Create buffered tiles from point layers

Buffers point parquet files and writes per-tile outputs named:

Fixed radius: <base>_r<radius>_<tileid>.parquet
Per-feature radii (radius_field): <base>_varradius_<tileid>.parquet

Where:

base = prefix of input filename before the first underscore (e.g., "pts100" from pts100_sauzeme.parquet)
tileid = unique value of split_field (default "tks50km") inside the input file

The buffer_mode determines how radii are assigned:

"dense": Buffers the best-matching pts100*.parquet (prefers pts100_sauzeme.parquet) for each tile by radii_dense (default: 500, 1250, 3000, 10000 m).
"sparse": Uses a file to radius mapping. Default mapping:
- "pts100_sauzeme.parquet" to c(500, 1250)
- "pts300_sauzeme.parquet" to 3000
- "pts1000_sauzeme.parquet" to 10000 You can override this via mapping_sparse (named list or data.frame with columns file, radius).
"specified": You provide points_path and either buffer_radius (uniform, one or more fixed radii) or radius_field (numeric column with per-feature radii in meters). Outputs are still split by split_field.

Usage

tiled_buffers(
  in_dir = "./Templates/TemplateGridPoints",
  out_dir = "./Templates/TemplateGridPoints/lapas",
  buffer_mode = "dense",
  radii_dense = c(500, 1250, 3000, 10000),
  mapping_sparse = list(pts100_sauzeme.parquet = c(500, 1250), pts300_sauzeme.parquet =
    3000, pts1000_sauzeme.parquet = 10000),
  points_path = NULL,
  buffer_radius = NULL,
  radius_field = NULL,
  split_field = "tks50km",
  n_workers = max(1L, parallel::detectCores()),
  future_max_mem_gb = 4,
  overwrite = FALSE,
  quiet = FALSE
)

Source

Zenodo grids/points example: https://zenodo.org/records/14277114

Arguments

in_dir: Character. Directory containing input points parquet files (default "./Templates/TemplateGridPoints"). Used in "dense" and "sparse" modes.
out_dir: Character. Output directory for buffered tiles (default "./Templates/TemplateGridPoints/lapas").
buffer_mode: Character. One of "dense", "sparse", "specified". Default "dense".
radii_dense: Numeric vector of radii (m) used when buffer_mode = "dense". Default c(500, 1250, 3000, 10000).
mapping_sparse: Named list or data.frame describing file to radii for buffer_mode = "sparse". Default: list("pts100_sauzeme.parquet" = c(500, 1250), "pts300_sauzeme.parquet" = 3000, "pts1000_sauzeme.parquet" = 10000). If a data.frame is supplied, it must have columns file and radius.
points_path: Character. Path to a single parquet file for buffer_mode = "specified".
buffer_radius: Numeric vector. Used in "specified" when you want fixed radii. Ignored if radius_field is provided.
radius_field: Character or NULL. Column name in points_path that provides per-feature radii (meters) for "specified". If given, buffer_radius is ignored.
split_field: Character. Field in the point data that defines tiles. Default "tks50km".
n_workers: Integer. Parallel workers. Default max(1L, parallel::detectCores()).
future_max_mem_gb: Numeric. Max size of exported globals per worker (GiB). Sets options(future.globals.maxSize = future_max_mem_gb * 1024^3). Default 4.
overwrite: Logical. Overwrite existing outputs? Default FALSE.
quiet: Logical. Suppress messages? Default FALSE.

Value

Invisibly returns a data.frame with columns: input, tileid, mode, radius_m (NA if radius_field is used), radius_field, out_file, wrote.

Details

Uses sfarrow for parquet I/O and sf for buffering.
Jobs are created per tile (unique split_field value) and (when applicable) per radius.
Workers open data from file paths (keeps RAM low).
Files are written atomically (temp file then move).
Safety: jobs are de-duplicated by content and by predicted output path, so concurrent workers never write to the same file.

Examples