add python script to compare lance performance vs parquet TPCH

[hzhang86]

Compare lance vs parquet for TPCH Q1 and Q6 using SF1 dataset
Steps to run the benchmark:

1. cd lance/benchmarks/tpch
2. mkdir dataset && cd dataset
3. download parquet file lineitem from : "https://github.com/cwida/duckdb-data/releases/download/v1.0/lineitemsf1.snappy.parquet"; then rename it to "lineitem_sf1.parquet"
4. generate lance file from the parquet file in the same directory
5. cd ..
6. python3 benchmark.py q1

[changhiskhan]

Thanks @hzhang86

Transcribing some context from the email thread:

"A few issues can make lance faster in TPC-H, including compression, better row-group stats and pruning, and maybe partitioning. Parquet is better in that regards."

see #738 for context as well

[eddyxu]

@hzhang86 , could you add a README.md with the same content you described here in the issue?

[hzhang86]

sure

…

On Wed, Apr 5, 2023 at 9:10 AM Lei Xu ***@***.***> wrote: @hzhang86 <https://github.com/hzhang86> , could you add a README.md with the same content you described here in the issue? — Reply to this email directly, view it on GitHub <#749 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACYCVQM2WOLCTS4GYJBPE7DW7WKQ3ANCNFSM6AAAAAAWTSGYDQ> . You are receiving this because you were mentioned.Message ID: ***@***.***>

-- Best regards Hui Zhang

[hzhang86]

@hzhang86 , could you add a README.md with the same content you described here in the issue?
what do you mean by add a README.md in the Issue? Shall I just create an
Issue and copy the description?

[eddyxu]

@hzhang86 , could you add a README.md with the same content you described here in the issue?
what do you mean by add a README.md in the Issue? Shall I just create an
Issue and copy the description?

Oh, i mean that add a README.md in this PR, and put the instruction of how to run TPCH in the README, so that it could be easier for other people to learn how to reproduce the benchmark you have later.

[changhiskhan]

Thanks @hzhang86 !