[R] Better error handling for DatasetFactory$Finish() when no format specified

[asfimport]

When I call the following code:

 

tf <- tempfile()
dir.create(tf)
on.exit(unlink(tf))
write_csv_arrow(mtcars[1:5,], file.path(tf, "file1.csv"))
write_csv_arrow(mtcars[6:11,], file.path(tf, "file2.csv"))
ds <- open_dataset(c(file.path(tf, "file1.csv"), file.path(tf, "file2.csv")))

I get the following error: 

 Error: IOError: Could not open parquet input source '/tmp/RtmpSug6P8/file714931976ac54/file1.csv': Invalid: Parquet magic bytes not found in footer. Either the file is corrupted or this is not a parquet file.

However, in the documentation for open_dataset(), there is nothing saying that the input source cannot be a CSV or must be a Parquet file.  

I think this is due to calling DataSetFactory$Finish() when schema is NULL and input files have no inherent schema (i.e. are CSVs).

Reporter: Nicola Crane / @thisisnic
Assignee: Nicola Crane / @thisisnic

PRs and other links:

• GitHub Pull Request #10326

_{Note: This issue was originally created as ARROW-12791. Please see the migration documentation for further details.}

[asfimport]

Neal Richardson / @nealrichardson:
Judging from the error message, it's because "parquet" is the default file format, and those aren't Parquet files. https://arrow.apache.org/docs/r/reference/dataset_factory.html

[asfimport]

Neal Richardson / @nealrichardson:
You could try to catch the "Parquet magic bytes not found in footer" error message inside open_dataset() and return a different/helpful message like "Looks like these are not parquet files, did you mean to specify a 'format'?" or something.

[asfimport]

Neal Richardson / @nealrichardson:
Issue resolved by pull request 10326
#10326