如何在BigQuery中使用非必填字段创建表？

I'm making a RESTful API in Go that writes rows in BigQuery. I'm using Google BigQuery package for Go.

In order to create the BigQuery scheme, I'm infering the schema from a struct as described in the example.

The problem is that, the resulting schema has all the non-repeated fields as "Required", so that, when I want to upload a struct with null values, the null values are uploaded as empy fields...

This is an example of my struct:

type Stats struct {
    Name            string        `bigquery:"name"`
    LastName        int           `bigquery:"last_name"`
    PhoneNumber     string        `bigquery:"phone_number"`
}

This is an example of how the schema is created:

testSchema, err := bigquery.InferSchema(Stats{})
if err != nil {
    // TODO: Handle error.
}

And, if I upload a struct with only one field set:

rows := []*Stats{
    {Name: "testA"},
}

u := table.Uploader()
err2 := u.Put(ctx, rows)

The result is that, in BigQuery, the fields "last_name" and "phone_number" is an empty string "" instead of NULL

For that, apparently you would have to use the specific BigQuery nullable types (NullInt64, NullFloat64, NullString, NullBool, NullTimestamp, NullDate, NullTime and NullDateTime).

Example using bigquery.NullBool: https://godoc.org/cloud.google.com/go/bigquery#ex-InferSchema--Tags

I am no Go expert, but looking at code at https://github.com/GoogleCloudPlatform/google-cloud-go/blob/master/bigquery/schema.go#L182 it looks like the inferFieldSchema method always sets Required: true. I would file a bug to allow to control it (although not clear how), or you can patch your schema after it was created - schema modification from REQUIRED to NULLABLE is supported by BigQuery.