CSV auto parsing regression in insertCSVFromPath

[domoritz]

There seems to be a regression between 1.17 and 1.24.

This code used to work but doesn't anymore.

  await database.open({ query: { castTimestampToDate: true } });
  const url = await file.url();
  await database.registerFileURL(file.name, url);
  const connection = await database.connect();
  await connection.insertCSVFromPath(file.name, {
    name: file.name,
    schema: "main"
  });

We get an error Error: Invalid Input Error: Error in file "us-state-capitals.tsv": CSV options could not be auto-detected. Consider setting parser options manually..

I can read the file fine in the CLI.

D from read_csv_auto('us-state-capitals.tsv');
┌────────────────┬────────────────┬────────────┬────────┬────────────┬─────────┬─────────┬───────┐
│     State      │    Capital     │   Since    │  Area  │ Population │ MSA/µSA │   CSA   │ Rank  │
│    varchar     │    varchar     │    date    │ double │   int64    │  int64  │  int64  │ int64 │
├────────────────┼────────────────┼────────────┼────────┼────────────┼─────────┼─────────┼───────┤
│ Alabama        │ Montgomery     │ 1846-01-01 │  159.8 │     198525 │  373290 │  461516 │     3 │
│ Alaska         │ Juneau         │ 1906-01-01 │ 2716.7 │      32113 │   32113 │         │     3 │
│ Arizona        │ Phoenix        │ 1912-01-01 │  517.6 │    1680992 │ 4948203 │ 5002221 │     1 │
│ Arkansas       │ Little Rock    │ 1821-01-01 │  116.2 │     197312 │  742384 │  908941 │     1 │

from read_csv('us-state-capitals.tsv', AUTO_DETECT=true); also works.

See https://observablehq.com/d/9df8918a02ad22c6

This is blocking observablehq/stdlib#350

[carlopi]

Thanks for the detailed bug report, I am looking into this.

[kimmolinna]

I don't think that this is a bug. Nowadays you have to specify DuckDBDataProtocol. A clip from Readme

await db.registerFileURL('remote.parquet', 'https://origin/remote.parquet', DuckDBDataProtocol.HTTP, false);

And you can skip the last boolean.
This one is working (4 equals DuckDBDataProtocol.HTTP:

await database.registerFileURL(file.name, url, 4);

[domoritz]

Closing since it's not a bug.