when HDFS is configured with fs.defaultsFS pointing to blob storage, we are not able to give local path as source location in feature definition config. this is because hdfsutils.scala uses FileSystem.get(conf) API. instead using FileSystem.get(URI, conf) solves the problem where one can give source location as file:///xyz even though defaultFS points to blob.
Testing
manually tested using notebook session configured with fs.defaultFS= wasbs://xyz
FileSystem.get(conf).exists("file:///path") returns false
FileSystem.get(URI, conf).exists("file:///path") returns true
Does this PR introduce any user-facing changes?
[X ] No. You can skip the rest of this section.
[ ] Yes. Make sure to clarify your proposed changes.
Description
when HDFS is configured with fs.defaultsFS pointing to blob storage, we are not able to give local path as source location in feature definition config. this is because hdfsutils.scala uses FileSystem.get(conf) API. instead using FileSystem.get(URI, conf) solves the problem where one can give source location as file:///xyz even though defaultFS points to blob.
Testing
manually tested using notebook session configured with fs.defaultFS= wasbs://xyz FileSystem.get(conf).exists("file:///path") returns false FileSystem.get(URI, conf).exists("file:///path") returns true
Does this PR introduce any user-facing changes?