feat(DRAFT): Add Expr.sort_by

[dangotbanned]

Closes #2534

What type of PR is this? (check all applicable)

• 💾 Refactor
• ✨ Feature
• 🐛 Bug Fix
• 🔧 Optimization
• 📝 Documentation
• ✅ Test
• 🐳 Other

Related issues

• Related issue [Enh]: add Expr.first #2526
• Closes [Enh]: Add Expr.sort_by #2534

Checklist

• Code follows style guide (ruff)
• Tests added
• Documented the changes

If you have comments or can explain your changes, please do so below

Important

Very early version, eager only, but solves the issue in (#2528 (comment))

• Pushing this mostly as a proof-of-concept.
• Still need to work through how this integrates with the lazy .over(order_by=...) rule

[dangotbanned]

If anyone comes across this, I'm putting on a pause on it to first explore:
(#2483 (comment)), (#2483 (comment)), (#2483 (comment)), (#2534 (comment))

Some very early notes:

• Things are coming together quite quickly, since we have the rust impl of polars to draw from
• A lot of the complexity of ExprMetadata, _expression_parsing.py is just capturing rules that would only appear on specific nodes/groups
  • We might only need a small subset of the options in (#polars/22573) to appear on every node/function

Hoping to have something more concrete to show off later this week 🙏

For now, here's some fluffier guidance I'm trying to work towards

Brainstorming an Expr internal represention.

Notes

• Each Expr method should be representable by a single node
  • But the node does not need to be unique to the method
• A chain of Expr methods should form a plan of operations
• We must be able to enforce rules on what plans are permitted:
  • Must be flexible to both eager/lazy and individual backends
  • Must be flexible to a given context (select, with_columns, filter, group_by)
• Nodes & plans are:
  • Immutable, but
    • Can be extended/re-written at both the Narwhals & Compliant levels
  • Introspectable, but
    • Store as little-as-needed for the common case
    • Provide properties/methods for computing the less frequent metadata

References

• https://github.com/pola-rs/polars/blob/dafd0a2d0e32b52bcfa4273bffdd6071a0d5977a/crates/polars-python/src/lazyframe/visitor/expr_nodes.rs
• https://github.com/pola-rs/polars/blob/dafd0a2d0e32b52bcfa4273bffdd6071a0d5977a/crates/polars-plan/src/dsl/expr.rs
• https://github.com/pola-rs/polars/blob/dafd0a2d0e32b52bcfa4273bffdd6071a0d5977a/crates/polars-plan/src/dsl/function_expr/mod.rs
• https://github.com/pola-rs/polars/blob/dafd0a2d0e32b52bcfa4273bffdd6071a0d5977a/crates/polars-plan/src/dsl/options/mod.rs#L137-L172
• https://github.com/pola-rs/polars/blob/3fd7ecc5f9de95f62b70ea718e7e5dbf951b6d1c/crates/polars-plan/src/plans/options.rs#L35-L106
• https://github.com/pola-rs/polars/blob/3fd7ecc5f9de95f62b70ea718e7e5dbf951b6d1c/crates/polars-plan/src/plans/options.rs#L131-L236
• https://github.com/pola-rs/polars/blob/3fd7ecc5f9de95f62b70ea718e7e5dbf951b6d1c/crates/polars-plan/src/plans/options.rs#L240-L267
• https://github.com/pola-rs/polars/blob/6df23a09a81c640c21788607611e09d9f43b1abc/crates/polars-plan/src/plans/aexpr/mod.rs

Related

• chore: Refactor expression parsing again (yes, again) #2483 (comment)
• chore: Refactor expression parsing again (yes, again) #2483 (comment)
• chore: Refactor expression parsing again (yes, again) #2483 (comment)
• chore: Refactor expression parsing again (yes, again) #2483 (comment)
• (a7eeb0d#r2083562677)
• api: formalise expression expansion in group-by #2225
• refactor: pass ExprMetadata down to compliant Exprs from narwhals.Expr #1848
• [Enh]: Add Expr.sort_by #2534 (comment)
• api (low-prio): allow for multi-output expressions on both side of binary operation if they're the same #2291

WIP branch

(https://github.com/narwhals-dev/narwhals/tree/oh-nodes)