + mapgl:

serve_tiles("tiles/")
maplibre() |>
  add_pmtiles_source(id = "src", url = "http://localhost:8080/data.pmtiles") |>
  add_fill_layer(source = "src", source_layer = "layer1")

Direct file://: Requires absolute path in add_pmtiles_source(url = "file://W:/...")

Common Mistakes

1. Parameter names: Use input, output, layer_name (NOT data, tileset, layer)
2. Source layer must match: layer_name in freestile() must match source_layer in mapgl
3. CRS must be WGS84: Use st_transform(4326) before tiling
4. Large sf objects: Use freestile_file() or freestile_query() to avoid loading into memory
5. Massive points (10M+): Add streaming = "always" to freestile_query()
6. Tile format: MLT (default, smaller files) vs MVT (broader compatibility, use for Python)
7. Large tilesets (>500MB): serve_tiles() struggles with byte-range requests; launch npx http-server from R via processx::process$new() instead. See references/mapping.md "Serving Large Tilesets" for a ready-made helper function.

When NOT to Use

• Small datasets (<10k features) - pass sf directly to mapgl
• Quick exploration - use maplibre_view() from mapgl
• Existing tile pipeline already works

Advanced

See references/ for:

• API.md: Complete function reference
• getting-started.md, mapping.md: CRAN vignettes
• Workflows, zoom strategy, SQL patterns, performance tips

Test with validator: Use lib/r-validators/plot-validator.R to verify map output in tests

Resources: Docs | GitHub | See r-mapgl skill