Reads the subset of Parquet that rmorie's own exporters and the open-data portals it consumes actually produce: PLAIN-encoded columns (INT32, INT64, DOUBLE, BYTE_ARRAY/UTF8), uncompressed or Snappy-compressed, in any number of row groups, with all values required (no nulls) or nullability encoded via RLE definition levels of max depth 1. Falls back with a clear error naming arrow for anything richer (dictionary encoding, nested types, other codecs).
References
Apache Parquet format specification (thrift compact protocol footer; PLAIN encoding; Snappy framing).
Examples
if (requireNamespace("arrow", quietly = TRUE)) {
tf <- tempfile(fileext = ".parquet")
arrow::write_parquet(data.frame(a = 1:3), tf, use_dictionary = FALSE,
compression = "snappy")
morie_fetch_parquet(tf)
}
#> a
#> 1 1
#> 2 2
#> 3 3