Strip paths during retrieval

[sbliven]

Summary

Should we change the behavior of datasetRetriever to strip leading components of the path when retrieving?

Current behavior

Say we have a dataset with creationLocation: /das/work/p12/p12345/data. When retrieving this with datasetRetriever PID retrieve/ it will actually put the dataset at retrieve/das/work/p12/p12335/data/. The user then needs to run mv retrieve/das/work/p12/p12335/* retrieve/; rm -r retrieve/das.

Proposed behavior

Several users have mentioned being annoyed by this behavior. They would prefer the dataset to be retrieved directly to retrieve/data.

Discussion

This current behavior the following advantages:

• Datasets can be retrieved to their original location with datasetRetriever PID /
• It works for datasets spread over multiple directories (specified by the ingestor fileListing.txt option)

Is there some way we could support the proposed behavior for the common single-folder case without breaking the multi-folder case? Maybe add a --skip-directories option or similar?