We weren't respecting storage adapter configuration for valkyrie! This fixes
that.

…modeltransformer

monograph indexer spec

… to generator PR

…rb; call the event after_create_concern

`Hyrax::FilterByType` already includes infrastructure for filtering by
collection and work types. Rather than implement a new switch, use its existing
`only_works?` and `only_collections?` to filter by "generic type".

Adds tests for collection member search using Valkyrie `Hyrax::PcdmCollection`.

Use the configured storage adapter

* adjust approach to converting files
* add custom query find_file_metadata_by_use which is a valkyrie replacement for af object filter_files_by_type
* add helper methods original_file, extracted_text, and thumbnail in FileSet resource

Co-authored-by: cjcolvar <cjcolvard@indiana.edu>

Refactor file use:
* added methods to Hyrax::FileSet defining which Valkyrie::Vocab::PCDMUse URIs should be used in all cases for the primary uses of original_file, extracted_text, and thumbnail.
* updated all uses of Valkyrie::Vocab::PCDMUse URIs for the primary uses to use the methods defined in FileSet

Added tests for new methods

Instead of putting the use in a use attributes, conversion from AF File copies over all types as they exist in AF into the types attribute of FileMetadata.  To find the use, the :include? method is used to check for a specific use in the types.  This avoids data duplication by not having both type and use.  And it avoids data integrity issues when we aren’t able to correctly identify the use out of the array of types.

Co-authored-by: cjcolvar <cjcolvard@indiana.edu>

Supersedes #4208.

…xt, thumbnail

2 changes here…

### Parts of FileSet

AF behavior #original_file, #extracted_text, and #thumbnail each return a single PCDM::File,  So this now has the methods that get these relatiionships from FileMetadata to return `.first`

### Versioning

Versioning was looking through to ActiveFedora through the original_file method defined in wings/hydra/pcdm/pcdm_valkyrization_behavior.  Since that was method was removed with the addition of Hyrax::FileSet#original_file, it was no longer getting the lookthrough behavior.  The tests related to versioning have been marked pending.  Issue #3923 addresses FileMetadata and versioning.

Add indexer to work_resource generator

Plant-based spicy

Requested code change to use each_with_object was expanded to include all places where the pattern applied in the effected file.
This also includes requested change to case statement.

handle conversion of custom use types for files

Use the test adapter in specs by doing:

`context 'some context', valkyrie_adapter: :test_adapter do`

Valkyrie-based FileSet objects don't have direct access to characterization
information. In ActiveFedora, we used a `#characterization_proxy` to provide
that information. Here, we provide a small service that can find the "primary"
file (`FileMetadata`) for a given FileSet, and ask it about its description.

The concept of "use" is more properly localized to `FileMetadata` than
`FileSet`, since `FileMetadata` holds the actual use (`#type`) attribute.

Rather than using some branching logic in `FileMetadata.for`, take advantage of
Valkyrie attributes' default values. When initializing a `FileMetadata`, assume
it's an `ORIGINAL_FILE` unless a different use is passed.

This query was added, but never registered. It needs to be added to the
`.queries` method for the query handler in order to be picked up by most
adapters.

This method is supplanted by a new custom query. Use that query instead
whereever `used_for?` was called.

Since `#type` is located in `FileMetadata` and there's a convenient custom
query, these methods mostly function to provide `.first`. My feeling is that
this is a bit obfuscating; why not encourage queriers to handle the edge case
where there are multiple matches in context appropriate ways instead? In some
cases that might be `.first`, but not always.

This also has the benefit of removing some business logic from the model.

Add an indexer for `Hyrax::PcdmCollection` and index `generic_type_sim`.

For valkyrie normalization, take advantage for the new
`Hyrax::FileMetadata::Use.uri_for` module method to avoid reproducing
logic. When the argument is already a URI, just return it to avoid casting
unnecessarily.

In the ActiveFedora normalizer, `RDF::URI` already implements equality
appropriately, so `#casecmp` isn't needed and we can flatten to a simple switch
statement.

The constant previously used an incorrect URI. `Thumbnail` should be
`ThumbnailImage`.

Avoid typing `Hyrax.query_service.custom_queries` everywhere. Provide a method
on the Hyrax level instead.

Implement Valkyrie indexing for Collections

Transitions from using the instance variable to using the already included method that returns the resource indexer.

Prior to this change, specs reported the following warning:

```console
DEPRECATION WARNING: Delegating constraints to arel is deprecated and
will be removed in Rails 6.0.
```

I removed the deprecation warnings with this change (e.g. adding an
explicit call to #arel).

…n-warning

Removing Rails 6.0 deprecation warning

Create `FileSetDescription` to find characterization for FileSets

When constructing a description of a Valkyrie description, prefer to use URIs.

Support custom URI use/type in `FileSetDescription`

This is a bit of a stab in the dark. However, we had a failed Ruby
build with the following message:

```console
1) Hyrax::FileSetCSVService when specifying terms and separator csv
     Failure/Error:
       ::CSV.generate do |csv|
         csv << terms.map do |term|
           values = file_set.send(term)
           values = values.respond_to?(:to_a) ? values.to_a : [values] # make sure we have an array
           values.join(multi_value_separator)
         end
       end

     NameError:
       uninitialized constant CSV
```

What I suspect to be in play is that the corresponding spec does not
similarly namespace the `CSV` constant. The production ruby file uses
`::CSV` whereas the spec ruby file uses `CSV`. This commit seeks to use
the same namespacing, hopefully reducing an intermittent error.

See the following  CircleCI build for the failures:

-  https://circleci.com/gh/samvera/hyrax/17063#tests/containers/3

…s-for-csv

Ensuring CSV usage has same namespace consideration

Valkyrize inherit_permissions_job

…modeltransformer

monograph indexer spec

… to generator PR

…yrie-indexer