Write an atlas surface to Parquet (arrow path) with the standard v8 options

Parquet V2 (version = "2.6"), zstd compression, row groups of chunk_size rows (arrow has no byte-based row-group control). Use in per-model / per-species ingest loops where the data is an in-memory tibble (and often inside mclapply / furrr forks). For DuckDB-resident data use copy_atlas_parquet() instead — both share the one option set in .atlas_pq.

Usage

write_atlas_parquet(x, path, chunk_size = .atlas_pq$arrow_chunk)

Arguments

x

a data.frame / tibble (typically mdl_key, cell_id, val)

path

output .parquet path

chunk_size

rows per row group (default from .atlas_pq)

Value

path, invisibly