Add tpch test cases with data.

[liurenjie1024]

Which issue does this PR close?

Closes #6405

Rationale for this change

Move tpch verification/plan test into sqllogictest.

What changes are included in this PR?

Are these changes tested?

Yes, tested with cargo test -p datafusion --test sqllogictests

Are there any user-facing changes?

No.

[jackwener]

Thanks @liurenjie1024 .
cc @alamb

[comphead]

Not sure we need to have all that inserts
Imho its better to use existing approach with sqllogictest like

CREATE EXTERNAL TABLE aggregate_test_100_by_sql (
  c1  VARCHAR NOT NULL,
  c2  TINYINT NOT NULL,
  c3  SMALLINT NOT NULL,
  c4  SMALLINT,
  c5  INT,
  c6  BIGINT NOT NULL,
  c7  SMALLINT NOT NULL,
  c8  INT NOT NULL,
  c9  BIGINT UNSIGNED NOT NULL,
  c10 VARCHAR NOT NULL,
  c11 FLOAT NOT NULL,
  c12 DOUBLE NOT NULL,
  c13 VARCHAR NOT NULL
)
STORED AS CSV
WITH HEADER ROW
LOCATION '../../testing/data/csv/aggregate_test_100.csv'

SELECT avg(c12) FROM aggregate_test_100

@alamb

ddl by file

[jackwener]

Not sure we need to have all that inserts Imho its better to use existing approach with sqllogictest like

agree with you, it's easier to maintain.

[alamb]

Thank you @liurenjie1024 . This is going to be awesome.

I was looking at what gets run on master for the tpch queries.

https://github.com/apache/arrow-datafusion/blob/444f0c42c4f37103c51eb8e0e92ad140727ca16f/.github/workflows/rust.yml#L198-L213

My reading is that the CI actually runs the tpch data generator and generates the SF1 data (about 1GB in total size)

Thus in terms of this test I suggest:

1. remove all the data from the files
2. use the CREATE EXTERNAL TABLE command as suggested by @comphead and @jackwener to create table that point at the data generated by the tpch data generator
3. Figure out some way to run the tpch tests conditionally in CI (maybe an environment variable or a flag 🤔 ) so they still pass even when the data generator hasn't been run.

What do you think?

[alamb]

In terms of conditionally running the tests, perhaps we can special case tests that start with tpch_ the way we do with pg_compat tests: https://github.com/apache/arrow-datafusion/tree/main/datafusion/core/tests/sqllogictests#running-tests-postgres-compatibility

[liurenjie1024]

Oh, I didn't notice that we already have tpch verification in ci, so another round of tpch verification would not be necessary. The main goal of this pr is to move tpch test in benchmark into sqllogictest to make it easier to maintain, so I will remove unnecessary changes to avoid.

[liurenjie1024]

I think this is ready for review. cc @comphead @alamb @jackwener

[alamb]

Thank you @liurenjie1024 -- I think this looks great.

One improvement I would like to look into is pre-creating the tpch data (perhaps we can create it in a docker image and then just copy it down). This would increase the CI speed greatly I think

[alamb]

I looked into this a little more -- and note we are running this CI check sf 0.1 (rather than SF1) It seems like a good idea and means it takes only 8 seconds to make the data 👍

https://github.com/apache/arrow-datafusion/actions/runs/5108217213/jobs/9181887855?pr=6435

[alamb]

I was able to run these tests locally following the directions 👍