`tbl_file()` doesn't support wildcards

[rplsmn]

Hello,

When reading a table that is split in multiple parquet files, duckdb supports wildcards like this :

SELECT * 
'test/*.parquet';

source

This can be achieved with tbl() and double quoting like this :

tbl(con, "'test/*.parquet'")

But it doesn't work with tbl_file, because tbl_file checks if the files exist with file.exist(), which doesn't work with wildcards out of the box.
Source

The common approach to wildcard expansion with R base would be to wrap the file.exist() in Sys.glob like this :

Sys.glob(file.exists(path))

Which yields a vector of booleans values for each path found after expanding the wildcard.
I believe that combining this with any() would update your check while keeping it's intent :

if (!any(Sys.glob(file.exists(path)))) {
    stop("File '", path, "' not found", call. = FALSE)
  }

Would you agree to change the behaviour of tbl_file to allow wildcards ?
I can submit a PR for it.

Motivation : I like what you are doing with tbl_file and tbl_function for safer tbl() alternatives, but currently cannot recommend tbl_file at my company because we have a lot of folder-like tables.

Thank you for your time and have a great day

[krlmlr]

Thanks. Happy to add a check = TRUE argument that, if set to FALSE, will skip the check. I also wonder if we really need the check -- DuckDB will emit an error anyway if the file(s) cannot be accessed.

Releasing duckdb 1.4.1 as soon as upstream is ready. Would you like to contribute?

[rplsmn]

Absolutely yes, thank you !
How would you like me to proceed from here ?

[krlmlr]

Let's just remove the check for file existence in R and add a test that checks what happens in this case.