clickwrap 0.0.0 → 0.1.0

data/lib/clickwrap/import/external_receipt.rb

@@ -0,0 +1,241 @@
+# frozen_string_literal: true
+
+module Clickwrap
+  module Import
+    # `Clickwrap.import_external_receipt!` — record that another system owned
+    # the presentation, and say so.
+    #
+    # ===========================================================================
+    # When Stripe or DocuSign owned the presentation, DO NOT PRETEND YOUR
+    # APPLICATION CAPTURED IT.
+    #
+    # Someone else rendered the page. Someone else chose the wording, the
+    # ordering, the call to action, and whether the control started unselected.
+    # Someone else observed the click, if a click is even what happened. Your
+    # application learned about it afterwards, through an API response or a
+    # webhook, and what you hold is their account of it.
+    #
+    # An event written here is therefore built so that it CANNOT be mistaken for
+    # a local capture, by a person or by a query:
+    #
+    #   * `event_type` is `external_receipt`, not `capture`;
+    #   * `capture_channel` is `imported_provider`;
+    #   * `attribution_method` is `imported_provider`;
+    #   * there is NO presentation manifest and no manifest digest, because
+    #     there is no offer of ours to reproduce;
+    #   * `Event#human_action?` is false, so it never satisfies a predicate that
+    #     asks whether a person acted through a Clickwrap presentation; and
+    #   * the receipt carries the provider's name, their event id, their raw
+    #     receipt, and the validation state of whatever check we ran on it.
+    #
+    # It participates in host verification — this is real evidence and refusing
+    # to record it would only push it somewhere worse — but it participates as
+    # what it is. A provider receipt is upgraded into no guarantee the provider
+    # did not make.
+    # ===========================================================================
+    class ExternalReceipt
+      def initialize(policy:, actor:, provider_name:, provider_event_id:, provider_receipt: nil,
+                     verified_with: nil, verified_at: nil, occurred_at: nil, subject: nil,
+                     tenant: nil, because: nil, statements: nil)
+        @policy = policy
+        @actor = actor
+        @provider_name = provider_name.to_s
+        @provider_event_id = provider_event_id.to_s
+        @provider_receipt = provider_receipt
+        @verified_with = verified_with&.to_s
+        @verified_at = coerce_time(verified_at)
+        @occurred_at = coerce_time(occurred_at)
+        @subject = subject
+        @tenant = tenant
+        @because = because
+        @statement_keys = statements&.map(&:to_s)
+      end
+
+      attr_reader :policy, :actor, :provider_name, :provider_event_id, :provider_receipt,
+                  :verified_with, :verified_at, :occurred_at, :subject, :tenant, :because
+
+      # Returns the `Clickwrap::Event`, not a Receipt. Callers of an importer
+      # almost always want to look straight at the labelling — `event_type`,
+      # `capture_channel`, `provider_name`, and the deliberately absent
+      # `presentation_manifest_digest` — because that labelling is the point.
+      # `Clickwrap.receipt(event.id)` is one call away when a receipt is wanted.
+      def import!
+        validate!
+
+        existing = Event.find_by(policy_key: policy.key, idempotency_key: idempotency_key)
+        return existing if existing
+
+        now = Clickwrap.now
+        revision = PolicyRevision.freeze_for(policy)
+        event = nil
+
+        ::ActiveRecord::Base.transaction do
+          event = build_event(now, revision)
+          build_statements(event, now)
+
+          event.save!
+          event.finalize_integrity!
+        end
+
+        event
+      end
+
+      alias call import!
+
+      # One provider event is one Clickwrap event, forever. A webhook delivered
+      # three times, a reconciliation sweep, and a backfill script all land on
+      # the same key, so the same provider receipt cannot become two pieces of
+      # evidence that an auditor would have to reconcile by hand.
+      def idempotency_key
+        "external_receipt:#{provider_name}:#{provider_event_id}"
+      end
+
+      private
+
+      def validate!
+        if provider_name.strip.empty?
+          raise ArgumentError,
+                "An imported receipt needs `provider_name:`. \"Some provider said so\" is not " \
+                "provenance, and the name is what tells a reader years from now whose account " \
+                "of events this row contains."
+        end
+
+        return unless provider_event_id.strip.empty?
+
+        raise ArgumentError,
+              "An imported receipt needs `provider_event_id:` — the provider's own identifier " \
+              "for this act. It is how the same webhook delivered twice becomes one event " \
+              "instead of two, and how anyone can go back and ask the provider about it."
+      end
+
+      def statements_to_import
+        return @statement_keys.map { |key| policy.statement!(key) } if @statement_keys
+
+        # Optional statements are excluded for the same reason they are excluded
+        # from a legacy import: a provider receipt that says the account holder
+        # accepted a service agreement says nothing about an optional purpose it
+        # never presented.
+        policy.required_statements
+      end
+
+      def build_event(now, revision)
+        Event.new(
+          event_type: "external_receipt",
+          policy_key: policy.key,
+          policy_revision: revision,
+          actor: actor.is_a?(::ActiveRecord::Base) ? actor : nil,
+          actor_reference: actor_reference,
+          tenant_key: tenant_key.presence,
+          subject: subject.is_a?(::ActiveRecord::Base) ? subject : nil,
+          subject_key: subject_key,
+          capture_channel: "imported_provider",
+          attribution_method: "imported_provider",
+          occurred_at: occurred_at,
+          recorded_at_by_server: now,
+          idempotency_key: idempotency_key,
+          provider_name: provider_name,
+          provider_event_id: provider_event_id,
+          provider_receipt: normalized_provider_receipt,
+          provider_verification: verification_record,
+          reason: because,
+          retention_class_key: policy.retention_class_key,
+          canonical_schema_version: Clickwrap::CANONICAL_SCHEMA_VERSION,
+          gem_version: Clickwrap::VERSION,
+          application_version: Clickwrap.config.resolved_application_version,
+          created_at: now
+          # No presentation_manifest. Deliberately, permanently. The provider
+          # rendered whatever was rendered; there is no offer of ours to sign,
+          # and a synthesized one would make this row look like a capture.
+        )
+      end
+
+      # What we did to check the provider's account of events, stated at the
+      # strength it actually has. `checked_against_provider` means we asked them
+      # and they confirmed. `not_checked` means the value arrived and we took
+      # it at face value — which is a perfectly ordinary thing to do and a
+      # completely different claim.
+      def verification_record
+        {
+          "verified_with" => verified_with,
+          "verified_at" => Receipt.format_time(verified_at),
+          "state" => verified_with.present? ? "checked_against_provider" : "not_checked",
+          "means" => "Provenance and validation state of one provider's receipt as recorded by " \
+                     "this application. It preserves exactly the assurance that provider " \
+                     "supplied, and adds none. Clickwrap did not present this content and did " \
+                     "not observe this action."
+        }.compact
+      end
+
+      def normalized_provider_receipt
+        case provider_receipt
+        when nil then nil
+        when Hash then provider_receipt.deep_stringify_keys
+        when String then { "value" => provider_receipt }
+        else
+          return { "value" => provider_receipt.to_s } unless provider_receipt.respond_to?(:to_h)
+
+          provider_receipt.to_h.deep_stringify_keys
+        end
+      end
+
+      def build_statements(event, now)
+        statements_to_import.each_with_index do |statement, index|
+          event.statements.build(
+            ordinal: index,
+            statement_key: statement.key,
+            kind: statement.kind,
+            action: statement.initial_action,
+            assertion_text: assertion_text_for(statement),
+            assertion_locale: "en",
+            required: statement.required?,
+            optional: statement.optional?,
+            answer: nil,
+            answered: false,
+            purpose_key: statement.purpose_key,
+            withdrawal_path: statement.withdrawal_path,
+            valid_from: occurred_at || now,
+            expires_at: statement.expires_after(occurred_at || now),
+            one_time: statement.one_time?,
+            requires: statement.requires,
+            created_at: now
+          )
+        end
+      end
+
+      def assertion_text_for(statement)
+        "Recorded from #{provider_name}'s receipt #{provider_event_id}: it states that this " \
+          "actor #{Vocabulary.initial_action_for(statement.kind)} #{statement.key}. " \
+          "#{provider_name} owned the presentation, so the exact wording, controls, and call to " \
+          "action belong to that provider and were not recorded here."
+      end
+
+      def actor_reference
+        @actor_reference ||= Reference.actor(actor)
+      end
+
+      # Provider timestamps arrive as whatever the API client handed over — a
+      # Time, a TimeWithZone, or an ISO 8601 String straight off the wire. All of
+      # them become UTC here, and an unparseable one becomes nil rather than
+      # crashing a webhook handler over a field that is provenance, not evidence.
+      def coerce_time(value)
+        return nil if value.nil?
+        return value.utc if value.respond_to?(:utc)
+
+        parsed = value.respond_to?(:to_time) ? value.to_time : Time.parse(value.to_s)
+        parsed&.utc
+      rescue ArgumentError, TypeError
+        nil
+      end
+
+      def subject_key = StatementState.subject_key_for(subject)
+
+      def tenant_key
+        return "" if tenant.nil?
+        return tenant.to_s if tenant.is_a?(String) || tenant.is_a?(Symbol)
+        return tenant.to_gid.to_s if tenant.respond_to?(:to_gid)
+
+        "#{tenant.class.name}/#{tenant.id}"
+      end
+    end
+  end
+end

data/lib/clickwrap/import/fine_print.rb

@@ -0,0 +1,290 @@
+# frozen_string_literal: true
+
+module Clickwrap
+  module Import
+    # `bin/rails clickwrap:import:fine_print[:plan]` — read FinePrint's tables
+    # and turn its signatures into explicit `imported_legacy` events.
+    #
+    # ===========================================================================
+    # FinePrint is established Rails prior art for versioned contracts and
+    # signature gates, and this importer exists because applications outgrow a
+    # question, not because they chose badly. FinePrint answers "did user U sign
+    # version N of contract X?", and it answers it well. What it does not record
+    # — because it was never trying to — is the presentation, the exact wording
+    # beside the control, the call to action, the request context, or the domain
+    # action the signature authorized.
+    #
+    # Those fields therefore come across as `unknown` / `not_collected`. This
+    # importer synthesizes none of them. A migration that filled in today's
+    # Terms digest for a 2019 signature would turn a modest, honest record into
+    # a confident false one, and the person reading the receipt in a dispute
+    # would have no way to tell.
+    #
+    # Read against FinePrint's signature model at the audited commit:
+    # https://github.com/openstax/fine_print/blob/3b75fbcbcfb048ecd2f4ee7c4f0b9bd3d10f7603/app/models/fine_print/signature.rb#L1-L33
+    #
+    # NOTE ON COUPLING: this class deliberately does NOT depend on the
+    # `fine_print` gem, require any of its files, or reference any of its
+    # constants. It reads two tables through the host's own connection, if they
+    # are there, and columns are discovered rather than assumed. A migration
+    # tool that forces you to keep the gem you are migrating away from installed
+    # is a migration tool with a hostage.
+    # ===========================================================================
+    class FinePrint
+      CONTRACTS_TABLE = "fine_print_contracts"
+      SIGNATURES_TABLE = "fine_print_signatures"
+
+      # What FinePrint's schema does not contain, recorded on every imported
+      # event so the gap is stated rather than inferred from silence.
+      UNKNOWN_FIELDS = %w[
+        exact_document_bytes
+        presentation_manifest
+        assertion
+        submit_button_text
+        protected_action
+        request_evidence
+        ip_address
+        browser_user_agent
+      ].freeze
+
+      # A plan or an import, described the same way either way.
+      Report = Data.define(:status, :policy_key, :tables_present, :contracts, :signatures,
+                           :results, :message) do
+        def possible? = status != :tables_absent
+        def imported = results.select(&:imported?)
+        def already_imported = results.select(&:already_imported?)
+        def planned = results.select(&:planned?)
+        def to_s = message
+      end
+
+      # One FinePrint contract version, and whether this application has a
+      # published Clickwrap document version that corresponds to it.
+      ContractMapping = Data.define(:name, :version, :title, :document_key, :version_label,
+                                    :published_version_id, :signature_count) do
+        def published? = !published_version_id.nil?
+      end
+
+      class << self
+        # Reads everything, writes nothing.
+        def plan(**) = new(dry_run: true, **).call
+
+        def import!(**) = new(**).call
+      end
+
+      # `map_contract_with` receives a contract row (a plain Hash of column name
+      # to value) and returns the Clickwrap document key that contract
+      # corresponds to. `find_actor_with` receives (user_type, user_id) and
+      # returns the actor record, or a stable actor reference string, or nil to
+      # skip that signature.
+      def initialize(policy_key:, find_actor_with:, map_contract_with: nil, contract_names: nil,
+                     because: nil, limit: nil, dry_run: false)
+        @policy_key = policy_key.to_s
+        @find_actor_with = find_actor_with
+        @map_contract_with = map_contract_with
+        @contract_names = contract_names&.map(&:to_s)
+        @because = because
+        @limit = limit
+        @dry_run = dry_run
+      end
+
+      attr_reader :policy_key, :find_actor_with, :map_contract_with, :contract_names,
+                  :because, :limit, :dry_run
+
+      def call
+        return tables_absent unless tables_present?
+
+        contracts = load_contracts
+        mappings = map_contracts(contracts)
+        results = import_signatures(contracts)
+
+        Report.new(
+          status: dry_run ? :planned : :imported,
+          policy_key: policy_key,
+          tables_present: true,
+          contracts: mappings,
+          signatures: results.length,
+          results: results,
+          message: summary(mappings, results)
+        )
+      end
+
+      private
+
+      def connection = ::ActiveRecord::Base.connection
+
+      def tables_present?
+        connection.data_source_exists?(CONTRACTS_TABLE) &&
+          connection.data_source_exists?(SIGNATURES_TABLE)
+      end
+
+      # Nothing to do, said plainly. An importer that raised here would make
+      # `clickwrap:doctor` and a migration checklist harder to run on an
+      # application that simply never used FinePrint.
+      def tables_absent
+        missing = [CONTRACTS_TABLE, SIGNATURES_TABLE].reject { |table| connection.data_source_exists?(table) }
+
+        Report.new(
+          status: :tables_absent, policy_key: policy_key, tables_present: false,
+          contracts: [], signatures: 0, results: [],
+          message: "FinePrint's tables are not in this database (missing: #{missing.join(", ")}), " \
+                   "so there is nothing to import and nothing was written. If you are migrating " \
+                   "from FinePrint, run this against the database that still has its tables."
+        )
+      end
+
+      # Columns are read, not assumed. FinePrint's schema is stable at the
+      # audited commit, but a host may have added to it, and an importer that
+      # hard-codes a SELECT list breaks on a database it could have read fine.
+      def contract_columns = @contract_columns ||= connection.columns(CONTRACTS_TABLE).map(&:name)
+      def signature_columns = @signature_columns ||= connection.columns(SIGNATURES_TABLE).map(&:name)
+
+      def load_contracts
+        rows = connection.select_all("SELECT * FROM #{quoted(CONTRACTS_TABLE)}").to_a
+        return rows if contract_names.nil?
+
+        rows.select { |row| contract_names.include?(row["name"].to_s) }
+      end
+
+      def load_signatures
+        sql = "SELECT * FROM #{quoted(SIGNATURES_TABLE)}"
+        sql << " ORDER BY #{quoted_column("id")}" if signature_columns.include?("id")
+        sql << " LIMIT #{limit.to_i}" if limit
+
+        connection.select_all(sql).to_a
+      end
+
+      # FinePrint contract versions become Clickwrap documents. The mapping is
+      # host-owned: only the application knows that its FinePrint contract named
+      # "terms_of_use" is the document this gem calls `:terms`. What this
+      # reports is whether the corresponding version is actually published here,
+      # because that is what decides whether an imported event can carry real
+      # document bytes or must record the label alone.
+      def map_contracts(contracts)
+        counts = signature_counts_by_contract_id
+
+        contracts.map do |row|
+          document_key = document_key_for(row)
+          label = row["version"].to_s
+          version = document_key && published_version(document_key, label)
+
+          ContractMapping.new(
+            name: row["name"].to_s,
+            version: label,
+            title: row["title"].to_s.presence,
+            document_key: document_key,
+            version_label: label,
+            published_version_id: version&.id,
+            signature_count: counts[row["id"]].to_i
+          )
+        end
+      end
+
+      def signature_counts_by_contract_id
+        return {} unless signature_columns.include?("contract_id")
+
+        connection
+          .select_all("SELECT contract_id, COUNT(*) AS signature_count FROM " \
+                      "#{quoted(SIGNATURES_TABLE)} GROUP BY contract_id")
+          .to_a
+          .to_h { |row| [row["contract_id"], row["signature_count"]] }
+      end
+
+      def import_signatures(contracts)
+        by_id = contracts.index_by { |row| row["id"] }
+
+        load_signatures.filter_map do |signature|
+          contract = by_id[signature["contract_id"]]
+          next if contract.nil?
+
+          actor = resolve_actor(signature)
+          next if actor.nil?
+
+          import_one(signature, contract, actor)
+        end
+      end
+
+      def import_one(signature, contract, actor)
+        Legacy.new(
+          policy: Clickwrap.policy!(policy_key),
+          actor: actor,
+          occurred_at: occurred_at_for(signature),
+          known: known_for(signature, contract),
+          unknown: UNKNOWN_FIELDS,
+          because: because || default_reason(signature, contract),
+          source: "fine_print",
+          capture_channel: "imported_provider",
+          dry_run: dry_run
+        ).import!
+      end
+
+      # FinePrint records when the signature row was created. That is the best
+      # time available and it is recorded as `occurred_at` — separate, as
+      # always, from when this import wrote the Clickwrap event down.
+      def occurred_at_for(signature)
+        signature["created_at"] || signature["updated_at"]
+      end
+
+      def known_for(signature, contract)
+        {
+          "source_system" => "fine_print",
+          "fine_print_signature_id" => signature["id"],
+          "fine_print_contract_id" => contract["id"],
+          "contract_name" => contract["name"],
+          "contract_title" => contract["title"],
+          # Deliberately NOT written as `document_version`: that key makes the
+          # legacy importer link published bytes, and a FinePrint version number
+          # is a label in another system's numbering, not a claim about ours.
+          "fine_print_contract_version" => contract["version"],
+          "signed_by_type" => signature["user_type"] || signature["signer_type"],
+          "signed_by_id" => signature["user_id"] || signature["signer_id"]
+        }.compact
+      end
+
+      def default_reason(signature, contract)
+        "Imported from FinePrint signature #{signature["id"]} for contract " \
+          "#{contract["name"]} version #{contract["version"]}"
+      end
+
+      def resolve_actor(signature)
+        find_actor_with.call(
+          signature["user_type"] || signature["signer_type"],
+          signature["user_id"] || signature["signer_id"]
+        )
+      end
+
+      def document_key_for(contract)
+        return map_contract_with.call(contract)&.to_s if map_contract_with
+
+        contract["name"].to_s.presence
+      end
+
+      def published_version(document_key, label)
+        document = ::Clickwrap::Document.find_by(document_key: document_key, tenant_key: nil)
+        return nil unless document
+
+        document.versions.find_by(version_label: label)
+      end
+
+      def summary(mappings, results)
+        unpublished = mappings.reject(&:published?).map { |mapping| "#{mapping.name}@#{mapping.version}" }
+
+        [
+          dry_run ? "Would import" : "Imported",
+          "#{results.length} FinePrint signature(s) into #{policy_key} as imported_legacy events.",
+          "Presentation manifest, IP address, call-to-action text, and protected action were not " \
+          "recorded by FinePrint and are marked unknown rather than filled in.",
+          if unpublished.empty?
+            nil
+          else
+            "No published Clickwrap document version matches: " \
+              "#{unpublished.join(", ")}. Those imports record the label " \
+              "only, not document bytes."
+          end
+        ].compact.join(" ")
+      end
+
+      def quoted(table) = connection.quote_table_name(table)
+      def quoted_column(column) = connection.quote_column_name(column)
+    end
+  end
+end