Column names can only contain a-z, A-Z, 0-9 and underscore (_), all other symbols will be replaced by underscores (_). The previous implementation was not fully compatible with Apache Spark implementations.