feat(queries): figuring out the perfect querie for getting objects

quite a few things are being considered, perf. wise: pagination, ordering, querying, depth.

modules/core/migrations/000-core.js

@@ -66,6 +66,8 @@ exports.up = async knex => {
     table.string( 'id' ).primary( )
     table.string( 'speckle_type' ).defaultTo( 'Base' ).notNullable( )
     table.string( 'applicationId' )
+    table.integer( 'totalChildrenCount' )
+    table.jsonb( 'totalChildrenCountByDepth' )
     table.jsonb( 'data' )
     table.string( 'author', 10 ).references( 'id' ).inTable( 'users' )
     table.string( 'description' )
@@ -74,6 +76,7 @@ exports.up = async knex => {
   } )
 
   await knex.raw( 'ALTER TABLE "objects" add column "serial_id" bigserial' )
+  await knex.raw( 'CREATE INDEX serial_idx ON objects(serial_id) ' )
 
   // Tree inheritance tracker
   await knex.schema.createTable( 'object_tree_refs', table => {
@@ -81,15 +84,16 @@ exports.up = async knex => {
     table.string( 'parent' ).index( null, 'HASH' )
     table.specificType( 'path', 'ltree' )
   } )
+  await knex.raw( `CREATE INDEX tree_path_idx ON object_tree_refs USING gist(path)` )
 
   await knex.schema.createTable( 'object_children_closure', table => {
-    table.string( 'parent' ).notNullable( )
-    table.string( 'child' ).notNullable( )
-    table.integer( 'minDepth' ).defaultTo( 1 ).notNullable()
-    // table.index( [ 'parent', 'child' ], 'pc_index' )
+    table.string( 'parent' ).notNullable( ).index()
+    table.string( 'child' ).notNullable( ).index()
+    table.integer( 'minDepth' ).defaultTo( 1 ).notNullable().index()
+    table.index( [ 'parent', 'child' ], 'parent_child_index' )
+    table.index( [ 'parent', 'minDepth' ], 'full_pcd_index' )
   } )
 
-  await knex.raw( `CREATE INDEX tree_path_idx ON object_tree_refs USING gist(path)` )
 
   // creates an enum type for db reference types (branch, tag).
   await knex.raw( `

modules/core/objects/services.js

@@ -88,15 +88,26 @@ module.exports = {
       batch.forEach( obj => {
 
         let insertionObject = prepInsertionObject( obj )
-
+        let totalChildrenCountByDepth = {}
+        let totalChildrenCountGlobal = 0
         if ( obj.__closure !== null ) {
           for ( const prop in obj.__closure ) {
             closures.push( { parent: insertionObject.id, child: prop, minDepth: obj.__closure[ prop ] } )
+            
+            totalChildrenCountGlobal++
+            
+            if( totalChildrenCountByDepth[ obj.__closure[prop].toString() ] )
+              totalChildrenCountByDepth[ obj.__closure[ prop ].toString() ]++
+            else 
+              totalChildrenCountByDepth[ obj.__closure[ prop ].toString() ] = 1
           }
         }
 
-        delete obj.__tree
-        delete obj.__closure
+        insertionObject.totalChildrenCount = totalChildrenCountGlobal
+        insertionObject.totalChildrenCountByDepth = JSON.stringify( totalChildrenCountByDepth )
+
+        delete insertionObject.__tree
+        delete insertionObject.__closure
 
         objsToInsert.push( insertionObject )
         ids.push( insertionObject.id )

modules/core/tests/objects.spec.js

@@ -156,13 +156,19 @@ describe( 'Objects', ( ) => {
 
     it( 'Should get object children', async ( ) => {
 
-      let nestedBoys = createAShitTonOfFuckingObjects( 300000 )
+      let objs_1 = createAShitTonOfFuckingObjects( 10000, 'noise__' )
+      let ids = await createObjects( objs_1 )
 
-      let ids = await createObjects( nestedBoys )
+      // let objs_2 = createAShitTonOfFuckingObjects( 20000, 'noise_2' )
+      // let ids2 = await createObjects( objs_2 )
+
+      // let objs_3 = createAShitTonOfFuckingObjects( 50000, 'noise_3' )
+      // let ids3 = await createObjects( objs_3 )
+      
       console.log( `base id is: ${ids[0]} ` )
+      console.log( `base id is: ${ids2[0]} ` )
+      console.log( `base id is: ${ids3[0]} ` )
 
-      let res = await getObjectChildren( '0_hash' )
-      // console.log( res )
     } ).timeout( 30000 )
 
   } )

test-queries/closure-fullcount.sql

@@ -0,0 +1,18 @@
+with objs as (
+	SELECT 
+-- 		child as id,
+		id,
+		serial_id, -- just for reference
+		"data"
+	FROM object_children_closure
+	JOIN objects ON objects.id = child
+	WHERE parent = '7919a52c017be262ee0daf1844c376d7'
+	AND "minDepth" < 1000
+-- 	AND (objects."data" -> 'sortValueA')::numeric <= 700
+-- 	AND (objects."data" -> 'sortValueA')::numeric > 100
+	ORDER BY id
+)
+SELECT * FROM objs
+RIGHT JOIN (SELECT count(*) FROM objs ) c(total_count) ON TRUE
+OFFSET 100
+LIMIT 200

test-queries/closure-simple-fast.sql

@@ -0,0 +1,7 @@
+-- Fast, because? 
+SELECT child, "data" FROM object_children_closure
+RIGHT JOIN objects ON objects.id = child
+WHERE parent = '7919a52c017be262ee0daf1844c376d7'
+ORDER BY id
+OFFSET 0
+LIMIT 20

test-queries/closure-simple-slow.sql

@@ -0,0 +1,20 @@
+-- SLOW, because? 
+-- because we were ordering it by the serial_id!
+-- sorting by the id (on which we actually do the join) is 10x faster. 
+-- nice, to say the least. 
+SELECT child as id, "data" FROM object_children_closure
+RIGHT JOIN objects ON objects.id = child
+WHERE parent = '509cb0c19594b731214d3ffed2c011df'
+-- minDepth is a way to limit asking for objects up to a specific nested depth. 
+-- this is useful, for example, when we want to get a stream's top level objects only. 
+-- AND "minDepth" < 1000
+-- better pagination routine:
+-- instead of using offset, we use the last item we "saw" as where clause (last seen id)
+-- this assumes that we are ordering results by their id. 
+-- if we would be ordering them by something else, this clause would need to change. 
+-- AND id > '5a29a1e000d94d8b9f4c6dd767235903'
+-- AND (objects."data" -> 'sortValueA')::numeric <= 700
+ORDER BY id
+-- ORDER BY serial_id 
+OFFSET 0
+LIMIT 200

test-queries/closure-two-stage.sql

@@ -0,0 +1,19 @@
+WITH ids AS (
+	SELECT child FROM object_children_closure
+	WHERE parent = '94a0a141c211f60c5e3f859baae125e9'
+	AND "minDepth" < 100 
+), 
+objs AS (
+	SELECT 
+		id, 
+		speckle_type,
+		"data"
+	FROM ids
+	JOIN objects ON ids.child = objects.id
+	WHERE
+		(objects."data" -> 'sortValueA')::numeric >= 100
+)
+SELECT * FROM objs 
+RIGHT JOIN (SELECT count(*) FROM objs ) c(totalCount) ON TRUE
+OFFSET 120
+LIMIT 1000

test-queries/materialised-fullcount.sql

@@ -0,0 +1,18 @@
+WITH ids AS (
+        SELECT DISTINCT unnest( string_to_array( ltree2text( subltree("path", 1, 2) ), '.') ) as obj_id
+        FROM object_tree_refs
+        WHERE parent = '0_hash'
+      ),
+      objs AS (
+        SELECT obj_id as id, speckle_type, "data"
+        FROM ids
+        JOIN objects ON ids.obj_id = objects.id
+        -- WHERE objects."data" @> '{"text": "This is object 1"}'
+        ORDER BY jsonb_path_query(data, '$.nest.orderMe' ) DESC
+      ),
+      childrenCount AS (SELECT count(*) FROM ids),
+      resultCount AS (SELECT count(*) FROM objs)
+      SELECT * from objs
+      RIGHT JOIN (SELECT count(*) FROM objs) d(totalCount) ON TRUE
+      OFFSET 100
+      LIMIT 200

test-queries/materialised-simple-ordinality.sql

@@ -0,0 +1,11 @@
+	WITH ids as (
+		SELECT (subltree("path", 2, 3))::text as obj_id
+		FROM object_tree_refs
+		WHERE path ~ '0_hash.*{2}'
+		ORDER BY id
+	)
+	SELECT id, speckle_type, "data" -> 'nest' -> 'orderMe'
+	FROM ids
+	JOIN objects ON obj_id = objects.id WITH ORDINALITY
+	OFFSET 2
+	LIMIT 50

test-queries/materialised-simple.sql

@@ -0,0 +1,10 @@
+WITH ids AS(
+	SELECT unnest( string_to_array( ltree2text( subltree("path", 1, 3) ), '.') ) as obj_id
+	FROM object_tree_refs
+	WHERE parent = '0_hash'
+)
+SELECT obj_id, speckle_type, "data"
+FROM ids
+JOIN objects ON ids.obj_id = objects.id
+OFFSET 0
+LIMIT 100