Development Task

Description

The SQL standard syntax for starting a transaction is START TRANSACTION, which both MySQL and PostgreSQL supports.
They also support syntax like START TRANSACTION + [modifier], where modifier can be READ WRITE, READ ONLY, etc.
For the ease of use, they also support use BEGIN instead of START TRANSACTION
However, they treat BEGIN + [modifier] differently. (like BEGIN READ WRITE)

In Postgres this is completely fine, BEGIN/BEGIN WORK/BEGIN TRANSACTION are all treated as aliases for START TRANSACTION

MySQL, on the other hand, does not allow modifier after BEGIN, a limitation which TiDB inherits.

In TiDB, the standard way to start new transaction via go code is

txn, err := db.Begin()

which calls underlying go-mysql-driver via go's database module, and return a transaction handle (the txn above).
After we switch the driver from lib/pq, because pq starts transaction by BEGIN READ WRITE when no modifier provided, it will cause parser to complain.

Previously in unit tests fixing, we have been solving the problem by skipping Begin()method and execute SQL statement START TRANSACTION; and execute COMMIT; at the end of query block. But since we are not using Begin(), we can't get txn handle. It will cause problem when multiple transactions are nested or intersected.

The ideal solution is for parser to support begin + modifier syntax.

References

PostgreSQL: Documentation: 9.1: BEGIN
MySQL :: MySQL 8.0 Reference Manual :: 13.3.1 START TRANSACTION, COMMIT, and ROLLBACK Statements
MySQL Difference between BEGIN and START TRANSACTION - Database Administrators Stack Exchange

Many systems are built on postgres, mssql, and db2. Their all have database, schema, table lalyer.

It is hard to do migration many applications to TiDB for it lack of the schema layer.

Does this project has timeline that support main feature of postgres that build a release version?

Development Task

MySQL can use single quotes(') or double quotes(") to represent values, but PgSQL can only use single quotes(') to represent values. PgSQL's double quotes(") represent system identifiers, such as table names or field names . MySQL can use back quotes(`) to indicate system identifiers, such as table names and field names. PgSQL also does not support it.

We need to implement the use of these marks in PgSQL，replace their usage in MySQL.

• Implement the use of double, single and back quotation marks in PgSQL

Description

Since TiDB is Mysql compliant database, we need to adjust DCParser yacc paser files to remove mysql keywords and add pg keywords, and then implement the keywords logic functions in TiDB4PG project.

The things you need to do are :

1. spot statements in parser.y file
2. remove mysql keywords of the statement
3. make sure build successfully
4. correct related unit tests
5. commit PR

Helpful reference:
Understand Delete Statement - Chinese

Task List

DCParser:

• Delete Statement:

1. remove mysql keywords
2. correct unit tests

• Select Statement Parser

1. remove mysql keywords
2. correct unit tests

• Insert Statement Parser

1. remove mysql keywords
2. correct unit tests

• Update Statement Parser

1. remove mysql keywords
2. correct unit tests

TiDB4PG:

Please check the issue
DigitalChinaOpenSource/TiDB-for-PostgreSQL#40