tidb/parser/docs/quickstart.md

Quickstart

This parser is highly compatible with MySQL syntax. You can use it as a library, parse a text SQL into an AST tree, and traverse the AST nodes.

In this example, you will build a project, which can extract all the column names from a text SQL.

Prerequisites

• Golang version 1.13 or above. You can follow the instructions in the official installation page (check it by go version)

Create a Project

mkdir colx && cd colx
go mod init colx && touch main.go

Import Dependencies

First, you need to use go get to fetch the dependencies through git hash. The git hashes are available in release page. Take v5.3.0 as an example:

go get -v github.com/pingcap/tidb/parser@4a1b2e9

Note

The parser was merged into TiDB repo since v5.3.0. So you can only choose version v5.3.0 or higher in this TiDB repo.

You may want to use advanced API on expressions (a kind of AST node), such as numbers, string literals, booleans, nulls, etc. It is strongly recommended using the types package in TiDB repo with the following command:

go get -v github.com/pingcap/tidb/types/parser_driver@4a1b2e9

and import it in your golang source code:

import _ "github.com/pingcap/tidb/types/parser_driver"

Your directory should contain the following three files:

.
├── go.mod
├── go.sum
└── main.go

Now, open main.go with your favorite editor, and start coding!

Parse SQL text

To convert a SQL text to an AST tree, you need to:

1. Use the parser.New() function to instantiate a parser, and
2. Invoke the method Parse(sql, charset, collation) on the parser.

package main

import (
	"fmt"

	"github.com/pingcap/tidb/parser"
	"github.com/pingcap/tidb/parser/ast"
	_ "github.com/pingcap/tidb/parser/test_driver"
)

func parse(sql string) (*ast.StmtNode, error) {
	p := parser.New()

	stmtNodes, _, err := p.Parse(sql, "", "")
	if err != nil {
		return nil, err
	}

	return &stmtNodes[0], nil
}

func main() {
	astNode, err := parse("SELECT a, b FROM t")
	if err != nil {
		fmt.Printf("parse error: %v\n", err.Error())
		return
	}
	fmt.Printf("%v\n", *astNode)
}

Test the parser by running the following command:

go run main.go

If the parser runs properly, you should get a result like this:

&{{{{SELECT a, b FROM t}}} {[]} 0xc0000a1980 false 0xc00000e7a0 <nil> 0xc0000a19b0 <nil> <nil> [] <nil> <nil> none [] false false 0 <nil>}

Note

Here are a few things you might want to know:

• To use a parser, a parser_driver is required. It decides how to parse the basic data types in SQL.

  You can use github.com/pingcap/tidb/parser/test_driver as the parser_driver for test. Again, if you need advanced features, please use the parser_driver in TiDB (run go get -v github.com/pingcap/tidb/types/parser_driver@4a1b2e9 and import it).

• The instantiated parser object is not goroutine safe. It is better to keep it in a single goroutine.

• The instantiated parser object is not lightweight. It is better to reuse it if possible.

• Warning: the 'parser.result' object is being reused without being properly reset or copied. This can cause unexpected behavior or errors if the object is used for multiple parsing operations or concurrently in multiple goroutines. To avoid these issues, make a copy of the 'parser.result' object before calling 'parser.Parse()' again or before using it in another goroutine, or create a new 'parser' object altogether for each new parsing operation.

• The 2nd and 3rd arguments of parser.Parse() are charset and collation respectively. If you pass an empty string into it, a default value is chosen.

Traverse AST Nodes

Now you get the AST tree root of a SQL statement. It is time to extract the column names by traverse.

Parser implements the interface ast.Node for each kind of AST node, such as SelectStmt, TableName, ColumnName. ast.Node provides a method Accept(v Visitor) (node Node, ok bool) to allow any struct that has implemented ast.Visitor to traverse itself.

ast.Visitor is defined as follows:

type Visitor interface {
	Enter(n Node) (node Node, skipChildren bool)
	Leave(n Node) (node Node, ok bool)
}

Now you can define your own visitor, colX(columnExtractor):

type colX struct{
	colNames []string
}

func (v *colX) Enter(in ast.Node) (ast.Node, bool) {
	if name, ok := in.(*ast.ColumnName); ok {
		v.colNames = append(v.colNames, name.Name.O)
	}
	return in, false
}

func (v *colX) Leave(in ast.Node) (ast.Node, bool) {
	return in, true
}

Finally, wrap colX in a simple function:

func extract(rootNode *ast.StmtNode) []string {
	v := &colX{}
	(*rootNode).Accept(v)
	return v.colNames
}

And slightly modify the main function:

func main() {
	if len(os.Args) != 2 {
		fmt.Println("usage: colx 'SQL statement'")
		return
	}
	sql := os.Args[1]
	astNode, err := parse(sql)
	if err != nil {
		fmt.Printf("parse error: %v\n", err.Error())
		return
	}
	fmt.Printf("%v\n", extract(astNode))
}

Test your program:

go build && ./colx 'select a, b from t'

[a b]

You can also try a different SQL statement as an input. For example:

$ ./colx 'SELECT a, b FROM t GROUP BY (a, b) HAVING a > c ORDER BY b'
[a b a b a c b]

If necessary, you can deduplicate by yourself.

$ ./colx 'SELECT a, b FROM t/invalid_str'
parse error: line 1 column 19 near "/invalid_str"

Enjoy!