CsvProvider - possible to retrieve exception when row fails to parse?

[NickLydon]

If a row in a csv file contains invalid data, but I still want to process the rest of the file, is it possible to do this and get the exception for the erroneous row?

I've supplied IgnoreErrors=true, but I believe this swallows the exception...

[ovatsus]

How would that look in terms of API?

[NickLydon]

I was just wondering if it was possible with the existing API, like subscribing to an event raised for on exception for instance.

If it isn't at the moment, perhaps there could be an exception property on the row object. Or the row itself could be a union type of a successfully parsed row or an exception.

[smoothdeveloper]

We could add an overload to Load + few supporting types, here is rough draft to get discussion going

// declared in FSharp.Data
type FailedRecord {
  FailureLocation: int * int // zero based (line * column) indices in the file
  RawRecord: string // always contain the record
  RawFields : string array // parsed fields, might be empty array
  // maybe some type of error as a DU?
}


type MyCsv = CsvProvider<....>

let handler (failedRecord: FailedRecord) =
  // can fail
  // failwith ""
  // or return a default record
  MyCsv.Record("Field1", 2, Guid.NewGuid())

let csv = MyCsv.LoadWithHandler(filename, fun (failedRecord: FailedRecord) -> failwithf "" )

What lacks from this IMHO is ability to obtain fields that were parsed successfully, maybe someone can chime-in with good idea for this or better proposal.

Another more robust optoin, would be to provide a IDataReader on which we have generated properties to access each column with correct type, plus IsValid method taking a column as argument, as well as exposing whole record. I think this would be better:

use reader = MyCsv.ExecuteReader(filename)
let results = [
  while reader.Read() do

   if reader.IsRecordValid then
     yield reader.Record
   else
     if not (reader.IsValid MyCsv.Columns.Column1) then
       // handle this case
       printfn "%s" reader.RawValue MyCsv.Columns.Column1
       ()
     // ...
]

[github-actions]

🤖 This is an automated response from Repo Assist.

Currently IgnoreErrors=true silently skips bad rows without any way to inspect the failure — the exception is swallowed internally (see CsvRuntime.fs). There is no built-in "error callback" or per-row error type yet.

Workaround — if you need error access, you can drop down to the lower-level CsvFile API and iterate rows manually:

let csv = CsvFile.Load("data.csv")
for row in csv.Rows do
    try
        let value = row.[0] |> int  // or your own type conversion
        // process row
        ()
    with ex ->
        printfn "Row failed: %s — %s" (String.concat "," row.Columns) ex.Message

This gives you full control: you can log the raw row, the exception, and continue processing the rest.

The structured approach suggested above (an IDataReader or LoadWithHandler overload) remains a valid enhancement idea, but hasn't been implemented in the typed provider layer. If you'd like to see it, a PR along those lines would likely be welcomed — the relevant code is in CsvRuntime.fs around the ignoreErrors path.

Generated by Repo Assist

To install this workflow, run gh aw add githubnext/agentics/workflows/repo-assist.md@b87234850bf9664d198f28a02df0f937d0447295. View source at https://github.com/githubnext/agentics/tree/b87234850bf9664d198f28a02df0f937d0447295/workflows/repo-assist.md.

[dsyme]

Good answer 👍