improvements.md

Areas for improvement

• Organization of Spectrum::Config::

  • source.rb needs to be split up because it contains multiple classes.
  • Lots of other classes should be split up for conceptual and organization reasons.

• Solr and Summon query generation

  • Remove dependency on CLIO and Blacklight (spectrum:app/models/spectrum/search_engines/solr.rb, spectrum:app/models/spectrum/search_engines/summon.rb).
  • Put more query generation logic in the same place.
  • Generate queries that work as planned for solr (i.e. expand aliases so they have a real pf or bq)
    • field:value => _query_:"{!dismax qf=$field.qf pf=$field.pf bq=$field.bq v=$field.clauseN}"
    • query generation will need to add arbitrary variables to the query (i.e. $field.clauseN)
  • Some of this is in spectrum-json:lib/spectrum/field_tree/ files
  • facet generation is handled some in spectrum-config:lib/spectrum/config/source.rb

• Anything that needs to access record metadata for making decisions.

  • Could use a decorator or something informed by the configuration.
  • Export code (RIS, and CSL, COinS) was shoehorned in late and in a hurry
    • spectrum-config:lib/spectrum/config/focus.rb
  • Holdings processing
  • Get This url generation

• Pride's query parser was never really developed.

  • Code is at: https://github.com/mlibrary/pride/blob/master/source/parser/parser.pegjs
  • It uses https://pegjs.org/
  • What we have is a stub that was being used as a placeholder when things were launched.
  • We don't have agreement on what the parse should be for various edge cases.

• Filters / data rewriting altering

  • Filters like the proxy prefix needs to know about the user's affiliation.
  • Currently accomplished by passing the request object all the way down to the filter to make that decision.
  • Maybe move towards building each request's filter stack with that information in mind, rather than assembling one stack to apply every time.