perf sliding window

[codspeed-hq[bot]]

CodSpeed Performance Report

Merging #535 will not alter performance

_{Comparing 422-perf-sliding-window (4c26d80) with main (f521d03)}

Summary

✅ 7 untouched benchmarks

[github-actions[bot]]

Looks like some formatting rules failed.

✨ The action has attempted automatic fixes ✨

If any were succesful, they were committed to the branch. We suggest using git pull --rebase to apply them locally.

If some errors could not be fixed automatically, you can:

🏎️ Get results locally by running pre-commit run --all-files 🕵️ Examine the results in the Run pre-commit section of this workflow pre-commit

We also strongly recommend setting up the ruff and black extensions to auto-format on save in your chosen editor.

[sarakolding]

@MartinBernstorff something fishy is going on here... pre-commit says it fails though no errors and unrelated tests are failing

[sarakolding]

just an aside, but I don't think I get what is going on here: if isinstance(processed_spec.df, pl.LazyFrame): dfs.append(processed_spec.collect().lazy()) else: dfs.append(processed_spec.df) if it's lazy then collect and back to lazy?

[sarakolding]

@MartinBernstorff apropos tsflattener; what is next step here? review?

[MartinBernstorff]

We should get this finished. I'd suggest:

• Change the input type of timedelta_days: int to dt.datetime and make required downstream edits
• Make it so no "stepping" is run if timedelta_days is None, and set that as the default
• Any other cleanup you think is required

After that, I'll give it a review and we can get it merged. Sounds good? 😊

[MartinBernstorff]

@sarakolding I've unassigned myself. Feel free to re-request a review when it's ready 👍

I'd prefer if this were completed relatively soon, so we don't forget what we've talked about :-)

[sarakolding]

• Change the input type of timedelta_days: int to dt.datetime and make required downstream edits

did you mean dt.timedelta?

(i just went ahead and assumed you did)