Skip to contents

Run the full soilKey benchmark suite and (optionally) write a report

Source: R/benchmark-suite.R
run_all_benchmarks.Rd

Auto-detects which reference datasets are available locally, runs each via benchmark_unified, adds the offline canonical sanity row and the AfSP sample when present, and returns a tidy accuracy summary. When report_path is given, a consolidated Markdown report is written.

Usage

run_all_benchmarks(
  datasets = "auto",
  paths = NULL,
  max_n = 300L,
  level = "order",
  report_path = NULL,
  verbose = TRUE
)

Arguments

datasets

"auto" (default) detects available datasets; otherwise any subset of c("bdsolos", "febr", "kssl", "lucas_esdb", "redape"), the literal "canonical" (only the fixture sanity row), or "all" (every dataset regardless of availability – absent ones are skipped).

paths

Named list of dataset paths (see benchmark_unified). NULL uses the package defaults (override the root via options(soilKey.benchmark_root = "...")).

max_n

Cap on pedons per dataset (keeps the run fast). Default 300.

level

Comparison level forwarded where supported (currently the suite reports at "order" / top level).

report_path

File to write the Markdown report to, TRUE to auto-name one under inst/benchmarks/reports/, or NULL (default) for no file.

verbose

Print progress.

Value

Invisibly, a list with summary (data.frame: dataset, system, n_compared, accuracy), per_system (pooled), raw (full benchmark_unified output), weak (zero-recall classes) and config.

See also

benchmark_unified, benchmark_redape.

Examples

if (FALSE) { # \dontrun{
res <- run_all_benchmarks(max_n = 250,
                          report_path = TRUE)
res$summary
} # }