BigQuery SQL Driver

This library is compatible with Go 1.17+

Please refer to CHANGELOG.md if you encounter breaking changes.

• DSN
• Usage
• Benchmark
• License
• Credits and Acknowledgements
• Custom OAuth 2.0 authentication

This library provides fast implementation of the BigQuery Client as a database/sql drvier.

DSN Data Source Name

The BigQuery driver accepts the following DSN

• 'bigquery://projectID/[location/]datasetID?queryString'

  Where queryString can optionally configure the following option:

  • credURL: (url encoded) local location or URL supported by Scy
  • credKey: optional (url encoded) Scy secret manager key or key location
  • credID: Scy resource secret ID
  • credJSON: rawURL base64 encoded cred JSON (not recommended)
  • endpoint
  • userAgent
  • apiKey
  • quotaProject
  • scopes

Since this library uses Google Cloud API you can pass your credentials via GOOGLE_APPLICATION_CREDENTIALS environment variable.

Usage:

package main

import (
	"database/sql"
	"fmt"
	"context"
	"log"
	_ "github.com/exponentiateam/bigquery"
)

type Participant struct {
	Name   string
	Splits []float64
}

func main() {

	db, err := sql.Open("bigquery", "bigquery://myProjectID/mydatasetID")
	if err != nil {
		log.Fatal(err)
	}
	defer db.Close()

	SQL := `WITH races AS (
		  SELECT "800M" AS race,
		    [STRUCT("Ben" as name, [23.4, 26.3] as splits), 
		 	 STRUCT("Frank" as name, [23.4, 26.3] as splits)
			]
		       AS participants)
		SELECT
		  race,
		  participant
		FROM races r
		CROSS JOIN UNNEST(r.participants) as participant`

	rows, err := db.QueryContext(context.Background(), SQL)
	if err != nil {
		log.Fatal(err)
	}
	defer rows.Close()
	for rows.Next() {
		if err := rows.Err(); err != nil {
			log.Fatal(err)
		}
		var race string
		var participant Participant
		err = rows.Scan(&race, &participant)
		fmt.Printf("fetched: %v %+v\n", race, participant)
		if err != nil {
			log.Fatal(err)
		}
	}
}

Data Ingestion (Load/Stream)

This driver implements LOAD/STREAM operation with the following SQL:

Loading data

To load data register a reader for supported source format, followed by LOAD SQL.

LOAD 'Reader:<SOURCE_FORMAT>:<READER_ID>' DATA INTO TABLE myproject.mydataset.mytable

The following snippet register READER_ID

 err := reader.Register(readerID, myReader)

The following code loads CSV data

package mypkg

import (
	"context"
	"database/sql"
	"fmt"
	"github.com/exponentiateam/bigquery/reader"
	"log"
	"strings"
)

func ExampleOfCSVLoad() {
	projectID := "my-gcp-project"
	db, err := sql.Open("bigquery", fmt.Sprintf("bigquery://%v/test", projectID))
	if err != nil {
		log.Fatal(err)
	}
	readerID := "123456"
	csvReader := strings.NewReader("ID,Name,Desc\n1,Name 1,Test\n2,Name 2,Test 2")
	err = reader.Register(readerID, csvReader)
	if err != nil {
		log.Fatal(err)
	}
	defer reader.Unregister(readerID)
	SQL := fmt.Sprintf(`LOAD 'Reader:csv:%v' DATA INTO TABLE mytable`, readerID)
	result, err := db.ExecContext(context.TODO(), SQL)
	if err != nil {
		log.Fatal(err)
	}
	affected, _ := result.RowsAffected()
	fmt.Printf("loaded: %v rows", affected)
}

Custom OAuth 2.0 authentication

Sometimes you already have a Google OAuth 2.0 client (client-id / secret) and an access or refresh token that you would like the driver to use.
OAuth2Manager can serialise such an oauth2.Config together with the corresponding oauth2.Token and make them addressable via URLs that you embed in the DSN.

Step-by-step example

package main

import (
    "context"
    "database/sql"
    "encoding/json"
    "fmt"
    "log"
    "net/url"
    "time"

    _ "github.com/exponentiateam/bigquery"
    "github.com/exponentiateam/bigquery"

    "golang.org/x/oauth2"
)

func main() {
    projectID := "my-project"
    datasetID := "my_dataset"

    // 1. Build the OAuth2 client configuration. In real life you would usually
    //    load this from a secret manager or configuration file.
    // IMPORTANT: the endpoints **must** be Google's authorization server
    // endpoints; BigQuery only accepts Google-issued access tokens.
    oauthCfg := &oauth2.Config{
        ClientID:     "<CLIENT_ID>",
        ClientSecret: "<CLIENT_SECRET>",
        Endpoint: oauth2.Endpoint{
            AuthURL:  "https://accounts.google.com/o/oauth2/v2/auth",
            TokenURL: "https://oauth2.googleapis.com/token",
        },
        RedirectURL: "http://localhost/oauth2callback",
        Scopes:      []string{"https://www.googleapis.com/auth/bigquery"},
    }

    // 2. Provide a *valid* access or refresh token obtained beforehand.
    //    The token expiry _must_ be set so that the helper can decide when to
    //    refresh it.
    oauthToken := &oauth2.Token{
        AccessToken:  "<ACCESS_TOKEN>",
        RefreshToken: "<REFRESH_TOKEN>", // optional – only when you want automatic refresh
        TokenType:    "Bearer",
        Expiry:       time.Now().Add(time.Hour),
    }

    // 3. Serialise both config and token to in-memory URLs that can be used in
    //    the DSN. You may also store them in any other fs supported by Afero
    //    (e.g. Google Cloud Storage, local disk, S3, …).
    helper := bigquery.NewOAuth2Manager()
    ctx := context.Background()

    cfgURL, err := helper.WithConfigURL(ctx, oauthCfg)
    if err != nil {
        log.Fatalf("failed to persist oauth2 config: %v", err)
    }

    tokenURL, err := helper.WithTokenURL(ctx, oauthToken)
    if err != nil {
        log.Fatalf("failed to persist oauth2 token: %v", err)
    }

    // 4. Build the DSN using the two generated URLs. Be sure to URL-escape them
    //    because the DSN itself is an URL as well.
    dsn := fmt.Sprintf(
        "bigquery://%s/%s?oauth2ClientURL=%s&oauth2TokenURL=%s",
        projectID,
        datasetID,
        url.QueryEscape(cfgURL),
        url.QueryEscape(tokenURL),
    )

    db, err := sql.Open("bigquery", dsn)
    if err != nil {
        log.Fatal(err)
    }
    defer db.Close()

    // Now use the db as usual – the driver will pick the token source built
    // from your custom OAuth2 config and token and will automatically refresh
    // the access token when it expires.
    var version string
    if err := db.QueryRowContext(ctx, "SELECT version()" ).Scan(&version); err != nil {
        log.Fatal(err)
    }
    fmt.Printf("connected, BigQuery engine version: %s\n", version)

    // For demonstration purposes print the DSN with the embedded URLs so that
    // you can see what happens under the hood.
    if raw, _ := json.MarshalIndent(struct {
        DSN       string `json:"dsn"`
        ConfigURL string `json:"oauth2ClientURL"`
        TokenURL  string `json:"oauth2TokenURL"`
    }{dsn, cfgURL, tokenURL}, "", "  "); raw != nil {
        fmt.Printf("%s\n", raw)
    }
}

Important notes

1. OAuth2Manager uses an in-memory ( mem:// ) file-system backend by default.
   The URLs produced in the example therefore live only for the lifetime of the current process. If you want to persist them, provide your own afs.Service or store the JSON blobs yourself and pass their URLs in the DSN – the driver only requires that the URLs are reachable, not that they were created by OAuth2Manager.
2. The provided token must have the Expiry field set. When the expiry is reached the token will be refreshed automatically using the refresh token and the OAuth2 endpoint supplied in the config.
3. The oauth2.Config.Endpoint must point to Google's authorization server (https://accounts.google.com/o/oauth2/v2/auth and https://oauth2.googleapis.com/token). BigQuery accepts only Google-issued access tokens.

Loading application data

The following code loads CSV data

package mypkg

import (
	"context"
	"database/sql"
	"fmt"
	"github.com/exponentiateam/bigquery/reader"
	"github.com/viant/sqlx/io/load/reader/parquet"
	"github.com/google/uuid"
	"log"
)

type Record struct {
	ID     int     `parquet:"id,plain,optional"`
	Name   string  `parquet:"name,plain,optional"`
	Active bool    `parquet:"active,plain,optional"`
	Amount float64 `parquet:"amount,plain,optional"`
}

func ExampleOfAppDataLoad() {
	projectID := "my-gcp-project"
	db, err := sql.Open("bigquery", fmt.Sprintf("bigquery://%v/test", projectID))
	if err != nil {
		log.Fatal(err)
	}
	var records = []*Record{
		{
			ID:     1,
			Name:   "record 1",
			Amount: 12.2,
		},
		{
			ID:     2,
			Name:   "record 2",
			Amount: 12.3,
		},
	}
	readerID := uuid.New().String()
    parquetReader, err := parquet.NewReader(records)
    if err != nil {
        log.Fatal(err)
    }
	err = reader.Register(readerID, parquetReader)
	if err != nil {
		log.Fatal(err)
	}
	defer reader.Unregister(readerID)
	SQL := fmt.Sprintf(`LOAD 'Reader:parquet:%v' DATA INTO TABLE mytable`, readerID)
	result, err := db.ExecContext(context.TODO(), SQL)
	if err != nil {
		log.Fatal(err)
	}
	affected, _ := result.RowsAffected()
	fmt.Printf("loaded: %v rows", affected)
}

Load option control

To customize Load you can inline JobConfigurationLoad you can as JSON as a hint LOAD 'Reader:<SOURCE_FORMAT>:<READER_ID>' /*+ <HINT> +*/ DATA INTO TABLE mytable for example:

LOAD 'Reader:CSV:201F973D-9BAB-4E0A-880F-7830B876F210' /*+ {
    "AllowJaggedRows": true,
    "AllowQuotedNewlines":true
  } +*/  DATA INTO TABLE mytable

Benchmark

Benchmark runs 3 times the following queries:

• primitive types:

   SELECT state,gender,year,name,number 
   FROM `bigquery-public-data.usa_names.usa_1910_2013` LIMIT 10000

• complex type (repeated string/repeated record)

   SELECT  t.publication_number, t.inventor, t.assignee, t.description_localized 
   FROM `patents-public-data.patents.publications_202105` t ORDER BY 1 LIMIT 1000

In both case database/sql driver is faster and allocate way less memory than GCP BigQuery API Client

goos: darwin
goarch: amd64
pkg: github.com/exponentiateam/bigquery/bench
cpu: Intel(R) Core(TM) i9-9980HK CPU @ 2.40GHz
Benchmark_Primitive_GCPClient
database/gcp: primitive types 3.918388531s
Benchmark_Primitive_GCPClient-16               1    3918442974 ns/op    42145144 B/op      830647 allocs/op
Benchmark_Primitive_SQLDriver
database/sql: primitive types 2.880091637s
Benchmark_Primitive_SQLDriver-16               1    2880149491 ns/op    22301848 B/op      334547 allocs/op
Benchmark_Complex_GCPClient
database/gcp: structured types 3.303497894s
Benchmark_Complex_GCPClient-16               1    3303548761 ns/op    11551216 B/op      154660 allocs/op
Benchmark_Complex_SQLDriver
database/sql: structured types 2.690012577s
Benchmark_Complex_SQLDriver-16               1    2690056675 ns/op     6643176 B/op       71562 allocs/op
PASS

License

The source code is made available under the terms of the Apache License, Version 2, as stated in the file LICENSE.

Individual files may be made available under their own specific license, all compatible with Apache License, Version 2. Please see individual files for details.

Credits and Acknowledgements

Library Author: Adrian Witas