Write a PhysioExperiment to a Parquet dataset directory

Serialises a PhysioExperiment (or MultiRatePhysioExperiment) to a directory of Apache Parquet tables and JSON sidecars. Each assay is stored as a columnar Parquet table (one column per channel), and colData, rowData, events, metadata, provenance and per-assay sampling rates are stored alongside so that readParquet() reproduces the object exactly.

Usage

writeParquet(x, dir, overwrite = FALSE)

Arguments

x: A PhysioExperiment or MultiRatePhysioExperiment.
dir: Output directory (created if absent).
overwrite: Overwrite a non-empty directory (default FALSE).

Value

The output dir, invisibly.

Details

Parquet is a columnar, self-describing format that DuckDB can query in place via read_parquet(), enabling out-of-core aggregation over the assays without loading them into R (see registerParquetAssay()).

A MultiRatePhysioExperiment is written as a streams/ sub-directory (one Parquet dataset per stream) plus its common clock (t0, reference rate, per-stream offsets).

References

Apache Software Foundation. "Apache Parquet." https://parquet.apache.org/.

Examples

if (requireNamespace("arrow", quietly = TRUE)) {
  pe <- PhysioExperiment(
    assays = S4Vectors::SimpleList(raw = matrix(rnorm(30), 10, 3)),
    colData = S4Vectors::DataFrame(label = c("Fp1", "Fp2", "Cz")),
    samplingRate = 256)
  d <- tempfile("pe-parquet")
  writeParquet(pe, d)
  pe2 <- readParquet(d)
}