[CALCITE-2088] Add more complex end-to-end tests in "plus" module, using Chinook data set

• Type: Improvement
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 1.16.0
• Assignee: Piotr Bojko
• Reporter: Piotr Bojko
• Created: 15/Dec/17 11:28
• Updated: 27/Feb/24 22:23
• Resolved: 04/Mar/18 21:24

Description

As in following correspondence - we would like to have more tests with more complexity. This should lead to lower regressions.

Yes, please do.

On Dec 8, 2017, at 7:13 AM, [e-mail] wrote:

I've taken a look into quidem.

I will log jira for that and assign it to myself, ok?

On Tue, Dec 5, 2017, 22:26 Julian Hyde <[e-mail]> wrote:

More tests are always welcome.

I would be wary of adding a new approach (assertj-db). Over time we end up
with as many approaches as there are contributors, and so the code becomes
hard to maintain. Consider using quidem (see QuidemTest and various .iq
files in the code base); it combines assertion-based testing with the
simplicity of script-based tests.

This could be added to the “plus” module, where we don’t mind extra
dependencies, and don’t mind if the test suite takes a long time.

Julian

On Dec 2, 2017, at 3:40 PM, [e-mail] wrote:

Hello fellow calcite dev team,

I am building a database with use of calcite framework and decided that
instead of simple unit tests I will go only with integration tests. This
is
due the fact that my code only glues the calcite with data and configures
the whole thing decorating with web api and jdbc access (with avatica
).

I have some problems with calcite, possible bugs - some of them in apache
jira for calcite logged already. Those problem are visible through my
tests.

And with that in mind I have an idea for a new maven artifact for
calcite -
end to end tests for an example h2 database. Database could have some
tables with data - maybe 100k rows in all tables. Tests with assertj and
its derivatives, something like I've done in my project - see the
pastebin
https://pastebin.com/raw/mevih4k6 .

Such test set can help with lowering regressions establishing a common
ground for talking about the calcite behaviour on specific cases (which
can
be described through end2end tests).

The tech under such maven artifact can be pretty simple:

• h2 as a data source, maybe some other
• one properly complicated json calcite schema
• some tech for populating h2 with data (just for having data with some
  descriptive language, not a binary format)
• assertj-db for DSL in tests

What do You think?

Cheers,
Pete

Issue Links

• is related to: CALCITE-2009 Possible bug in interpreting ( IN ) OR ( IN ) logic
• links to: Github fork
• links to: Pull request