Remove SINGLE_FILE_OUTPUT
option from COPY statement
#8621
Labels
enhancement
New feature or request
SINGLE_FILE_OUTPUT
option from COPY statement
#8621
Is your feature request related to a problem or challenge?
Figuring out how to use
SINGLE_FILE_OUTPUT
while writing https://github.com/apache/arrow-datafusion/blob/98a5a4eb1ea1277f5fe001e1c7602b37592452f1/datafusion/sqllogictest/test_files/repartition_scan.slt#L35-L38 was actually quite painful for me.This option controls if the COPY statement writes a single file or multiple files into a directory.
Having datafusion automatically figure out that writing to
foo.parquet
should be a single file certainly would be a better UX in my opinionDescribe the solution you'd like
Yes, COPY still uses single_file_output option of the FileSinkConfig struct. We could potentially update COPY to instead rely on inference based on the path rather than an explicit option. E.g. copy table to file.parquet vs copy table to folder/. Then single_file_output could be removed entirely.
Describe alternatives you've considered
No response
Additional context
@devinjdangelo removed the corresponding option from
LIstingTable
(akaCREATE EXTERNAL TABLE
) in #8604 and suggsted this in #8604 (comment)The text was updated successfully, but these errors were encountered: