🤖 This is an automated PR from Repo Assist.

Fixes the regex bug in CSV schema parsing reported in #348.

Root Cause

nameAndTypeRegex in CsvInference.fs was incorrectly using RegexOptions.RightToLeft. This caused the regex engine to scan right-to-left and match the first ( it found (scanning from the right), rather than the last ( from the left.

For a schema item like "Revenue (USD)(float)", the intended parse is:

• name = "Revenue (USD)"
• type = "float"

But with RightToLeft, the regex matched:

• name = "Revenue " (stopped at the inner ( in (USD))
• type = "USD)(float" (incorrect!)

Without RightToLeft, the standard greedy .+ in the name group correctly finds the last parenthesized group as the type, because it consumes as many characters as possible before backtracking.

Fix

Remove RegexOptions.RightToLeft from nameAndTypeRegex only (line 44). The other two regexes (typeAndUnitRegex and overrideByNameRegex) correctly retain RightToLeft because they need to handle cases like:

• "float(unit<sub)>" — unit string can contain <
• "Revenue->newName=float" — column names can contain =

Test Status

All 465 existing tests pass ✅. A new regression test is added:

[(Test)]
let ``Can read CSV with schema where column name contains parentheses``() =
  let csv = "Revenue (USD),Count\n100.0,1\n200.0,2"
  use reader = new System.IO.StringReader(csv)
  let df = Frame.ReadCsv(reader, schema="Revenue (USD)(float),Count(int)")
  List.ofSeq df.ColumnKeys |> shouldEqual ["Revenue (USD)"; "Count"]
  df.GetColumn(float)("Revenue (USD)") |> Series.values |> List.ofSeq |> shouldEqual [100.0; 200.0]

Trade-offs

None. The greedy (non-RightToLeft) behavior is the correct and intended behavior for nameAndTypeRegex. This is a pure bug fix with no behaviour changes for existing valid schemas.

Generated by Repo Assist · ◷

To install this agentic workflow, run

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