fix(rust): Reject non-integral offset and length in AExpr::Slice

[wence-]

• Fixes [QST]: should pl.Expr.slice raise if the numeric arguments are not integral #16855

[codspeed-hq]

CodSpeed Performance Report

Merging #16874 will not alter performance

_{Comparing wence-:wence/fix/16855 (cbe4091) with main (e29d28e)}

Summary

✅ 37 untouched benchmarks

[wence-]

Signposts

[wence-]

This is arguably a "feature" improvement. Previously one could not ask for the tail for an expression with an expression value of unsigned integer type. That is, this enables the following:

df.select(pl.col("a").tail(pl.lit(1, pl.UInt32()))

The casting is similar to the rust side implementation of tail:

    /// Get the last `n` elements of the Expr result.
    pub fn tail(self, length: Option<usize>) -> Self {
        let len = length.unwrap_or(10);
        self.slice(lit(-(len as i64)), lit(len as u64))
    }

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 81.09%. Comparing base (8965a68) to head (cbe4091).

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #16874      +/-   ##
==========================================
- Coverage   81.10%   81.09%   -0.02%     
==========================================
  Files        1435     1435              
  Lines      189546   189555       +9     
  Branches     2712     2712              
==========================================
- Hits       153732   153719      -13     
- Misses      35314    35336      +22     
  Partials      500      500              

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

[ritchie46]

Ah.. Sorry for the inconvenience, but can you rebase? Seems to be some conflicts.

[wence-]

Ah.. Sorry for the inconvenience, but can you rebase? Seems to be some conflicts.

Done.