Implement all TPC-H benchmarks in https://github.com/narwhals-dev/narwhals/tree/main/tpch

[MarcoGorelli]

In https://github.com/narwhals-dev/narwhals/tree/main/tpch/notebooks, we have notebooks with which we can run the TPC-H queries via Narwhals for different backends

We're a bit inconsistent with their format, and are missing a few queries (because when we started, some functionality in Narwhals was missing).

I've made https://github.com/narwhals-dev/narwhals/tree/main/tpch for this purpose. Let's define the queries there, and then we can just import those into the notebooks.

The task is:

• choose a query. You can look at https://github.com/narwhals-dev/narwhals/tree/main/tpch/notebooks to see what's already been implemented (ignore the gpu folder, we'll get rid of that anyway)
• follow the setup instructions in https://github.com/narwhals-dev/narwhals-tpch on the README
• submit that new query to https://github.com/narwhals-dev/narwhals-tpch. For example, if you did Q2, then your pull request should add files queries/q2.py and execute/q2.py

---

Example pull request: narwhals-dev/narwhals-tpch#1

[FBruzzesi]

Maybe we should make a separate for these, like narwhals-tpch

I like the idea, maybe even benchmarking gets easier. #819 is a bit pointless with the tiny tiny subset of data we have in narwhals itself.

[MarcoGorelli]

Yup! Cool, I've made a subfolder, with a generate_data.py script https://github.com/narwhals-dev/narwhals/tree/main/tpch

We can use this to store queries, and check that they run, then in Kaggle I just clone this repo, import the queries function, and time them (taking the best of 7)

[marenwestermann]

Working on q10.

[montanarograziano]

Working on query11

[MarcoGorelli]

yay thanks! as a heads-up, I've moved the queries to the Narwhals repo itself (https://github.com/narwhals-dev/narwhals/tree/main/tpch) so we have everything together - I've excluded that folder from the build anyway

[montanarograziano]

yay thanks! as a heads-up, I've moved the queries to the Narwhals repo itself (https://github.com/narwhals-dev/narwhals/tree/main/tpch) so we have everything together - I've excluded that folder from the build anyway

I just read this, after making a PR in the other repo 😅.
Should I make a PR here with the same changes too?

[MarcoGorelli]

😆 sorry about that, my fault

sure, if it's not too much trouble (this repo will just be a bit pickier about linting, but if you look at the other queries in this repo that should help), else we can merge it there and I'll port it over here tomorrow

[MarcoGorelli]

looks like q8 is the only one left, exciting!

[IsaiasGutierrezCruz]

I'll take the q8 c:

[FBruzzesi]

All implemented 🙌🏼🚀