Test data

The files under testdata/ are small Parquet files produced by PyArrow 25. The unit tests decode them and, for a subset, encode them again and decode the result.

File	What it checks
opt_i32.parquet	Optional int32, one null, plain encoding, no compression.
required_i64.parquet	A non-null int64 field.
strings.parquet	Optional UTF-8 byte arrays.
dict_str.parquet	Dictionary indexes and a null.
bools.parquet	Optional booleans.
floats.parquet	Float64, negative zero, and NaN.
list_i32.parquet	A list column, including null and empty lists.
struct.parquet	A struct with an int and a string.
map.parquet	A map from string to int32.
decimal.parquet	Decimal128 stored as a fixed-length byte array.
ts.parquet	Timestamp in microseconds, adjusted to UTC.
date.parquet	Date stored as days.
v2.parquet	Data page v2 and a page index.
multi_rg.parquet	Several row groups.
bss.parquet	Byte stream split on float32.
snappy_i32.parquet, gzip_i32.parquet, zstd_i32.parquet, lz4_i32.parquet	The four compression codecs.

testdata/schema/record.json is the JSON Schema input for gld-parquetgen-mojo. The checked-in generator output is tests/generated/.

PyArrow is a test oracle. It is not linked into the library and it is not required to encode or decode.