feat: Support homedir expansion in lazy/scan read functions

[alexander-beedie]

(@ritchie46: as mentioned yesterday)

We already expand the homedir ~/ for eager read function paths, but seem to have missed it for the lazy/scan equivalents; this adds the missing expansion there too.

[codspeed-hq]

CodSpeed Performance Report

Merging #16869 will not alter performance

_{Comparing alexander-beedie:lazy-homedir-expansion (d236a44) with main (ff34c69)}

Summary

✅ 37 untouched benchmarks

[ritchie46]

Not entirely sure, but I believe we can collect into a Arc<[]> directly.

[alexander-beedie]

(thinking out loud)
Will poke at it some more 😆

[alexander-beedie]

Hmm, can't see a way to avoid the intermediate Vec; the Arc seems to want a constructed object to reference? 🤔 If you can spot something cunning, feel free to point me at it or commit over the top :))

[ritchie46]

Now I am sure. :D

https://play.rust-lang.org/?version=stable&mode=debug&edition=2021&gist=82ca9fa7f2c267ad4811d7714747b7ae

[alexander-beedie]

Now I am sure. :D

https://play.rust-lang.org/?version=stable&mode=debug&edition=2021&gist=82ca9fa7f2c267ad4811d7714747b7ae

Lol... pretty sure that the collect there still creates an intermediate Vec, it just gets deallocated after conversion to Arc[i32] :)

[ritchie46]

Yes, but Rust might optimize that later.

[codecov]

Codecov Report

Attention: Patch coverage is 90.00000% with 3 lines in your changes missing coverage. Please review.

Project coverage is 81.42%. Comparing base (13d68ae) to head (d236a44).
Report is 7 commits behind head on main.

Files	Patch %	Lines
crates/polars-lazy/src/scan/csv.rs	87.50%	1 Missing ⚠️
crates/polars-lazy/src/scan/ipc.rs	87.50%	1 Missing ⚠️
crates/polars-lazy/src/scan/parquet.rs	85.71%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #16869      +/-   ##
==========================================
+ Coverage   81.36%   81.42%   +0.06%     
==========================================
  Files        1425     1425              
  Lines      187910   187988      +78     
  Branches     2705     2704       -1     
==========================================
+ Hits       152885   153070     +185     
+ Misses      34529    34421     -108     
- Partials      496      497       +1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[etiennebacher]

@alexander-beedie Just wondering if resolve_homedir() should also be used for write_ functions? For context, we want to fix a bug on this in r-polars (pola-rs/r-polars#1085) and we're not sure if the path expansion should be done in Rust or in R. I see that write_csv() calls file = normalize_filepath(file) in Python but with this PR this is handled in Rust (for read/scan functions).

Bottom line: if the user calls .write_csv("~/foo.csv"), should this path be directly passed to Rust or should it be modified before?