How to specify sql dialect when creating spark dataframe from JDBC?

问题

I'm having an issue reading data via custom JDBC with Spark. How would I go about about overriding the sql dialect inferred via jdbc url?

The database in question is vitess (https://github.com/youtube/vitess) which runs a mysql variant, so I want to specify a mysql dialect. The jdbc url begins with jdbc:vitess/

Otherwise the DataFrameReader is inferring a default dialect which uses """ as a quote identifier. As a result, queries via spark.read.jdbc get sent as

Select 'id', 'col2', col3', 'etc' from table

which selects the string representations instead of the column values instead of

Select id, col2, col3, etc from table

回答1:

Maybe it's too late. But answer will be next:

Create your custom dialect, as I did for ClickHouse database(my jdbc connection url looks like this jdbc:clickhouse://localhost:8123)

 private object ClickHouseDialect extends JdbcDialect {
    //override here quoting logic as you wish
    override def quoteIdentifier(colName: String): String = colName

    override def canHandle(url: String): Boolean = url.startsWith("jdbc:clickhouse")
  }

And register it somewhere in your code, like this:

JdbcDialects.registerDialect(ClickHouseDialect)

回答2:

You can do something like this.

val jdbcDF = spark.read
  .format("jdbc")
  .option("url", "jdbc:postgresql:dbserver")
  .option("dbtable", "schema.tablename")
  .option("user", "username")
  .option("password", "password")
  .load()

For more info check this

You can also specify in this way.

val connectionProperties = new Properties()
    connectionProperties.put("user", "username")
    connectionProperties.put("password", "password")
    val jdbcDF2 = spark.read
      .jdbc("jdbc:postgresql:dbserver", "schema.tablename", connectionProperties)

来源：https://stackoverflow.com/questions/40371433/how-to-specify-sql-dialect-when-creating-spark-dataframe-from-jdbc