PGQL implementation on TiDB

[sleepymole]

Motivation

This PR aims to implement all PGQL features except graph modification on TiDB.

Status

• Graph DDL
  • Parser
  • DDL Executor
• Graph Show
  • SHOW GRAPHS
  • SHOW CREATE PROPERTY GRAPH `xxx`
• Graph Query
  • Parser
  • Simple Path Matching (Complete Subgraph Matching)
  • Variable-Length Paths
    • Any Path
    • Shortest Path
    • Cheapest Path
  • Subquery Compatible
  • Path Pattern Macro
  • Vertex and Edge functions
    • ID
    • LABEL
    • LABELS (function)
    • HAS_LABEL
    • ALL_DIFFERENT
    • IN_DEGREE
    • OUT_DEGREE

Try it out now

Prerequisites

• Go (1.18+)

Run a standalone tidb-server from current branch.

git clone https://github.com/gozssky/tidb.git
cd tidb && git checkout graph
go run tidb-server/main.go --path=/tmp/tigraph

Create a property graph

Underlying tables

CREATE TABLE `Persons` (
  `id` bigint(20) NOT NULL,
  `name` varchar(255) DEFAULT NULL,
  `company_id` bigint(20) DEFAULT NULL,
  PRIMARY KEY (`id`) /*T![clustered_index] CLUSTERED */
);
INSERT INTO Persons VALUES (1, 'Nikita', NULL);
INSERT INTO Persons VALUES (2, 'Camille', 1);
INSERT INTO Persons VALUES (3, 'Liam', NULL);

CREATE TABLE `Companies` (
  `id` bigint(20) NOT NULL,
  `name` varchar(255) DEFAULT NULL,
  PRIMARY KEY (`id`) /*T![clustered_index] CLUSTERED */
);
INSERT INTO Companies VALUES (1, 'Oracle');

CREATE TABLE `Accounts` (
  `number` bigint(20) NOT NULL,
  `account_type` int(11) DEFAULT NULL,
  `person_id` bigint(20) DEFAULT NULL,
  `company_id` bigint(20) DEFAULT NULL,
  PRIMARY KEY (`number`) /*T![clustered_index] CLUSTERED */
);
INSERT INTO Accounts VALUES (10039, 100, 2, NULL);
INSERT INTO Accounts VALUES (2090, 100, 3, NULL);
INSERT INTO Accounts VALUES (8021, 100, 1, NULL);
INSERT INTO Accounts VALUES (1001, 200, NULL, 1);

CREATE TABLE `Transactions` (
  `id` bigint(20) NOT NULL,
  `from_account` bigint(20) DEFAULT NULL,
  `to_account` bigint(20) DEFAULT NULL,
  `date` datetime DEFAULT NULL,
  `amount` double DEFAULT NULL,
  PRIMARY KEY (`id`) /*T![clustered_index] CLUSTERED */
);
INSERT INTO Transactions VALUES (1, 2090, 10039, '2022-03-12 14:53:32', 9900.00);
INSERT INTO Transactions VALUES (2, 10039, 8021, '2022-03-12 14:56:05', 1000.00);
INSERT INTO Transactions VALUES (3, 1001, 2090, '2022-03-12 14:55:46', 9999.50);
INSERT INTO Transactions VALUES (4, 8021, 1001, '2022-03-12 14:53:14', 1500.30);
INSERT INTO Transactions VALUES (5, 8021, 1001, '2022-03-12 14:54:05', 3000.70);

Graph financial_transactions

CREATE PROPERTY GRAPH financial_transactions
  VERTEX TABLES (
    Persons KEY ( id ) LABEL Person PROPERTIES ( name ),
    Companies KEY ( id ) LABEL Company PROPERTIES ( name ),
    Accounts KEY ( number ) LABEL Account PROPERTIES ( number )
  )
  EDGE TABLES (
    Transactions KEY ( from_account, to_account, date )
      SOURCE KEY ( from_account ) REFERENCES Accounts
      DESTINATION KEY ( to_account ) REFERENCES Accounts
      LABEL transaction PROPERTIES ( amount ),
    Accounts AS PersonOwner KEY ( number )
      SOURCE KEY ( number ) REFERENCES Accounts
      DESTINATION KEY ( person_id ) REFERENCES Persons
      LABEL owner NO PROPERTIES,
    Accounts AS CompanyOwner KEY ( number )
      SOURCE KEY ( number ) REFERENCES Accounts
      DESTINATION KEY ( company_id ) REFERENCES Companies
      LABEL owner NO PROPERTIES,
    Persons AS worksFor KEY ( id )
      SOURCE KEY ( id ) REFERENCES Persons
      DESTINATION KEY ( company_id ) REFERENCES Companies
      NO PROPERTIES
  );

Graph query

An example in spec.

SELECT label(owner),
       COUNT(*) AS numTransactions,
       SUM(out.amount) AS totalOutgoing,
       GROUP_CONCAT(out.amount) AS amounts
FROM MATCH (a:Account) -[:owner]-> (owner:Person|Company) ON financial_transactions
     , MATCH (a) -[out:transaction]-> (:Account) ON financial_transactions
GROUP BY label(owner)
ORDER BY label(owner);

Output

+--------------+-----------------+---------------+-------------------------+
| label(owner) | numTransactions | totalOutgoing | amounts                 |
+--------------+-----------------+---------------+-------------------------+
| Company      |               1 |        9999.5 | 9999.5                  |
| Person       |               4 |         15401 | 9900,3000.7,1500.3,1000 |
+--------------+-----------------+---------------+-------------------------+
2 rows in set (0.00 sec)