Keep local copies of files in a separate mets:FLocat

[kba]

One early decision that has haunted us for years now is that we have been using a single mets:FLocat for both the original URL of a mets:file and the local copy in the workspace we use for processing.

This PR tries to solve #323 by changing OcrdFile and the download logic in Resolver and Workspace:

• the original URL (OcrdFile.url) remains in mets:FLocat[@LOCTYPE="URL"]/xlink:href
• the local filename (OcrdFile.local_filename) will now be written to an additional mets:Flocat[@LOCTYPE="OTHER"][@OTHERLOCTYPE="FILE"]/xlink:href
• The contract of Workspace.download_file is that after calling it with OcrdFile f, f will have a local_filename attribute and that is what processors should use rather than the url.
• The logic of when to download and when to copy files in Resolver.download_to_directory and Workspace.download_file has been adapted accordingly.

The goal here is to make the OCR-D processing non-invasive. Currently, once you do ocrd workspace find --download, the original URL will be gone. With this PR, ocrd workspace find --download will add an additional mets:Flocat which can then be removed after processing is finished (to be compliant with the DFG Viewer METS profile) with ocrd workspace find --undo-download.

[MehmedGIT]

When the files are downloaded there is an output with relative file path per line. When I did undo the downloading it returned just None on each line. Since the local path is removed, this makes sense, however, how should the general user interpret that output?

[kba]

Since the local path is removed, this makes sense, however, how should the general user interpret that output

Good point, I hadn't thought about that. Should be fixed in 0f26809

For the kant_aufklaerung_1784 test asset:

ocrd workspace find --download
OCR-D-IMG/INPUT_0017.tif
OCR-D-IMG/INPUT_0020.tif
OCR-D-GT-PAGE/PAGE_0017_PAGE.xml
OCR-D-GT-PAGE/PAGE_0020_PAGE.xml
OCR-D-GT-ALTO/PAGE_0017_ALTO.xml
OCR-D-GT-ALTO/PAGE_0020_ALTO.xml

and the reverse:

ocrd workspace find --undo-download
Removed local_filename OCR-D-IMG/INPUT_0017.tif
Removed local_filename OCR-D-IMG/INPUT_0020.tif
Removed local_filename OCR-D-GT-PAGE/PAGE_0017_PAGE.xml
Removed local_filename OCR-D-GT-PAGE/PAGE_0020_PAGE.xml
Removed local_filename OCR-D-GT-ALTO/PAGE_0017_ALTO.xml
Removed local_filename OCR-D-GT-ALTO/PAGE_0020_ALTO.xml

[MehmedGIT]

Yes, that output is more convenient. Great!