Feature-complete with decent tests

Author: pond

app/controllers/scimitar/schemas_controller.rb

@@ -14,6 +14,7 @@ class EngineConfiguration
       :application_controller_mixin,
       :exception_reporter,
       :optional_value_fields_required,
+      :schema_list_from_attribute_mappings,
     )
 
     def initialize(attributes = {})
@@ -22,7 +23,8 @@ def initialize(attributes = {})
       # Set defaults that may be overridden by the initializer.
       #
       defaults = {
-        optional_value_fields_required: true
+        optional_value_fields_required:      true,
+        schema_list_from_attribute_mappings: []
       }
 
       super(defaults.merge(attributes))

app/models/scimitar/engine_configuration.rb

@@ -139,11 +139,12 @@ module Resources
     #       # ...
     #       groups: [
     #         {
-    #           list: :users,          # <-- i.e. Team.users,
+    #           list:  :users,         # <-- i.e. Team.users,
     #           using: {
     #             value:   :id,        # <-- i.e. Team.users[n].id
     #             display: :full_name  # <-- i.e. Team.users[n].full_name
     #           },
+    #           class: Team, # Optional; see below
     #           find_with: -> (scim_list_entry) {...} # See below
     #         }
     #       ],
@@ -159,7 +160,10 @@ module Resources
     # example above, "find_with"'s Proc might look at a SCIM entry value which
     # is expected to be a user ID and find that User. The mapped set of User
     # data thus found would be written back with "#users=", due to the ":list"
-    # key declaring the method name ":users".
+    # key declaring the method name ":users". The optional "class" key is
+    # recommended but not really *needed* unless the configuration option
+    # Scimitar::EngineConfiguration::schema_list_from_attribute_mappings is
+    # defined; see documentation of that option for more information.
     #
     # Note that you can only use either:
     #
@@ -176,7 +180,8 @@ module Resources
     # == scim_mutable_attributes
     #
     # Define this method to return a Set (preferred) or Array of names of
-    # attributes which may be written in the mixing-in class.
+    # attributes which may be written in the mixing-in class. The names MUST be
+    # expressed as Symbols, *not* Strings.
     #
     # If you return +nil+, it is assumed that +any+ attribute mapped by
     # ::scim_attributes_map which has a write accessor will be eligible for
@@ -291,7 +296,7 @@ module Mixin
         # the result in an instance variable.
         #
         def scim_mutable_attributes
-          @scim_mutable_attributes ||= self.class.scim_mutable_attributes()
+          @scim_mutable_attributes ||= self.class.scim_mutable_attributes()&.map(&:to_sym)
 
           if @scim_mutable_attributes.nil?
             @scim_mutable_attributes = Set.new

app/models/scimitar/resources/mixin.rb

@@ -106,6 +106,47 @@
     # whatever that means for you receiving system in your model code.
     #
     #     optional_value_fields_required: false
+
+    # The SCIM standard `/Schemas` endpoint lists, by default, all known schema
+    # definitions with the mutabilty (read-write, read-only, write-only) state
+    # described by those definitions, and includes all defined attributes. For
+    # user-defined schema, this will typically exactly match your underlying
+    # mapped attribute and model capability - it wouldn't make sense to define
+    # your own schema that misrepresented the implementation! For core SCIM RFC
+    # schema, though, you might want to only list actually mapped attributes.
+    # Further, if you happen to have a non-compliant implementation especially
+    # in relation to mutability of some attributes, you may want to report that
+    # accurately in the '/Schemas' list, for auto-discovery purposes. To switch
+    # to a significantly slower but more accurate render method for the list,
+    # driven by your resource subclasses and their attribute maps, set:
+    #
+    #     schema_list_from_attribute_mappings: [...array...]
+    #
+    # ...where you provide an Array of *models*, your classes that include the
+    # Scimitar::Resources::Mixin module and, therefore, define an attribute map
+    # translating SCIM schema attributes into actual implemented data. These
+    # must *uniquely* describe, via the Scimitar resources they each declare in
+    # their Scimitar::Resources::Mixin::scim_resource_type implementation, the
+    # set of schemas and extended schemas you want to render. Should resources
+    # share schema, the '/Schemas' endpoint will fail since it cannot determine
+    # which model attribute map it should use and it needs the map in order to
+    # resolve the differences (if any) between what the schema might say, and
+    # what the actual underlying model supports.
+    #
+    # It is further _very_ _strongly_ _recommended_ that, for any
+    # +scim_attributes_map+ containing a collection which has "list:" key (for
+    # an associative array of zero or more entities; the Groups to which a User
+    # might belong is a good example) then you should also specify the "class:"
+    # key, giving the class used for objects in that associated collection. The
+    # class *must* include Scimitar::Resources::Mixin, since its own attribute
+    # map is consulted in order to render the part of the schema describing
+    # those associated properties in the owning resource. If you don't do this,
+    # and if you're using ActiveRecord, then Scimitar attempts association
+    # reflection to determine the collection class - but that's more fragile
+    # than just being told the exact class in the attribute map. No matter how
+    # this class is determined, though, it must be possible to create a simple
+    # instance with +new+ and no parameters, since that's needed in order to
+    # call Scimitar::Resources::Mixin#scim_mutable_attributes.
   })
 
 end