update remap method to remove undefined mapping properties (#53)

If a field is not explicitly defined in an index's mapping we should not
persist it via the documents we copy over as part of remap.  This allows
for a way to get around the limitation in lucene where you can't delete
fields from index mappings without recreating the index from scratch.

CHANGELOG

@@ -1,3 +1,6 @@
+v0.10.0
+	- update remap to removing fields from the mapping that are not explicitly
+	defined.
 v0.9.1
 	- fix search enumerator, missing first result set
 v0.8.3

lib/elasticity/strategies/alias_index.rb

@@ -66,10 +66,17 @@ def remap(index_def)
             docs = @client.mget(body: { docs: id_docs }, refresh: true)["docs"]
             break if docs.empty?
 
+            # Modify document hashes to match the mapping definition so that legacy fields aren't added
+            defined_mapping_fields = index_def[:mappings][docs.first["_type"]]["properties"].keys
+
             # Move only documents that still exists on the old index, into the new index.
             ops = []
             docs.each do |doc|
-              ops << { index: { _index: new_index, _type: doc["_type"], _id: doc["_id"], data: doc["_source"] } } if doc["found"]
+              if doc["found"]
+                legacy_fields = doc["_source"].keys - defined_mapping_fields
+                legacy_fields.each { |field| doc["_source"].delete(field) }
+                ops << { index: { _index: new_index, _type: doc["_type"], _id: doc["_id"], data: doc["_source"] } }
+              end
             end
 
             @client.bulk(body: ops)

lib/elasticity/version.rb

@@ -1,3 +1,3 @@
 module Elasticity
-  VERSION = "0.9.1"
+  VERSION = "0.10.0"
 end

spec/functional/persistence_spec.rb

@@ -16,7 +16,7 @@ def self.name
           c.strategy        = Elasticity::Strategies::SingleIndex
 
           c.mapping = {
-            properties: {
+            "properties" => {
               name: { type: "string", index: "not_analyzed" },
               birthdate: { type: "date" },
             },
@@ -85,7 +85,7 @@ class Cat < Animal
         c.index_base_name = "cats_and_dogs"
         c.strategy = Elasticity::Strategies::SingleIndex
         c.document_type  = "cat"
-        c.mapping = { properties: {
+        c.mapping = { "properties" => {
           name: { type: "string", index: "not_analyzed" },
           age: { type: "integer" }
         } }
@@ -103,7 +103,7 @@ class Dog < Animal
         c.index_base_name = "cats_and_dogs"
         c.strategy = Elasticity::Strategies::SingleIndex
         c.document_type = "dog"
-        c.mapping = { properties: {
+        c.mapping = { "properties" => {
           name: { type: "string", index: "not_analyzed" },
           age: { type: "integer" },
           hungry: { type: "boolean" }
@@ -165,7 +165,7 @@ def self.name
           c.strategy        =  Elasticity::Strategies::AliasIndex
 
           c.mapping = {
-            properties: {
+            "properties" => {
               id: { type: "integer" },
               name: { type: "string", index: "not_analyzed" },
               birthdate: { type: "date" },
@@ -243,6 +243,50 @@ def to_document
       expect(results.total).to eq(2010)
     end
 
+    it "does not copy over fields not defined in the mapping" do
+      john = subject.new(_id: 1, id: 1, name: "John", birthdate: "1985-10-31", sort: ['john'])
+      mari = subject.new(_id: 2, id: 2, name: "Mari", birthdate: "1986-09-24", sort: ['mari'])
+
+      john.update
+      mari.update
+
+      subject.flush_index
+      results = subject.search({})
+      expect(results.first.birthdate).to be
+
+      # no birthdate
+      subject = Class.new(Elasticity::Document) do
+        def self.name
+          "SomeClass"
+        end
+
+        configure do |c|
+          c.index_base_name = "users"
+          c.document_type   = "user"
+          c.strategy        =  Elasticity::Strategies::AliasIndex
+
+          c.mapping = {
+            "properties" => {
+              id: { type: "integer" },
+              name: { type: "string", index: "not_analyzed" },
+            },
+          }
+        end
+
+        attr_accessor :id, :name
+
+        def to_document
+          { id: id, name: name }
+        end
+      end
+
+      subject.remap!
+      subject.flush_index
+
+      results = subject.search({})
+      expect(results.first.respond_to?(:birthdate)).to be false
+    end
+
     it "recover from remap interrupts" do
       number_of_docs = 2000
       docs = number_of_docs.times.map do |i|