For Developers 

• The import
  • How a run works
    • An example run
    • Why persisting takes several rounds
    • When something fails
  • Records, identity and languages
    • How a record is recognised
      • Records without an @id of their own
    • How far references are followed
    • Languages
    • Values that disappear upstream
    • Entities are shared between records
  • Adding a record kind
    • The town as an example
    • The table
    • The entity
    • Can the kind be a root?
    • Relations to and from the kind
    • Checklist
  • Relations
    • References to other upstream objects
      • What the log says about references
      • One property, several target tables
    • Child records built from nested data
    • Sets of shared targets
      • Several upstream shapes, one set
      • Collect during the run, write once at the end
      • Removal, and when not to remove
      • Each property gets its own everything
      • Tests a new property of this shape needs
    • Events matched to places
  • Values stored as category trees
    • Where a tree lives: anchors and scopes
    • Creating the records
    • The @type hierarchy
      • Building one chain
    • Keywords
  • Media files
    • Where an image goes
    • Download, staging and promotion
    • Re-imports and removal
  • The import log
    • Severity
    • Entry types
    • Adding an entry type
  • What a run imports
    • The existing types
    • Adding a type
  • Import tuning
    • Recovered retries
    • Fetch cache
  • Testing an import
    • The test case
    • Habits that keep tests honest
  • Terms used throughout
• Filter fields
  • Registering a filter field
    • What each argument means
    • What a field does not decide
    • Adding a further storage shape
  • How the search form finds its list
• Backend relation scoping
  • The criterion
  • Event listeners
  • Adding a record kind