🤖 This PR was created by Repo Assist, an automated AI assistant.

Summary

Adds Frame.distinctRowsBy, a new frame-transformation function that retains only the first row (by index order) for each unique combination of values in the specified columns.

This is analogous to SQL SELECT DISTINCT col1, col2 FROM table — a pattern that comes up regularly (see #558).

Example

// Given a frame with duplicate rows:
//   A  B   C
// 0 x  1.0 10
// 1 y  2.0 20
// 2 x  1.0 30   ← duplicate of row 0 on (A, B)
// 3 y  2.0 40   ← duplicate of row 1 on (A, B)

df |> Frame.distinctRowsBy ["A"; "B"]
//   A  B   C
// 0 x  1.0 10
// 1 y  2.0 20

C#:

df.DistinctRowsBy("A", "B")

Implementation

distinctRowsBy is built on top of the existing filterRows primitive. For each row, it computes a key as an obj list of the requested column values. F# lists have structural equality and GetHashCode() so HashSet(obj list) correctly deduplicates any combination of standard .NET value types (strings, ints, floats, DateTime, etc.).

A C#-friendly [(Extension)] DistinctRowsBy(frame, [(ParamArray)] columns) overload is added to FrameExtensions.

Test Status

• src/Deedle/Deedle.fsproj — builds without errors ✅
• tests/Deedle.Tests/Deedle.Tests.fsproj — compiles without errors ✅
• Unit tests cannot be executed locally (requires .NET 5, environment has .NET 8+); three new tests cover partial deduplication, no-op, and full deduplication. Please verify via CI.

Trade-offs

• Uses obj list key rather than a typed tuple — keeps the implementation simple and avoids arity limits, at the cost of boxing value types per row per call. For large frames with many unique columns, a custom IEqualityComparer could improve performance, but this is sufficient for typical use.
• Missing values are treated as null in the key (two rows with missing values in the same columns are considered equal).

Generated by Repo Assist · ◷

To install this agentic workflow, run

gh aw add githubnext/agentics/workflows/repo-assist.md@30f2254f2a7a944da1224df45d181a3f8faefd0d