QLBridge - Go SQL Runtime Engine
A SQL execution engine for embedded use as a library for SQL or SQL-Like functionality. Hackable, add datasources ("Storage" can be rest apis, or anything), and add functions. See usage in https://github.com/dataux/dataux a federated Sql Engine mysql-compatible with backends (Elasticsearch, Google-Datastore, Mongo, Cassandra, Files).
QLBridge Features and Goals
- expression engine for evaluation of single expressions
- 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)
Dialects
- SQL see examples
- FilterQL (just Where clause) with more of a DSL for filter see examples
- Simple Expressions see examples
Example of Expression Evaluation Engine
These expressions can be used stand-alone embedded usage in your app. But, are the same expressions which might be columns, where, group-by clauses in SQL. see example
func main() {
// Add a custom function to the VM to make available to expression language
expr.FuncAdd("email_is_valid", &EmailIsValid{})
// This is the evaluation context which will be the data-source
// to be evaluated against the expressions. There is a very simple
// interface you can use to create your own.
evalContext := datasource.NewContextSimpleNative(map[string]interface{}{
"int5": 5,
"str5": "5",
"created": dateparse.MustParse("12/18/2015"),
"bvalt": true,
"bvalf": false,
"user_id": "abc",
"urls": []string{"http://google.com", "http://nytimes.com"},
"hits": map[string]int64{"google.com": 5, "bing.com": 1},
"email": "[email protected]",
"emailbad": "bob",
"mt": map[string]time.Time{
"event0": dateparse.MustParse("12/18/2015"),
"event1": dateparse.MustParse("12/22/2015"),
},
})
// Example list of expressions
exprs := []string{
"int5 == 5",
`6 > 5`,
`6 > 5.5`,
`(4 + 5) / 2`,
`6 == (5 + 1)`,
`2 * (3 + 5)`,
`todate("12/12/2012")`,
`created > "now-1M"`, // Date math
`created > "now-10y"`,
`user_id == "abc"`,
`email_is_valid(email)`,
`email_is_valid(emailbad)`,
`email_is_valid("not_an_email")`,
`EXISTS int5`,
`!exists(user_id)`,
`mt.event0 > now()`, // step into child of maps
`["portland"] LIKE "*land"`,
`email contains "bob"`,
`email NOT contains "bob"`,
`[1,2,3] contains int5`,
`[1,2,3,5] NOT contains int5`,
`urls contains "http://google.com"`,
`split("chicago,portland",",") LIKE "*land"`,
`10 BETWEEN 1 AND 50`,
`15.5 BETWEEN 1 AND "55.5"`,
`created BETWEEN "now-50w" AND "12/18/2020"`,
`toint(not_a_field) NOT IN ("a","b" 4.5)`,
`
OR (
email != "[email protected]"
AND (
NOT EXISTS not_a_field
int5 == 5
)
)`,
}
for _, expression := range exprs {
// Same ast can be re-used safely concurrently
exprAst := expr.MustParse(expression)
// Evaluate AST in the vm
val, _ := vm.Eval(evalContext, exprAst)
v := val.Value()
u.Debugf("Output: %-35v T:%-15T expr: %s", v, v, expression)
}
}
// Example of a custom Function, that we are making available in the Expression VM
type EmailIsValid struct{}
func (m *EmailIsValid) Validate(n *expr.FuncNode) (expr.EvaluatorFunc, error) {
if len(n.Args) != 1 {
return nil, fmt.Errorf("Expected 1 arg for EmailIsValid(arg) but got %s", n)
}
return func(ctx expr.EvalContext, args []value.Value) (value.Value, bool) {
if args[0] == nil || args[0].Err() || args[0].Nil() {
return value.BoolValueFalse, true
}
if _, err := mail.ParseAddress(args[0].ToString()); err == nil {
return value.BoolValueTrue, true
}
return value.BoolValueFalse, true
}, nil
}
func (m *EmailIsValid) Type() value.ValueType { return value.BoolType }
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 stdin where email_is_valid(email);' < users.csv
func main() {
if sqlText == "" {
u.Errorf("You must provide a valid select query in argument: --sql=\"select ...\"")
return
}
// load all of our built-in functions
builtins.LoadAllBuiltins()
// Add a custom function to the VM to make available to SQL language
expr.FuncAdd("email_is_valid", &EmailIsValid{})
// We are registering the "csv" datasource, to show that
// the backend/sources can be easily created/added. This csv
// reader is an example datasource that is very, very simple.
exit := make(chan bool)
src, _ := datasource.NewCsvSource("stdin", 0, bytes.NewReader([]byte("##")), exit)
schema.RegisterSourceAsSchema("example_csv", src)
db, err := sql.Open("qlbridge", "example_csv")
if err != nil {
panic(err.Error())
}
defer db.Close()
rows, err := db.Query(sqlText)
if err != nil {
u.Errorf("could not execute query: %v", err)
return
}
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]
}
fmt.Printf("\n\nScanning through CSV: (%v)\n\n", strings.Join(cols, ","))
for rows.Next() {
rows.Scan(readCols...)
fmt.Println(strings.Join(writeCols, ", "))
}
fmt.Println("")
}
// 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 stdin
// WHERE email_is_valid(email)
type EmailIsValid struct{}
func (m *EmailIsValid) Validate(n *expr.FuncNode) (expr.EvaluatorFunc, error) {
if len(n.Args) != 1 {
return nil, fmt.Errorf("Expected 1 arg for EmailIsValid(arg) but got %s", n)
}
return func(ctx expr.EvalContext, args []value.Value) (value.Value, bool) {
if args[0] == nil || args[0].Err() || args[0].Nil() {
return value.BoolValueFalse, true
}
if _, err := mail.ParseAddress(args[0].ToString()); err == nil {
return value.BoolValueTrue, true
}
return value.BoolValueFalse, true
}, nil
}
func (m *EmailIsValid) Type() value.ValueType { return value.BoolType }
[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
- http://prestosql.io/
- https://crate.io/docs/current/sql/index.html
- http://senseidb.com/
- http://influxdb.com/docs/v0.8/api/query_language.html
- https://github.com/crosbymichael/dockersql
- http://harelba.github.io/q/
- https://github.com/dinedal/textql
- https://github.com/cloudson/gitql
- https://github.com/brendandburns/ksql