/qlbridge

A golang [x]QL lex, parse and expression VM. Embed QL in streaming, or custom storage

Primary LanguageGoMIT LicenseMIT

QLBridge - a Go SQL Runtime Engine

This is a SQL execution engine to process data with sql for embedded use, includes a native go lexer, parser. Extend with native go functions.

QLBridge Features and Goals

  • execution of sql queries against your data, embedable, not coupled to storage layer
  • extend VM with custom go functions, provide rich basic library of functions
  • provide example backends (csv, elasticsearch, etc)

Example SQL Runtime for Reading a Csv via Stdio, File

See example in qlcsv folder for a CSV reader, parser, evaluation engine.

./qlcsv -sql 'select 
		user_id, email, item_count * 2, yy(reg_date) > 10 
	FROM stdio where email_is_valid(email);' < users.csv
func main() {

	// load the libray of pre-built functions for usage in sql queries
	builtins.LoadAllBuiltins()

	// Add a custom function to the VM to make available to SQL language
	// showing lexer/parser accepts it
	expr.FuncAdd("email_is_valid", EmailIsValid)

	// Datasources are easy to write and can be added
	datasource.Register("csv", &datasource.CsvDataSource{})

	// now from here down is standard go database/sql query handling
	db, err := sql.Open("qlbridge", "csv:///dev/stdin")
	if err != nil {
		panic(err.Error())
	}
	defer db.Close()

	rows, err := db.Query(sqlText)
	if err != nil {
		panic(err.Error())
	}
	defer rows.Close()
	cols, _ := rows.Columns()

	// this is just stupid hijinx for getting pointers for unknown len columns
	readCols := make([]interface{}, len(cols))
	writeCols := make([]string, len(cols))
	for i, _ := range writeCols {
		readCols[i] = &writeCols[i]
	}

	for rows.Next() {
		rows.Scan(readCols...)
		fmt.Println(strings.Join(writeCols, ", "))
	}
}

// Example of a custom Function, that we are adding into the Expression VM
//
//         select
//              user_id AS theuserid, email, item_count * 2, reg_date
//         FROM stdio
//         WHERE email_is_valid(email)
func EmailIsValid(ctx vm.EvalContext, email value.Value) (value.BoolValue, bool) {
	emailstr, ok := value.ToString(email.Rv())
	if !ok || emailstr == "" {
		return value.BoolValueFalse, true
	}
	if _, err := mail.ParseAddress(emailstr); err == nil {
		return value.BoolValueTrue, true
	}

	return value.BoolValueFalse, true
}

[x]QL languages are making a comeback. It is still an easy, approachable way of working with data. Also, we see more and more ql's that are xql'ish but un-apologetically non-standard. This matches our observation that data is stored in more and more formats in more tools, services that aren't traditional db's but querying that data should still be easy. Examples Influx, GitQL, Presto, Hive, CQL, yql, ql.io, etc

Projects that access non-sql data via [x]ql

Projects that value-add at proxy

Inspiration/Other works

Go Script/VM interpreters

Creating a custom Lexer/Parser

See example in exampledialect folder for a custom ql dialect, this example creates a mythical SUBSCRIBETO query language...

// Tokens Specific to our PUBSUB
var TokenSubscribeTo lex.TokenType = 1000

// Custom lexer for our maybe hash function
func LexMaybe(l *ql.Lexer) ql.StateFn {

	l.SkipWhiteSpaces()

	keyWord := strings.ToLower(l.PeekWord())

	switch keyWord {
	case "maybe":
		l.ConsumeWord("maybe")
		l.Emit(lex.TokenIdentity)
		return ql.LexExpressionOrIdentity
	}
	return ql.LexExpressionOrIdentity
}

func main() {

	// We are going to inject new tokens into qlbridge
	lex.TokenNameMap[TokenSubscribeTo] = &lex.TokenInfo{Description: "subscribeto"}

	// OverRide the Identity Characters in qlbridge to allow a dash in identity
	ql.IDENTITY_CHARS = "_./-"

	ql.LoadTokenInfo()
	ourDialect.Init()

	// We are going to create our own Dialect that uses a "SUBSCRIBETO" keyword
	pubsub = &ql.Statement{TokenSubscribeTo, []*ql.Clause{
		{Token: TokenSubscribeTo, Lexer: ql.LexColumns},
		{Token: lex.TokenFrom, Lexer: LexMaybe},
		{Token: lex.TokenWhere, Lexer: ql.LexColumns, Optional: true},
	}}
	ourDialect = &ql.Dialect{
		"Subscribe To", []*ql.Statement{pubsub},
	}

	l := ql.NewLexer(`
			SUBSCRIBETO
				count(x), Name
			FROM ourstream
			WHERE 
				k = REPLACE(LOWER(Name),'cde','xxx');
		`, ourDialect)

}