Tim Sweña (Swast)

So long as we leave out description (and other column metadata like IAM roles), which can contain arbitrary text (possibly, though unlikely to be PII) I do think it would...

@chansinhui This does sound like a bug. Do you have a small example we can use to reproduce this?

Also, this will allow us to remove our Circle CI config.

Type casting sounds like a good solution for this. It's actually similar to how we provide a `dtypes` dictionary parameter for `to_dataframe` which suffers a similar issue (nullable int64 columns...

The main problem is the server doesn't even send the full precision over the wire with the `tabledata.list` API. It does send the full precision with the BQ Storage API,...

> ANY help is welcome. Will update when I get to AVRO to R object representation. You'll likely get better performance with Arrow, as I assume R uses a column-oriented...

Discussed offline. Datasets are strongly consistent within a region, but we aren't using regional endpoints to create the table, so likely occasionally get a stale read for if the dataset...

For the project I'm working on, my ideal scenario would be the same kind of node as a table expression, but a special case for literal values wouldn't be so...

I don't think I had an existing node in mind. A coworker was experimenting with Table-Valued-Functions, which if implemented in general could get this working for the BigQuery backend. I...

Is this writing to an existing table? Could you share the schema of the destination table?