I'd note that applying a search or filter to data that returns no results is in general not in error, and arguably should result in an empty result set, not an exception. It seems the picture here is complicated by the fact that based on filtering alone the result set isn't empty, but becomes empty once we remove rows that don't reference a valid image? Then the question is, is it an error to have rows without image to begin with, or is it not.
Regardless of the answer, I feel this doesn't need to be sorted out here. We're at the prototype stage, and now the function at least either returns a result of a consistent type, or it doesn't return a result (by raising an exception), which seems good enough. So perhaps open an issue with the above question to be considered down the road if and when there's a lot more continued development.
Also noted in this review: Timing of filtering out "unknown" Image_filename and file_url values. It may be more efficient to filter these first with large data; however, do we wish to preserve (display) information about entries that may match user filters but be missing filename or path information?
Originally posted by @hlapp in https://github.com/Imageomics/dashboard-prototype/pull/33#pullrequestreview-1540235417
Also noted in this review: Timing of filtering out "unknown"
Image_filenameandfile_urlvalues. It may be more efficient to filter these first with large data; however, do we wish to preserve (display) information about entries that may match user filters but be missing filename or path information?