tpcbench.py add --query support to run custom query

[zhangxffff]

1. add --query argument suport for tpcbench.py to run custom query with tpch tables.
2. fix scripts in docs/contributing.md

[zhangxffff]

test in local environment

>> RAY_COLOR_PREFIX=1 RAY_DEDUP_LOGS=0 python tpcbench.py --data=tpch --concurrency=2 --batch-size=8182 --worker-pool-min=10 --validate --query 'select c.c_name, sum(o.o_totalprice) as total from orders o inner join customer c on o.o_custkey = c.c_custkey group by c_name order by c_name limit 1' 
Executing custom query:  select c.c_name, sum(o.o_totalprice) as total from orders o inner join customer c on o.o_custkey = c.c_custkey group by c_name order by c_name limit 1
2025-03-12 11:30:36,038 INFO worker.py:1832 -- Started a local Ray instance. View the dashboard at http://127.0.0.1:8265 
Registering table customer using path tpch/customer.parquet
Registering table lineitem using path tpch/lineitem.parquet
Registering table nation using path tpch/nation.parquet
Registering table orders using path tpch/orders.parquet
Registering table part using path tpch/part.parquet
Registering table partsupp using path tpch/partsupp.parquet
Registering table region using path tpch/region.parquet
Registering table supplier using path tpch/supplier.parquet
Writing results to datafusion-ray-tpch-1741750238210.json
statements = ['select c.c_name, sum(o.o_totalprice) as total from orders o inner join customer c on o.o_custkey = c.c_custkey group by c_name order by c_name limit 1']
executing  select c.c_name, sum(o.o_totalprice) as total from orders o inner join customer c on o.o_custkey = c.c_custkey group by c_name order by c_name limit 1
+--------------------+-----------+
| c_name             | total     |
+--------------------+-----------+
| Customer#000000001 | 587762.91 |
+--------------------+-----------+
done with query custom query
{
    "engine": "datafusion-ray",
    "benchmark": "tpch",
    "settings": {
        "concurrency": 2,
        "batch_size": 8182,
        "prefetch_buffer_size": 0,
        "partitions_per_worker": null
    },
    "data_path": "tpch",
    "queries": {
        "custom query": 0.19914793968200684
    },
    "validated": {
        "custom query": true
    }
}
benchmark complete. sleeping for 3 seconds for ray to clean up

[zhangxffff]

Hi @robtandy , Would you mind reviewing this PR when you get a chance? This is the follow-up work we had discussed in #82 . Thanks for your time!

[jazracherif]

Hello,

Thank you for this PR!

I'm new to this project and ran into the exact issues handled by the PR, was going to submit one until I found there was already one, so just providing some of my thoughts on it!

[jazracherif]

minor suggestion, extract this into its own functions, for example

from typing import List
def get_sql_queries(tpch_qnum: str = None, sql_statement: str= None) -> List[(str, str)]:
    """
    Get the list of SQL statements from either the TPCH or user provided SQL statements.
    At most one of these parameters can be provided.

    :param tpch_qnum: the TPCH Query number. If none, return all TPCH queries supported
    :param sql_statement: SQL string statement on available data tables (e.g ingested through make_data.py)
    :return: a list of tuples with name of the Query and the string SQL statement
    """

[jazracherif]

Suggested change

	help="Custom query to run with tpch tables")
	help="Custom SQL query statement to run with tpch tables")

[jazracherif]

explicitly mention TPCH in the id

Suggested change

	queries.append((str(args.qnum), tpch_query(args.qnum)))
	queries.append((f"TPCH-{args.qnum)}", tpch_query(args.qnum)))

[jazracherif]

Suggested change

	queries = [(str(i), tpch_query(i)) for i in range(1, 23)]
	queries = [(f"TPCH-{i}", tpch_query(i)) for i in range(1, 23)]

[jazracherif]

I would recommend standardizing the data file directory to testdata/tpch and add the correct make_file.py command just above, for example

Suggested change

	RAY_COLOR_PREFIX=1 RAY_DEDUP_LOGS=0 python tpcbench.py --data=file:///path/to/your/tpch/directory/ --concurrency=2 --batch-size=8182 --worker-pool-min=10 --qnum 2
	RAY_COLOR_PREFIX=1 RAY_DEDUP_LOGS=0 python tpcbench.py --data=../testdata/tpch --concurrency=2 --batch-size=8182 --worker-pool-min=10 --qnum 2

add before this more documentation one make_file

• In the tpch directory, use make_data.py to create a TPCH dataset at a provided scale factor and an output director, such as the testdata directory

python make_data.py 1 "../testdata/tpch"

could also specify a env variable for this in the setup
TPCH_DATA=../testdata/tpch

and replace the examples with $TPCH_DATA