Loading Parquet data from Cloud Storage

This page provides an overview of loading Parquet data from Cloud Storage into BigQuery.

Parquet is an open source column-oriented data format that is widely used in the Apache Hadoop ecosystem.

When you load Parquet data from Cloud Storage, you can load the data into a new table or partition, or you can append to or overwrite an existing table or partition. When your data is loaded into BigQuery, it is converted into columnar format for Capacitor (BigQuery's storage format).

When you load data from Cloud Storage into a BigQuery table, the dataset that contains the table must be in the same regional or multi-regional location as the Cloud Storage bucket.

For information about loading Parquet data from a local file, see Loading data into BigQuery from a local data source.

Parquet schemas

When you load Parquet files into BigQuery, the table schema is automatically retrieved from the self-describing source data. When BigQuery retrieves the schema from the source data, the alphabetically last file is used.

For example, you have the following Parquet files in Cloud Storage:

gs://mybucket/00/
  a.parquet
  z.parquet
gs://mybucket/01/
  b.parquet

This command loads all of the files in a single CLI command (as a comma-separated list), and the schema is derived from mybucket/01/b.parquet:

bq --location=US load --source_format=PARQUET [DATASET].[TABLE] "gs://mybucket/00/*.parquet","gs://mybucket/01/*.parquet"

When you load multiple Parquet files that have different schemas, identical columns specified in multiple schemas must have the same mode in each schema definition.

When BigQuery detects the schema, some Parquet data types are converted to BigQuery data types to make them compatible with BigQuery SQL syntax. For more information, see Parquet conversions.

Parquet compression

Compressed Parquet files are not supported, but compressed data blocks are. BigQuery supports Snappy, GZip, and LZO_1X codecs for compressed data blocks in Parquet files.

Loading Parquet data into a new table

To load Parquet data from Cloud Storage into a new BigQuery table, or append data into an existing table:

Console

1. Open the BigQuery web UI in the GCP Console.
   Go to the BigQuery web UI

2. In the navigation panel, in the Resources section, expand your project and select a dataset.

3. On the right side of the window, in the details panel, click Create table. The process for loading data is the same as the process for creating an empty table.

   

4. On the Create table page, in the Source section:

   • For Create table from, select your desired source type.

     

   • In the source field, browse for the File/Cloud Storage bucket, or enter the Cloud Storage URI. Note that you cannot include multiple URIs in the BigQuery web UI, but wildcards are supported. The Cloud Storage bucket must be in the same location as the dataset that contains the table you're creating.

     

   • For File format, select Parquet.

5. On the Create table page, in the Destination section:

   • For Dataset name, choose the appropriate dataset.

     

   • In the Table name field, enter the name of the table you're creating in BigQuery.

   • Verify that Table type is set to Native table.

6. In the Schema section, no action is necessary. The schema is self-described in Parquet files.

7. Select applicable items in the Advanced options section and then click Create table.

Classic UI

Go to the BigQuery web UI.
Go to the BigQuery web UI

In the navigation panel, hover on a dataset, click the down arrow icon , and click Create new table. The process for loading data is the same as the process for creating an empty table.

On the Create Table page, in the Source Data section:

• For Location, select Cloud Storage and in the source field, enter the Cloud Storage URI. Note that you cannot include multiple URIs in the BigQuery web UI, but wildcards are supported. The Cloud Storage bucket must be in the same location as the dataset that contains the table you're creating.
• For File format, select Parquet.

On the Create Table page, in the Destination Table section:

• For Table name, choose the appropriate dataset, and in the table name field, enter the name of the table you're creating in BigQuery.
• Verify that Table type is set to Native table.

In the Schema section, no action is necessary. The schema is self-described in Parquet files.

Click Create Table.

Command-line

Use the bq load command, specify PARQUET as the source_format, and include a Cloud Storage URI. You can include a single URI, a comma-separated list of URIs or a URI containing a wildcard.

Supply the --location flag and set the value to your location.

bq --location=[LOCATION] load --source_format=[FORMAT] [DATASET].[TABLE] [PATH_TO_SOURCE]

where:

• [LOCATION] is your location. The --location flag is optional. For example, if you are using BigQuery in the Tokyo region, you can set the flag's value to asia-northeast1. You can set a default value for the location using the .bigqueryrc file.
• [FORMAT] is PARQUET.
• [DATASET] is an existing dataset.
• [TABLE] is the name of the table into which you're loading data.
• [PATH_TO_SOURCE] is a fully-qualified Cloud Storage URI or a comma-separated list of URIs. Wildcards are also supported.

Examples:

• The following command loads data from gs://mybucket/mydata.parquet into a table named mytable in mydataset. mybucket and mydataset were created in the US multi-region location.

  bq --location=US load --source_format=PARQUET mydataset.mytable gs://mybucket/mydata.parquet
  

• The following command loads data from multiple files in gs://mybucket/ into a table named mytable in mydataset. The Cloud Storage URI uses a wildcard. mybucket and mydataset were created in the US multi-region location.

  bq --location=US load --source_format=PARQUET mydataset.mytable gs://mybucket/mydata*.parquet
  

• The following command loads data from multiple files in gs://mybucket/ into a table named mytable in mydataset. The command includes a comma-separated list of Cloud Storage URIs with wildcards. mybucket and mydataset were created in the asia-northeast1 region.

  bq --location=asia-northeast1 load --autodetect --source_format=PARQUET mydataset.mytable "gs://mybucket/00/*.parquet","gs://mybucket/01/*.parquet"
  

API

Set the following properties to load Parquet data using the API.

1. Create a load job that points to the source data in Cloud Storage.

2. Specify your location in the location property in the jobReference section of the job resource.

3. The source URIs must be fully-qualified, in the format gs://[BUCKET]/[OBJECT]. Each URI can contain one '*' wildcard character.

4. Specify the Parquet data format by setting the configuration.load.sourceFormat property to PARQUET.

5. To check the job status, call jobs.get([JOB_ID]*), where [JOB_ID] is the ID of the job returned by the initial request.

   • If status.state = DONE, the job completed successfully.
   • If the status.errorResult property is present, the request failed, and that object will include information describing what went wrong. When a request fails, no table is created and no data is added.
   • If status.errorResult is absent, the job finished successfully, although there might have been some non-fatal errors, such as problems importing a few rows. Non-fatal errors are listed in the returned job object's status.errors property.

API notes:

• Load jobs are atomic and consistent; if a load job fails, none of the data is available, and if a load job succeeds, all of the data is available.

• As a best practice, generate a unique ID and pass it as jobReference.jobId when calling jobs.insert() to create a load job. This approach is more robust to network failure because the client can poll or retry on the known job ID.

• Calling jobs.insert() on a given job ID is idempotent; in other words, you can retry as many times as you like on the same job ID, and at most one of those operations will succeed.

Go

Before trying this sample, follow the Go setup instructions in the BigQuery Quickstart Using Client Libraries . For more information, see the BigQuery Go API reference documentation .

View on GitHub Feedback

// To run this sample, you will need to create (or reuse) a context and
// an instance of the bigquery client.  For example:
// import "cloud.google.com/go/bigquery"
// ctx := context.Background()
// client, err := bigquery.NewClient(ctx, "your-project-id")
gcsRef := bigquery.NewGCSReference("gs://cloud-samples-data/bigquery/us-states/us-states.parquet")
gcsRef.SourceFormat = bigquery.Parquet
gcsRef.AutoDetect = true
loader := client.Dataset(datasetID).Table(tableID).LoaderFrom(gcsRef)

job, err := loader.Run(ctx)
if err != nil {
	return err
}
status, err := job.Wait(ctx)
if err != nil {
	return err
}

if status.Err() != nil {
	return fmt.Errorf("Job completed with error: %v", status.Err())
}

Java

Before trying this sample, follow the Java setup instructions in the BigQuery Quickstart Using Client Libraries . For more information, see the BigQuery Java API reference documentation .

View on GitHub Feedback

String sourceUri = "gs://cloud-samples-data/bigquery/us-states/us-states.parquet";
TableId tableId = TableId.of(datasetName, "us_states");
LoadJobConfiguration configuration =
        LoadJobConfiguration.builder(tableId, sourceUri)
                .setFormatOptions(FormatOptions.parquet())
                .build();
// Load the table
Job loadJob = bigquery.create(JobInfo.of(configuration));
loadJob = loadJob.waitFor();
// Check the table
StandardTableDefinition destinationTable = bigquery.getTable(tableId).getDefinition();
System.out.println("State: " + loadJob.getStatus().getState());
System.out.printf("Loaded %d rows.\n", destinationTable.getNumRows());

Node.js

Before trying this sample, follow the Node.js setup instructions in the BigQuery Quickstart Using Client Libraries . For more information, see the BigQuery Node.js API reference documentation .

View on GitHub Feedback

// Import the Google Cloud client libraries
const {BigQuery} = require('@google-cloud/bigquery');
const {Storage} = require('@google-cloud/storage');

/**
 * TODO(developer): Uncomment the following lines before running the sample.
 */
// const datasetId = "my_dataset";
// const tableId = "my_table";

/**
 * This sample loads the Parquet file at
 * https://storage.googleapis.com/cloud-samples-data/bigquery/us-states/us-states.parquet
 *
 * TODO(developer): Replace the following lines with the path to your file.
 */
const bucketName = 'cloud-samples-data';
const filename = 'bigquery/us-states/us-states.parquet';

async function loadTableGCSParquet() {
  // Imports a GCS file into a table with Parquet source format.

  // Instantiate clients
  const bigqueryClient = new BigQuery();
  const storageClient = new Storage();

  // Configure the load job. For full list of options, see:
  // https://cloud.google.com/bigquery/docs/reference/rest/v2/jobs#configuration.load
  const metadata = {
    sourceFormat: 'PARQUET',
    location: 'US',
  };

  // Load data from a Google Cloud Storage file into the table
  const [job] = await bigqueryClient
    .dataset(datasetId)
    .table(tableId)
    .load(storageClient.bucket(bucketName).file(filename), metadata);

  // load() waits for the job to finish
  console.log(`Job ${job.id} completed.`);

  // Check the job's status for errors
  const errors = job.status.errors;
  if (errors && errors.length > 0) {
    throw errors;
  }
}
loadTableGCSParquet();

PHP

Before trying this sample, follow the PHP setup instructions in the BigQuery Quickstart Using Client Libraries . For more information, see the BigQuery PHP API reference documentation .

View on GitHub Feedback

use Google\Cloud\BigQuery\BigQueryClient;
use Google\Cloud\Core\ExponentialBackoff;

/** Uncomment and populate these variables in your code */
// $projectId  = 'The Google project ID';
// $datasetId  = 'The BigQuery dataset ID';

// instantiate the bigquery table service
$bigQuery = new BigQueryClient([
    'projectId' => $projectId,
]);
$dataset = $bigQuery->dataset($datasetId);
$table = $dataset->table('us_states');

// create the import job
$gcsUri = 'gs://cloud-samples-data/bigquery/us-states/us-states.parquet';
$loadConfig = $table->loadFromStorage($gcsUri)->sourceFormat('PARQUET');
$job = $table->runJob($loadConfig);
// poll the job until it is complete
$backoff = new ExponentialBackoff(10);
$backoff->execute(function () use ($job) {
    print('Waiting for job to complete' . PHP_EOL);
    $job->reload();
    if (!$job->isComplete()) {
        throw new Exception('Job has not yet completed', 500);
    }
});
// check if the job has errors
if (isset($job->info()['status']['errorResult'])) {
    $error = $job->info()['status']['errorResult']['message'];
    printf('Error running job: %s' . PHP_EOL, $error);
} else {
    print('Data imported successfully' . PHP_EOL);
}

Python

Before trying this sample, follow the Python setup instructions in the BigQuery Quickstart Using Client Libraries . For more information, see the BigQuery Python API reference documentation .

Use the Client.load_table_from_uri() method to start a load job from Cloud Storage. To use Parquet, set the LoadJobConfig.source_format property to the SourceFormat constant PARQUET and pass the job config as the job_config argument to the load_table_from_uri() method.

View on GitHub

# from google.cloud import bigquery
# client = bigquery.Client()
# dataset_id = 'my_dataset'

dataset_ref = client.dataset(dataset_id)
job_config = bigquery.LoadJobConfig()
job_config.source_format = bigquery.SourceFormat.PARQUET
uri = "gs://cloud-samples-data/bigquery/us-states/us-states.parquet"

load_job = client.load_table_from_uri(
    uri, dataset_ref.table("us_states"), job_config=job_config
)  # API request
print("Starting job {}".format(load_job.job_id))

load_job.result()  # Waits for table load to complete.
print("Job finished.")

destination_table = client.get_table(dataset_ref.table("us_states"))
print("Loaded {} rows.".format(destination_table.num_rows))

Overwriting a table with Parquet data

You can load additional data into a table either from source files or by appending query results.

In the console and the classic BigQuery web UI, you use the Write preference option to specify what action to take when you load data from a source file or from a query result.

You have the following options when you load additional data into a table:

Console option	Classic UI option	CLI flag	BigQuery API property	Description
Write if empty	Write if empty	None	WRITE_EMPTY	Writes the data only if the table is empty.
Append to table	Append to table	--noreplace or --replace=false; if --[no]replace is unspecified, the default is append	WRITE_APPEND	(Default) Appends the data to the end of the table.
Overwrite table	Overwrite table	--replace or --replace=true	WRITE_TRUNCATE	Erases all existing data in a table before writing the new data.

By default, load jobs will append data to a table unless the write disposition is altered. If you instead wish to replace data with the data from a load job, you can choose to overwrite data in a BigQuery table:

Console

1. Open the BigQuery web UI in the GCP Console.
   Go to the BigQuery web UI

2. In the navigation panel, in the Resources section, expand your project and select a dataset.

3. On the right side of the window, in the details panel, click Create table. The process for loading data is the same as the process for creating an empty table.

   

4. On the Create table page, in the Source section:

   • For Create table from, select your desired source type.

     

   • In the source field, browse for the File/Cloud Storage bucket, or enter the Cloud Storage URI. Note that you cannot include multiple URIs in the BigQuery web UI, but wildcards are supported. The Cloud Storage bucket must be in the same location as the dataset that contains the table you're creating.

     

   • For File format, select Parquet.

5. On the Create table page, in the Destination section:

   • For Dataset name, choose the appropriate dataset.

     

   • In the Table name field, enter the name of the table you're creating in BigQuery.

   • Verify that Table type is set to Native table.

6. In the Schema section, enter the schema definition.

   • Enter schema information manually by:

     • Enabling Edit as text and entering the table schema as a JSON array.

       Note: You can view the schema of an existing table in JSON format by entering the following command: bq show --format=prettyjson [DATASET].[TABLE].

     • Using Add field to manually input the schema.

7. In the Advanced options section, for Write preference, choose Write if empty, Append to table, or Overwrite table.

   

8. Click Create table.

Classic UI

1. Go to the BigQuery web UI.
   Go to the BigQuery web UI

2. In the navigation panel, hover on a dataset, click the down arrow icon , and click Create new table. The process for loading data is the same as the process for creating an empty table.

3. On the Create Table page, in the Source Data section:

   • For Location, select Cloud Storage and in the source field, enter the Cloud Storage URI. Note that you cannot include multiple URIs in the UI, but wildcards are supported. The Cloud Storage bucket must be in the same location as the dataset that contains the table you're appending or overwriting.
   • For File format, select Parquet.

4. On the Create Table page, in the Destination Table section:

   • For Table name, choose the appropriate dataset, and in the table name field, enter the name of the table you're appending or overwriting.
   • Verify that Table type is set to Native table.

5. In the Schema section, no action is necessary. Schema information is self-described in Parquet files.

6. In the Options section, for Write preference, choose Write if empty, Append to table, or Overwrite table.

   

7. Click Create Table.

Command-line

Enter the bq load command with the --replace flag to overwrite the table. Supply the --location flag and set the value to your location. Use the --noreplace flag to append data to the table. If no flag is specified, the default is to append data.

bq --location=[LOCATION] load --[no]replace [DATASET].[TABLE] [PATH_TO_SOURCE]

where:

• [LOCATION] is your location. The --location flag is optional. You can set a default value for the location by using the .bigqueryrc file.
• [DATASET] is an existing dataset.
• [TABLE] is the name of the table into which you're loading data.
• [PATH_TO_SOURCE] is a fully-qualified Cloud Storage URI or a comma-separated list of URIs. Wildcards are also supported.

Examples:

• The following command loads data from gs://mybucket/mydata.parquet and overwrites a table named mytable in mydataset. mybucket and mydataset were created in the US multi-region location.

  bq --location=US load --replace --source_format=PARQUET mydataset.mytable gs://mybucket/mydata.parquet
  

• The following command loads data from gs://mybucket/mydata.parquet and appends data to a table named mytable in mydataset. mybucket and mydataset were created in the asia-northeast1 region.

  bq --location=asia-northeast1 load --noreplace --source_format=PARQUET mydataset.mytable gs://mybucket/mydata.parquet
  

API

Set the following properties to load CSV data using the API.

1. Create a load job that points to the source data in Cloud Storage.

2. Specify your location in the location property in the jobReference section of the job resource.

3. The source URIs must be fully-qualified, in the format gs://[BUCKET]/[OBJECT]. You can include multiple URIs as a comma-separated list. Note that wildcards are also supported when loading CSV data from Cloud Storage.

4. Specify the data format by setting the configuration.load.sourceFormat property to PARQUET.

5. Specify the write preference by setting the configuration.load.writeDisposition property to WRITE_TRUNCATE, WRITE_APPEND, or WRITE_EMPTY.

Go

Before trying this sample, follow the Go setup instructions in the BigQuery Quickstart Using Client Libraries . For more information, see the BigQuery Go API reference documentation .

View on GitHub Feedback

// To run this sample, you will need to create (or reuse) a context and
// an instance of the bigquery client.  For example:
// import "cloud.google.com/go/bigquery"
// ctx := context.Background()
// client, err := bigquery.NewClient(ctx, "your-project-id")
gcsRef := bigquery.NewGCSReference("gs://cloud-samples-data/bigquery/us-states/us-states.parquet")
gcsRef.SourceFormat = bigquery.Parquet
gcsRef.AutoDetect = true
loader := client.Dataset(datasetID).Table(tableID).LoaderFrom(gcsRef)
loader.WriteDisposition = bigquery.WriteTruncate

job, err := loader.Run(ctx)
if err != nil {
	return err
}
status, err := job.Wait(ctx)
if err != nil {
	return err
}

if status.Err() != nil {
	return fmt.Errorf("Job completed with error: %v", status.Err())
}

Node.js

Before trying this sample, follow the Node.js setup instructions in the BigQuery Quickstart Using Client Libraries . For more information, see the BigQuery Node.js API reference documentation .

View on GitHub Feedback

// Import the Google Cloud client libraries
const {BigQuery} = require('@google-cloud/bigquery');
const {Storage} = require('@google-cloud/storage');

/**
 * TODO(developer): Uncomment the following lines before running the sample.
 */
// const datasetId = "my_dataset";
// const tableId = "my_table";

/**
 * This sample loads the CSV file at
 * https://storage.googleapis.com/cloud-samples-data/bigquery/us-states/us-states.csv
 *
 * TODO(developer): Replace the following lines with the path to your file.
 */
const bucketName = 'cloud-samples-data';
const filename = 'bigquery/us-states/us-states.parquet';

async function loadParquetFromGCSTruncate() {
  /**
   * Imports a GCS file into a table and overwrites
   * table data if table already exists.
   */

  // Instantiate clients
  const bigqueryClient = new BigQuery();
  const storageClient = new Storage();

  // Configure the load job. For full list of options, see:
  // https://cloud.google.com/bigquery/docs/reference/rest/v2/jobs#configuration.load
  const metadata = {
    sourceFormat: 'PARQUET',
    // Set the write disposition to overwrite existing table data.
    writeDisposition: 'WRITE_TRUNCATE',
    location: 'US',
  };

  // Load data from a Google Cloud Storage file into the table
  const [job] = await bigqueryClient
    .dataset(datasetId)
    .table(tableId)
    .load(storageClient.bucket(bucketName).file(filename), metadata);
  // load() waits for the job to finish
  console.log(`Job ${job.id} completed.`);

  // Check the job's status for errors
  const errors = job.status.errors;
  if (errors && errors.length > 0) {
    throw errors;
  }
}
loadParquetFromGCSTruncate();

PHP

Before trying this sample, follow the PHP setup instructions in the BigQuery Quickstart Using Client Libraries . For more information, see the BigQuery PHP API reference documentation .

View on GitHub Feedback

use Google\Cloud\BigQuery\BigQueryClient;
use Google\Cloud\Core\ExponentialBackoff;

/** Uncomment and populate these variables in your code */
// $projectId = 'The Google project ID';
// $datasetId = 'The BigQuery dataset ID';
// $tableID = 'The BigQuery table ID';

// instantiate the bigquery table service
$bigQuery = new BigQueryClient([
    'projectId' => $projectId,
]);
$table = $bigQuery->dataset($datasetId)->table($tableId);

// create the import job
$gcsUri = 'gs://cloud-samples-data/bigquery/us-states/us-states.parquet';
$loadConfig = $table->loadFromStorage($gcsUri)->sourceFormat('PARQUET')->writeDisposition('WRITE_TRUNCATE');
$job = $table->runJob($loadConfig);

// poll the job until it is complete
$backoff = new ExponentialBackoff(10);
$backoff->execute(function () use ($job) {
    print('Waiting for job to complete' . PHP_EOL);
    $job->reload();
    if (!$job->isComplete()) {
        throw new Exception('Job has not yet completed', 500);
    }
});

// check if the job has errors
if (isset($job->info()['status']['errorResult'])) {
    $error = $job->info()['status']['errorResult']['message'];
    printf('Error running job: %s' . PHP_EOL, $error);
} else {
    print('Data imported successfully' . PHP_EOL);
}

Python

Before trying this sample, follow the Python setup instructions in the BigQuery Quickstart Using Client Libraries . For more information, see the BigQuery Python API reference documentation .

To replace the rows in an existing table, set the LoadJobConfig.write_disposition property to the WriteDisposition constant WRITE_TRUNCATE.

View on GitHub

# from google.cloud import bigquery
# client = bigquery.Client()
# table_ref = client.dataset('my_dataset').table('existing_table')

job_config = bigquery.LoadJobConfig()
job_config.write_disposition = bigquery.WriteDisposition.WRITE_TRUNCATE
job_config.source_format = bigquery.SourceFormat.PARQUET
uri = "gs://cloud-samples-data/bigquery/us-states/us-states.parquet"
load_job = client.load_table_from_uri(
    uri, table_ref, job_config=job_config
)  # API request
print("Starting job {}".format(load_job.job_id))

load_job.result()  # Waits for table load to complete.
print("Job finished.")

destination_table = client.get_table(table_ref)
print("Loaded {} rows.".format(destination_table.num_rows))

Parquet conversions

BigQuery converts Parquet data types to the following BigQuery data types:

Type conversions

Parquet type	Parquet converted type(s)	BigQuery data type
BOOLEAN	NONE	Boolean
INT32	NONE, UINT_8, UINT_16, UINT_32, INT_8, INT_16, INT_32	Integer
INT32	DECIMAL (see DECIMAL annotation)	Numeric
INT32	DATE	Date
INT64	NONE, UINT_64, INT_64	Integer
INT64	DECIMAL (see DECIMAL annotation)	Numeric
INT64	TIMESTAMP_MILLIS	Timestamp
INT64	TIMESTAMP_MICROS	Timestamp
INT96	NONE	Timestamp
FLOAT	NONE	Floating point
DOUBLE	NONE	Floating point
BYTE_ARRAY	NONE	Bytes
BYTE_ARRAY	UTF8	String
FIXED_LEN_BYTE_ARRAY	DECIMAL (see DECIMAL annotation)	Numeric
FIXED_LEN_BYTE_ARRAY	NONE	Bytes

Other combinations of Parquet types and converted types are not supported.

Decimal annotation

Parquet types with the DECIMAL annotation may have at most a precision of 38 (total number of digits) and at most a scale of 9 (digits to the right of the decimal). The number of integer digits, which is the precision minus the scale, may be at most 29. For example, DECIMAL(38, 9) is supported because the precision is 38 and the scale is 9. In this example, the number of integer digits is 29. DECIMAL(38, 5) is not supported because it has a precision of 38 and a scale of 5. In this example, the number of integer digits is 33.

Column name conversions

A column name must contain only letters (a-z, A-Z), numbers (0-9), or underscores (_), and it must start with a letter or underscore. The maximum column name length is 128 characters. A column name cannot use any of the following prefixes:

• _TABLE_
• _FILE_
• _PARTITION

Duplicate column names are not allowed even if the case differs. For example, a column named Column1 is considered identical to a column named column1.

Currently, you cannot load Parquet files containing columns that have a period (.) in the column name.

If a Parquet column name contains other characters (aside from a period), the characters are replaced with underscores. Trailing underscores may be added to column names to avoid collisions. For example, if a Parquet file contains 2 columns Column1 and column1, the columns are loaded as Column1 and column1_ respectively.