Skip to content

Write parquet with decimal columns in load_test #217

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
ChrisJar wants to merge 3 commits into
base: main
Choose a base branch
from
Open

Write parquet with decimal columns in load_test #217

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
0317afb
Write parquet with decimal columns in load_test
May 21, 2021
caad91c
Read in as string
May 24, 2021
515e4dc
Cleanup
May 25, 2021
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
22 changes: 14 additions & 8 deletions gpu_bdb/queries/load_test/gpu_bdb_load_test.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,5 +1,6 @@
from bdb_tools.utils import benchmark, gpubdb_argparser, run_query
from bdb_tools.readers import build_reader
from cudf.core.dtypes import Decimal64Dtype
import os, subprocess, math, time


Expand Down Expand Up @@ -33,22 +34,21 @@ def get_schema(table):
schema = fp.read()
names = [line.replace(",", "").split()[0] for line in schema.split("\n")]
types = [
line.replace(",", "")
line.replace(", ", "")
.split()[1]
.replace("bigint", "int")
.replace("string", "str")
for line in schema.split("\n")
]
types = [
col_type.split("(")[0].replace("decimal", "float") for col_type in types
]

return names, types


def read_csv_table(table, chunksize="256 MiB"):
# build dict of dtypes to use when reading CSV
names, types = get_schema(table)
dtype = {names[i]: types[i] for i in range(0, len(names))}
types_dec_as_str = ['str' if 'decimal' in t else t for t in types]
dtype_dec_as_str = {names[i]: types_dec_as_str[i] for i in range(0, len(names))}

data_dir = config["data_dir"].split('parquet_')[0]
base_path = f"{data_dir}/data/{table}"
Expand All @@ -67,7 +67,7 @@ def read_csv_table(table, chunksize="256 MiB"):
if "audit" not in fn and os.path.getsize(f"{base_path}/{fn}") > 0
]
df = dask_cudf.read_csv(
paths, sep="|", names=names, dtype=dtype, chunksize=chunksize, quoting=3
paths, sep="|", names=names, dtype=dtype_dec_as_str, chunksize=chunksize, quoting=3
)
else:
paths = [
Expand All @@ -83,9 +83,15 @@ def read_csv_table(table, chunksize="256 MiB"):
if os.path.getsize(f"{base_path}/{fn}") > 0
]
df = dask_cudf.read_csv(
paths, sep="|", names=names, dtype=types, chunksize=chunksize, quoting=3
paths, sep="|", names=names, dtype=types_dec_as_str, chunksize=chunksize, quoting=3
)

for i, dtype in enumerate(types):
if 'decimal' in dtype:
precision = int(dtype[dtype.find("(")+1:dtype.find(",")])
scale = int(dtype[dtype.find(",")+1:dtype.find(")")])
df[names[i]] = df[names[i]].astype(Decimal64Dtype(precision, scale))

return df


Expand Down Expand Up @@ -133,7 +139,7 @@ def repartition(table, outdir, npartitions=None, chunksize=None, compression="sn
def main(client, config):
# location you want to write Parquet versions of the table data
data_dir = config["data_dir"].split('parquet_')[0]
outdir = f"{data_dir}/parquet_{part_size}gb/"
outdir = f"{data_dir}/parquet_{part_size}gb_decimal/"

t0 = time.time()
for table in tables:
Expand Down