From 09b78b5640fb2d9caf61e1cde110231655fd5ec2 Mon Sep 17 00:00:00 2001 From: Martin Durant Date: Mon, 25 Sep 2023 11:11:33 -0400 Subject: [PATCH] Allow DELTA for V1 pages (#883) --- fastparquet/core.py | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/fastparquet/core.py b/fastparquet/core.py index 5a05554b..eab78abb 100644 --- a/fastparquet/core.py +++ b/fastparquet/core.py @@ -159,6 +159,10 @@ def read_data_page(f, helper, header, metadata, skip_nulls=False, values = values.data[:nval] else: values = np.zeros(nval, dtype=np.int8) + elif daph.encoding == parquet_thrift.Encoding.DELTA_BINARY_PACKED: + values = np.empty(daph.num_values - num_nulls, dtype=np.int32) + o = encoding.NumpyIO(values.view('uint8')) + encoding.delta_binary_unpack(io_obj, o) else: raise NotImplementedError('Encoding %s' % daph.encoding) return definition_levels, repetition_levels, values[:nval]