Revamp age csv loader (apache#2044)

* Allow 0 as entry_id - No regression test were impacted by this change. * Use batch inserts to improve performance - Changed heap_insert to heap_multi_insert since it is faster than calling heap_insert() in a loop. When multiple tuples can be inserted on a single page, just a single WAL record covering all of them, and only need to lock/unlock the page once. - BATCH_SIZE is set to 1000, which is the number of tuples to insert in a single batch. This number was chosen after some experimentation. - Change some of the field names to avoid confusion. * Use sequence for generating ids for edge and vertex - Sequence is not used if the id_field_exists is true in load_labels_from_file function, since the entry id is present in the csv. * Add function to create temporary table for ids - Created a temporary table and populate it with already generated vertex ids. A unique index is created on id column to ensure that new ids generated (using entry id from csv) are unique. * Insert generated ids in the temporary table to enforce uniqueness - Insert ids in the temporary table and also update index to enforce uniqueness. - If the entry id provided in the CSV is greater than the current sequence value, the sequence value is updated to match the entry ID. For example: Suppose the current sequence value is 1, and the CSV entry ID is 2. If we use 2 but not update the sequence to 2, next time the CREATE clause is used, 2 will be returned by sequence as an entry id, resulting in duplicate. - Update batch functions * Add functions to create graph and label automatically - These functions will check existence of graph and label, and create them if they don't exist. * Add regression tests
MuhammadTahaNaveed · Aug 19, 2024 · 750a09e · 750a09e
1 parent be771e0
commit 750a09e
Show file tree

Hide file tree

Showing 13 changed files with 779 additions and 105 deletions.
diff --git a/regress/expected/age_load.out b/regress/expected/age_load.out
@@ -19,41 +19,87 @@
 \! cp -r regress/age_load/data regress/instance/data/age_load
 LOAD 'age';
 SET search_path TO ag_catalog;
+-- Create a country using CREATE clause
 SELECT create_graph('agload_test_graph');
 NOTICE:  graph "agload_test_graph" has been created
  create_graph 
 --------------
 
 (1 row)
 
-SELECT create_vlabel('agload_test_graph','Country');
-NOTICE:  VLabel "Country" has been created
- create_vlabel 
----------------
-
+SELECT * FROM cypher('agload_test_graph', $$CREATE (n:Country {__id__:1}) RETURN n$$) as (n agtype);
+                                        n                                         
+----------------------------------------------------------------------------------
+ {"id": 844424930131969, "label": "Country", "properties": {"__id__": 1}}::vertex
 (1 row)
 
+--
+-- Load countries with id
+--
 SELECT load_labels_from_file('agload_test_graph', 'Country',
-    'age_load/countries.csv');
+    'age_load/countries.csv', true);
  load_labels_from_file 
 -----------------------
 
 (1 row)
 
-SELECT create_vlabel('agload_test_graph','City');
-NOTICE:  VLabel "City" has been created
- create_vlabel 
----------------
-
+-- A temporary table should have been created with 54 ids; 1 from CREATE and 53 from file
+SELECT COUNT(*)=54 FROM "_agload_test_graph_ag_vertex_ids";
+ ?column? 
+----------
+ t
+(1 row)
+
+-- Sequence should be equal to max entry id i.e. 248
+SELECT currval('agload_test_graph."Country_id_seq"')=248;
+ ?column? 
+----------
+ t
 (1 row)
 
+-- Should error out on loading the same file again due to duplicate id
+SELECT load_labels_from_file('agload_test_graph', 'Country',
+    'age_load/countries.csv', true);
+ERROR:  Cannot insert duplicate vertex id: 844424930131970
+HINT:  Entry id 2 is already used
+--
+-- Load cities with id
+--
+-- Should create City label automatically and load cities
 SELECT load_labels_from_file('agload_test_graph', 'City',
-    'age_load/cities.csv');
+    'age_load/cities.csv', true);
+NOTICE:  VLabel "City" has been created
  load_labels_from_file 
 -----------------------
 
 (1 row)
 
+-- Temporary table should have 54+72485 rows now
+SELECT COUNT(*)=54+72485 FROM "_agload_test_graph_ag_vertex_ids";
+ ?column? 
+----------
+ t
+(1 row)
+
+-- Sequence should be equal to max entry id i.e. 146941
+SELECT currval('agload_test_graph."City_id_seq"')=146941;
+ ?column? 
+----------
+ t
+(1 row)
+
+-- Should error out on loading the same file again due to duplicate id
+SELECT load_labels_from_file('agload_test_graph', 'City',
+    'age_load/cities.csv', true);
+ERROR:  Cannot insert duplicate vertex id: 1125899906842777
+HINT:  Entry id 153 is already used
+--
+-- Load edges -- Connects cities to countries
+--
+-- Should error out for using vertex label
+SELECT load_edges_from_file('agload_test_graph', 'Country',
+    'age_load/edges.csv');
+ERROR:  label "Country" already exists as edge label
 SELECT create_elabel('agload_test_graph','has_city');
 NOTICE:  ELabel "has_city" has been created
  create_elabel 
@@ -68,6 +114,17 @@ SELECT load_edges_from_file('agload_test_graph', 'has_city',
 
 (1 row)
 
+-- Sequence should be equal to number of edges loaded i.e. 72485
+SELECT currval('agload_test_graph."has_city_id_seq"')=72485;
+ ?column? 
+----------
+ t
+(1 row)
+
+-- Should error out for using edge label
+SELECT load_labels_from_file('agload_test_graph', 'has_city',
+     'age_load/cities.csv');
+ERROR:  label "has_city" already exists as vertex label
 SELECT table_catalog, table_schema, lower(table_name) as table_name, table_type
 FROM information_schema.tables
 WHERE table_schema = 'agload_test_graph' ORDER BY table_name ASC;
@@ -83,7 +140,7 @@ WHERE table_schema = 'agload_test_graph' ORDER BY table_name ASC;
 SELECT COUNT(*) FROM agload_test_graph."Country";
  count 
 -------
-    53
+    54
 (1 row)
 
 SELECT COUNT(*) FROM agload_test_graph."City";
@@ -101,7 +158,7 @@ SELECT COUNT(*) FROM agload_test_graph."has_city";
 SELECT COUNT(*) FROM cypher('agload_test_graph', $$MATCH(n) RETURN n$$) as (n agtype);
  count 
 -------
- 72538
+ 72539
 (1 row)
 
 SELECT COUNT(*) FROM cypher('agload_test_graph', $$MATCH (a)-[e]->(b) RETURN e$$) as (n agtype);
@@ -110,6 +167,17 @@ SELECT COUNT(*) FROM cypher('agload_test_graph', $$MATCH (a)-[e]->(b) RETURN e$$
  72485
 (1 row)
 
+--
+-- Load countries and cities without id
+--
+-- Should load countries in Country label without error since it should use sequence now
+SELECT load_labels_from_file('agload_test_graph', 'Country',
+    'age_load/countries.csv', false);
+ load_labels_from_file 
+-----------------------
+
+(1 row)
+
 SELECT create_vlabel('agload_test_graph','Country2');
 NOTICE:  VLabel "Country2" has been created
  create_vlabel 
@@ -153,6 +221,7 @@ SELECT COUNT(*) FROM agload_test_graph."City2";
 SELECT id FROM agload_test_graph."Country" LIMIT 10;
        id        
 -----------------
+ 844424930131969
  844424930131970
  844424930131971
  844424930131974
@@ -162,7 +231,6 @@ SELECT id FROM agload_test_graph."Country" LIMIT 10;
  844424930131996
  844424930132002
  844424930132023
- 844424930132025
 (10 rows)
 
 SELECT id FROM agload_test_graph."Country2" LIMIT 10;
@@ -180,42 +248,57 @@ SELECT id FROM agload_test_graph."Country2" LIMIT 10;
  1688849860263946
 (10 rows)
 
+-- Should return 2 rows for Country with same properties, but different ids
 SELECT * FROM cypher('agload_test_graph', $$MATCH(n:Country {iso2 : 'BE'})
     RETURN id(n), n.name, n.iso2 $$) as ("id(n)" agtype, "n.name" agtype, "n.iso2" agtype);
       id(n)      |  n.name   | n.iso2 
 -----------------+-----------+--------
  844424930131990 | "Belgium" | "BE"
-(1 row)
+ 844424930132223 | "Belgium" | "BE"
+(2 rows)
 
+-- Should return 1 row
 SELECT * FROM cypher('agload_test_graph', $$MATCH(n:Country2 {iso2 : 'BE'})
     RETURN id(n), n.name, n.iso2 $$) as ("id(n)" agtype, "n.name" agtype, "n.iso2" agtype);
       id(n)       |  n.name   | n.iso2 
 ------------------+-----------+--------
  1688849860263942 | "Belgium" | "BE"
 (1 row)
 
+-- Should return 2 rows for Country with same properties, but different ids
 SELECT * FROM cypher('agload_test_graph', $$MATCH(n:Country {iso2 : 'AT'})
     RETURN id(n), n.name, n.iso2 $$) as ("id(n)" agtype, "n.name" agtype, "n.iso2" agtype);
       id(n)      |  n.name   | n.iso2 
 -----------------+-----------+--------
  844424930131983 | "Austria" | "AT"
-(1 row)
+ 844424930132221 | "Austria" | "AT"
+(2 rows)
 
+-- Should return 1 row
 SELECT * FROM cypher('agload_test_graph', $$MATCH(n:Country2 {iso2 : 'AT'})
     RETURN id(n), n.name, n.iso2 $$) as ("id(n)" agtype, "n.name" agtype, "n.iso2" agtype);
       id(n)       |  n.name   | n.iso2 
 ------------------+-----------+--------
  1688849860263940 | "Austria" | "AT"
 (1 row)
 
+-- Should return 2 rows for Country with same properties, but different ids
 SELECT * FROM cypher('agload_test_graph', $$
     MATCH (u:Country {region : "Europe"})
     WHERE u.name =~ 'Cro.*'
-    RETURN u.name, u.region
-$$) AS (result_1 agtype, result_2 agtype);
- result_1  | result_2 
------------+----------
- "Croatia" | "Europe"
+    RETURN id(u), u.name, u.region
+$$) AS ("id(u)" agtype, result_1 agtype, result_2 agtype);
+      id(u)      | result_1  | result_2 
+-----------------+-----------+----------
+ 844424930132023 | "Croatia" | "Europe"
+ 844424930132226 | "Croatia" | "Europe"
+(2 rows)
+
+-- There shouldn't be any duplicates
+SELECT * FROM cypher('agload_test_graph', $$return graph_stats('agload_test_graph')$$) as (a agtype);
+                                            a                                             
+------------------------------------------------------------------------------------------
+ {"graph": "agload_test_graph", "num_loaded_edges": 72485, "num_loaded_vertices": 145130}
 (1 row)
 
 SELECT drop_graph('agload_test_graph', true);
@@ -236,22 +319,11 @@ NOTICE:  graph "agload_test_graph" has been dropped
 --
 -- Test property type conversion
 --
-SELECT create_graph('agload_conversion');
-NOTICE:  graph "agload_conversion" has been created
- create_graph 
---------------
-
-(1 row)
-
 -- vertex: load as agtype
-SELECT create_vlabel('agload_conversion','Person1');
-NOTICE:  VLabel "Person1" has been created
- create_vlabel 
----------------
-
-(1 row)
-
+-- Should create graph and label automatically
 SELECT load_labels_from_file('agload_conversion', 'Person1', 'age_load/conversion_vertices.csv', true, true);
+NOTICE:  graph "agload_conversion" has been created
+NOTICE:  VLabel "Person1" has been created
  load_labels_from_file 
 -----------------------
 

diff --git a/regress/sql/age_load.sql b/regress/sql/age_load.sql
@@ -22,20 +22,65 @@
 LOAD 'age';
 
 SET search_path TO ag_catalog;
+
+-- Create a country using CREATE clause
 SELECT create_graph('agload_test_graph');
 
-SELECT create_vlabel('agload_test_graph','Country');
+SELECT * FROM cypher('agload_test_graph', $$CREATE (n:Country {__id__:1}) RETURN n$$) as (n agtype);
+
+--
+-- Load countries with id
+--
+SELECT load_labels_from_file('agload_test_graph', 'Country',
+    'age_load/countries.csv', true);
+
+-- A temporary table should have been created with 54 ids; 1 from CREATE and 53 from file
+SELECT COUNT(*)=54 FROM "_agload_test_graph_ag_vertex_ids";
+
+-- Sequence should be equal to max entry id i.e. 248
+SELECT currval('agload_test_graph."Country_id_seq"')=248;
+
+-- Should error out on loading the same file again due to duplicate id
 SELECT load_labels_from_file('agload_test_graph', 'Country',
-    'age_load/countries.csv');
+    'age_load/countries.csv', true);
+
+--
+-- Load cities with id
+--
+
+-- Should create City label automatically and load cities
+SELECT load_labels_from_file('agload_test_graph', 'City',
+    'age_load/cities.csv', true);
+
+-- Temporary table should have 54+72485 rows now
+SELECT COUNT(*)=54+72485 FROM "_agload_test_graph_ag_vertex_ids";
 
-SELECT create_vlabel('agload_test_graph','City');
+-- Sequence should be equal to max entry id i.e. 146941
+SELECT currval('agload_test_graph."City_id_seq"')=146941;
+
+-- Should error out on loading the same file again due to duplicate id
 SELECT load_labels_from_file('agload_test_graph', 'City',
-    'age_load/cities.csv');
+    'age_load/cities.csv', true);
+
+--
+-- Load edges -- Connects cities to countries
+--
+
+-- Should error out for using vertex label
+SELECT load_edges_from_file('agload_test_graph', 'Country',
+    'age_load/edges.csv');
 
 SELECT create_elabel('agload_test_graph','has_city');
 SELECT load_edges_from_file('agload_test_graph', 'has_city',
      'age_load/edges.csv');
 
+-- Sequence should be equal to number of edges loaded i.e. 72485
+SELECT currval('agload_test_graph."has_city_id_seq"')=72485;
+
+-- Should error out for using edge label
+SELECT load_labels_from_file('agload_test_graph', 'has_city',
+     'age_load/cities.csv');
+
 SELECT table_catalog, table_schema, lower(table_name) as table_name, table_type
 FROM information_schema.tables
 WHERE table_schema = 'agload_test_graph' ORDER BY table_name ASC;
@@ -48,6 +93,14 @@ SELECT COUNT(*) FROM cypher('agload_test_graph', $$MATCH(n) RETURN n$$) as (n ag
 
 SELECT COUNT(*) FROM cypher('agload_test_graph', $$MATCH (a)-[e]->(b) RETURN e$$) as (n agtype);
 
+--
+-- Load countries and cities without id
+--
+
+-- Should load countries in Country label without error since it should use sequence now
+SELECT load_labels_from_file('agload_test_graph', 'Country',
+    'age_load/countries.csv', false);
+
 SELECT create_vlabel('agload_test_graph','Country2');
 SELECT load_labels_from_file('agload_test_graph', 'Country2',
                              'age_load/countries.csv', false);
@@ -62,31 +115,39 @@ SELECT COUNT(*) FROM agload_test_graph."City2";
 SELECT id FROM agload_test_graph."Country" LIMIT 10;
 SELECT id FROM agload_test_graph."Country2" LIMIT 10;
 
+-- Should return 2 rows for Country with same properties, but different ids
 SELECT * FROM cypher('agload_test_graph', $$MATCH(n:Country {iso2 : 'BE'})
     RETURN id(n), n.name, n.iso2 $$) as ("id(n)" agtype, "n.name" agtype, "n.iso2" agtype);
+-- Should return 1 row
 SELECT * FROM cypher('agload_test_graph', $$MATCH(n:Country2 {iso2 : 'BE'})
     RETURN id(n), n.name, n.iso2 $$) as ("id(n)" agtype, "n.name" agtype, "n.iso2" agtype);
 
+-- Should return 2 rows for Country with same properties, but different ids
 SELECT * FROM cypher('agload_test_graph', $$MATCH(n:Country {iso2 : 'AT'})
     RETURN id(n), n.name, n.iso2 $$) as ("id(n)" agtype, "n.name" agtype, "n.iso2" agtype);
+-- Should return 1 row
 SELECT * FROM cypher('agload_test_graph', $$MATCH(n:Country2 {iso2 : 'AT'})
     RETURN id(n), n.name, n.iso2 $$) as ("id(n)" agtype, "n.name" agtype, "n.iso2" agtype);
 
+-- Should return 2 rows for Country with same properties, but different ids
 SELECT * FROM cypher('agload_test_graph', $$
     MATCH (u:Country {region : "Europe"})
     WHERE u.name =~ 'Cro.*'
-    RETURN u.name, u.region
-$$) AS (result_1 agtype, result_2 agtype);
+    RETURN id(u), u.name, u.region
+$$) AS ("id(u)" agtype, result_1 agtype, result_2 agtype);
+
+-- There shouldn't be any duplicates
+SELECT * FROM cypher('agload_test_graph', $$return graph_stats('agload_test_graph')$$) as (a agtype);
 
 SELECT drop_graph('agload_test_graph', true);
 
 --
 -- Test property type conversion
 --
-SELECT create_graph('agload_conversion');
 
 -- vertex: load as agtype
-SELECT create_vlabel('agload_conversion','Person1');
+
+-- Should create graph and label automatically
 SELECT load_labels_from_file('agload_conversion', 'Person1', 'age_load/conversion_vertices.csv', true, true);
 SELECT * FROM cypher('agload_conversion', $$ MATCH (n:Person1) RETURN properties(n) $$) as (a agtype);