🤖 Repo Assist — automated bug fix (Task 3: Issue Investigation and Fix).

Closes #347

Root Cause

In FrameUtils.readCsv, the inferTypes=false code path unconditionally set all columns to string, ignoring the schema parameter entirely:

| Some false ->
    // Schema was never consulted here
    headers |> Array.map (fun c -> PrimitiveInferedProperty.Create(c, typeof(string), true, None))

The schema argument was only passed to InferColumnTypes in the inferTypes=true/None branch.

Fix

When inferTypes=false and a non-empty schema is provided, we now call InferColumnTypes so that the schema overrides are respected. Columns explicitly specified in the schema use the declared type; columns not mentioned in the schema are still inferred from data (which is the most useful behaviour when the caller has declared overrides for a subset of columns).

| Some false when schema <> "" ->
    data.InferColumnTypes(inferRows, missingValues, cultureInfo, schema, safeMode, preferOptionals)
    |> Array.ofSeq
| Some false ->
    // no schema: fall back to all-string behaviour as before
    ...

The XML documentation comments on ReadCsv are also updated to clarify that schema overrides are respected even when inferTypes=false.

Usage example

// Schema types are now respected even with inferTypes=false
let df = Frame.ReadCsv(reader, inferTypes=false, schema="Value=int,Flag=bool")
// Previously: all columns were string
// Now: Value is int, Flag is bool

Test Status

• New regression test Schema is respected when inferTypes=false (regression #347) added to Frame.fs
• All 495 existing tests pass on Ubuntu (dotnet test tests/Deedle.Tests/Deedle.Tests.fsproj -c Release)

Generated by Repo Assist · ◷

To install this agentic workflow, run

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