Google Cloud BigQuery

Google BigQuery enables super-fast, SQL-like queries against massive datasets, using the processing power of Google's infrastructure. To learn more, read What is BigQuery?.

The goal of google-cloud is to provide an API that is comfortable to Rubyists. Your authentication credentials are detected automatically in Google Cloud Platform (GCP), including Google Compute Engine (GCE), Google Kubernetes Engine (GKE), Google App Engine (GAE), Google Cloud Functions (GCF) and Cloud Run. In other environments you can configure authentication easily, either directly in your code or via environment variables. Read more about the options for connecting in the Authentication Guide.

To help you get started quickly, the first few examples below use a public dataset provided by Google. As soon as you have signed up to use BigQuery, and provided that you stay in the free tier for queries, you should be able to run these first examples without the need to set up billing or to load data (although we'll show you how to do that too.)

Listing Datasets and Tables

A BigQuery project contains datasets, which in turn contain tables. Assuming that you have not yet created datasets or tables in your own project, let's connect to Google's bigquery-public-data project, and see what we find.

require "google/cloud/bigquery"

bigquery = Google::Cloud::Bigquery.new project: "bigquery-public-data"

bigquery.datasets.count #=> 1
bigquery.datasets.first.dataset_id #=> "samples"

dataset = bigquery.datasets.first
tables = dataset.tables

tables.count #=> 7
tables.map &:table_id #=> [..., "shakespeare", "trigrams", "wikipedia"]

In addition to listing all datasets and tables in the project, you can also retrieve individual datasets and tables by ID. Let's look at the structure of the shakespeare table, which contains an entry for every word in every play written by Shakespeare.

require "google/cloud/bigquery"

bigquery = Google::Cloud::Bigquery.new project: "bigquery-public-data"

dataset = bigquery.dataset "samples"
table = dataset.table "shakespeare"

table.headers #=> [:word, :word_count, :corpus, :corpus_date]
table.rows_count #=> 164656

Now that you know the column names for the Shakespeare table, let's write and run a few queries against it.

Running queries

BigQuery supports two SQL dialects: standard SQL and the older legacy SQl (BigQuery SQL), as discussed in the guide Migrating from legacy SQL.

Standard SQL

Standard SQL is the preferred SQL dialect for querying data stored in BigQuery. It is compliant with the SQL 2011 standard, and has extensions that support querying nested and repeated data. This is the default syntax. It has several advantages over legacy SQL, including:

  • Composability using WITH clauses and SQL functions
  • Subqueries in the SELECT list and WHERE clause
  • Correlated subqueries
  • ARRAY and STRUCT data types
  • Inserts, updates, and deletes
  • COUNT(DISTINCT <expr>) is exact and scalable, providing the accuracy of EXACT_COUNT_DISTINCT without its limitations
  • Automatic predicate push-down through JOINs
  • Complex JOIN predicates, including arbitrary expressions

For examples that demonstrate some of these features, see Standard SQL ghlights.

As shown in this example, standard SQL is the library default:

require "google/cloud/bigquery"

bigquery = Google::Cloud::Bigquery.new

sql = "SELECT word, SUM(word_count) AS word_count " \
      "FROM `bigquery-public-data.samples.shakespeare`" \
      "WHERE word IN ('me', 'I', 'you') GROUP BY word"
data = bigquery.query sql

Notice that in standard SQL, a fully-qualified table name uses the following format: my-dashed-project.dataset1.tableName.

Legacy SQL (formerly BigQuery SQL)

Before version 2.0, BigQuery executed queries using a non-standard SQL dialect known as BigQuery SQL. This variant is optional, and can be enabled by passing the flag legacy_sql: true with your query. (If you get an SQL syntax error with a query that may be written in legacy SQL, be sure that you are passing this option.)

To use legacy SQL, pass the option legacy_sql: true with your query:

require "google/cloud/bigquery"

bigquery = Google::Cloud::Bigquery.new

sql = "SELECT TOP(word, 50) as word, COUNT(*) as count " \
      "FROM [bigquery-public-data:samples.shakespeare]"
data = bigquery.query sql, legacy_sql: true

Notice that in legacy SQL, a fully-qualified table name uses brackets instead of back-ticks, and a colon instead of a dot to separate the project and the dataset: [my-dashed-project:dataset1.tableName].

Query parameters

With standard SQL, you can use positional or named query parameters. This example shows the use of named parameters:

require "google/cloud/bigquery"

bigquery = Google::Cloud::Bigquery.new

sql = "SELECT word, SUM(word_count) AS word_count " \
      "FROM `bigquery-public-data.samples.shakespeare`" \
      "WHERE word IN UNNEST(@words) GROUP BY word"
data = bigquery.query sql, params: { words: ['me', 'I', 'you'] }

As demonstrated above, passing the params option will automatically set standard_sql to true.

Data types

BigQuery standard SQL supports simple data types such as integers, as well as more complex types such as ARRAY and STRUCT.

The BigQuery data types are converted to and from Ruby types as follows:

BigQueryRubyNotes
BOOLtrue/false
INT64Integer
FLOAT64Float
NUMERICBigDecimalBigDecimal values will be rounded to scale 9.
BIGNUMERICconverted to BigDecimalPass data as String and map query param values in types.
STRINGString
DATETIMEDateTimeDATETIME does not support time zone.
DATEDate
TIMESTAMPTime
TIMEGoogle::Cloud::BigQuery::Time
BYTESFile, IO, StringIO, or similar
ARRAYArrayNested arrays, nil values are not supported.
STRUCTHashHash keys may be strings or symbols.

See Data Types for an overview of each BigQuery data type, including allowed values.

Running Queries

Let's start with the simplest way to run a query. Notice that this time you are connecting using your own default project. It is necessary to have write access to the project for running a query, since queries need to create tables to hold results.

require "google/cloud/bigquery"

bigquery = Google::Cloud::Bigquery.new

sql = "SELECT APPROX_TOP_COUNT(corpus, 10) as title, " \
      "COUNT(*) as unique_words " \
      "FROM `bigquery-public-data.samples.shakespeare`"
data = bigquery.query sql

data.next? #=> false
data.first #=> {:title=>[{:value=>"hamlet", :count=>5318}, ...}

The APPROX_TOP_COUNT function shown above is just one of a variety of functions offered by BigQuery. See the Query Reference (standard SQL) for a full listing.

Query Jobs

It is usually best not to block for most BigQuery operations, including querying as well as importing, exporting, and copying data. Therefore, the BigQuery API provides facilities for managing longer-running jobs. With this approach, an instance of Google::Cloud::Bigquery::QueryJob is returned, rather than an instance of Google::Cloud::Bigquery::Data.

require "google/cloud/bigquery"

bigquery = Google::Cloud::Bigquery.new

sql = "SELECT APPROX_TOP_COUNT(corpus, 10) as title, " \
      "COUNT(*) as unique_words " \
      "FROM `bigquery-public-data.samples.shakespeare`"
job = bigquery.query_job sql

job.wait_until_done!
if !job.failed?
  job.data.first
  #=> {:title=>[{:value=>"hamlet", :count=>5318}, ...}
end

Once you have determined that the job is done and has not failed, you can obtain an instance of Google::Cloud::Bigquery::Data by calling data on the job instance. The query results for both of the above examples are stored in temporary tables with a lifetime of about 24 hours. See the final example below for a demonstration of how to store query results in a permanent table.

Creating Datasets and Tables

The first thing you need to do in a new BigQuery project is to create a Google::Cloud::Bigquery::Dataset. Datasets hold tables and control access to them.

require "google/cloud/bigquery"

bigquery = Google::Cloud::Bigquery.new

dataset = bigquery.create_dataset "my_dataset"

Now that you have a dataset, you can use it to create a table. Every table is defined by a schema that may contain nested and repeated fields. The example below shows a schema with a repeated record field named cities_lived. (For more information about nested and repeated fields, see Preparing Data for Loading.)

require "google/cloud/bigquery"

bigquery = Google::Cloud::Bigquery.new
dataset = bigquery.dataset "my_dataset"

table = dataset.create_table "people" do |schema|
  schema.string "first_name", mode: :required
  schema.record "cities_lived", mode: :repeated do |nested_schema|
    nested_schema.string "place", mode: :required
    nested_schema.integer "number_of_years", mode: :required
  end
end

Because of the repeated field in this schema, we cannot use the CSV format to load data into the table.

Loading records

To follow along with these examples, you will need to set up billing on the Google Developers Console.

In addition to CSV, data can be imported from files that are formatted as Newline-delimited JSON, Avro, ORC, Parquet or from a Google Cloud Datastore backup. It can also be "streamed" into BigQuery.

Streaming records

For situations in which you want new data to be available for querying as soon as possible, inserting individual records directly from your Ruby application is a great approach.

require "google/cloud/bigquery"

bigquery = Google::Cloud::Bigquery.new
dataset = bigquery.dataset "my_dataset"
table = dataset.table "people"

rows = [
    {
        "first_name" => "Anna",
        "cities_lived" => [
            {
                "place" => "Stockholm",
                "number_of_years" => 2
            }
        ]
    },
    {
        "first_name" => "Bob",
        "cities_lived" => [
            {
                "place" => "Seattle",
                "number_of_years" => 5
            },
            {
                "place" => "Austin",
                "number_of_years" => 6
            }
        ]
    }
]
table.insert rows

To avoid making RPCs (network requests) to retrieve the dataset and table resources when streaming records, pass the skip_lookup option. This creates local objects without verifying that the resources exist on the BigQuery service.

require "google/cloud/bigquery"

bigquery = Google::Cloud::Bigquery.new
dataset = bigquery.dataset "my_dataset", skip_lookup: true
table = dataset.table "people", skip_lookup: true

rows = [
    {
        "first_name" => "Anna",
        "cities_lived" => [
            {
                "place" => "Stockholm",
                "number_of_years" => 2
            }
        ]
    },
    {
        "first_name" => "Bob",
        "cities_lived" => [
            {
                "place" => "Seattle",
                "number_of_years" => 5
            },
            {
                "place" => "Austin",
                "number_of_years" => 6
            }
        ]
    }
]
table.insert rows

There are some trade-offs involved with streaming, so be sure to read the discussion of data consistency in Streaming Data Into BigQuery.

Uploading a file

To follow along with this example, please download the names.zip archive from the U.S. Social Security Administration. Inside the archive you will find over 100 files containing baby name records since the year 1880.

require "google/cloud/bigquery"

bigquery = Google::Cloud::Bigquery.new
dataset = bigquery.dataset "my_dataset"
table = dataset.create_table "baby_names" do |schema|
  schema.string "name", mode: :required
  schema.string "gender", mode: :required
  schema.integer "count", mode: :required
end

file = File.open "names/yob2014.txt"
table.load file, format: "csv"

Because the names data, although formatted as CSV, is distributed in files with a .txt extension, this example explicitly passes the format option in order to demonstrate how to handle such situations. Because CSV is the default format for load operations, the option is not actually necessary. For JSON saved with a .txt extension, however, it would be.

Exporting query results to Google Cloud Storage

The example below shows how to pass the table option with a query in order to store results in a permanent table. It also shows how to export the result data to a Google Cloud Storage file. In order to follow along, you will need to enable the Google Cloud Storage API in addition to setting up billing.

require "google/cloud/bigquery"

bigquery = Google::Cloud::Bigquery.new
dataset = bigquery.dataset "my_dataset"
source_table = dataset.table "baby_names"
result_table = dataset.create_table "baby_names_results"

sql = "SELECT name, count " \
      "FROM baby_names " \
      "WHERE gender = 'M' " \
      "ORDER BY count ASC LIMIT 5"
query_job = dataset.query_job sql, table: result_table

query_job.wait_until_done!

if !query_job.failed?
  require "google/cloud/storage"

  storage = Google::Cloud::Storage.new
  bucket_id = "bigquery-exports-#{SecureRandom.uuid}"
  bucket = storage.create_bucket bucket_id
  extract_url = "gs://#{bucket.id}/baby-names.csv"

  result_table.extract extract_url

  # Download to local filesystem
  bucket.files.first.download "baby-names.csv"
end

If a table you wish to export contains a large amount of data, you can pass a wildcard URI to export to multiple files (for sharding), or an array of URIs (for partitioning), or both. See Exporting Data for details.

Configuring retries and timeout

You can configure how many times API requests may be automatically retried. When an API request fails, the response will be inspected to see if the request meets criteria indicating that it may succeed on retry, such as 500 and 503 status codes or a specific internal error code such as rateLimitExceeded. If it meets the criteria, the request will be retried after a delay. If another error occurs, the delay will be increased before a subsequent attempt, until the retries limit is reached.

You can also set the request timeout value in seconds.

require "google/cloud/bigquery"

bigquery = Google::Cloud::Bigquery.new retries: 10, timeout: 120

See the BigQuery error table for a list of error conditions.

Additional information

Google BigQuery can be configured to use logging. To learn more, see the Logging guide.