Flink 1.12.0 java.lang.NullPointerException: buffer at org.apache.flink.core.memory.MemorySegment

Flink 1.12.0 java.lang.NullPointerException: buffer at org.apache.flink.core.memory.MemorySegment.<init>

Flink 1.12.0 版本，当你使用Temporal Table Join ORC格式的Hive维度表时出现如下异常

java.lang.NullPointerException: buffer
	at org.apache.flink.core.memory.MemorySegment.<init>(MemorySegment.java:161) ~[flink-core-1.12.0.jar:1.12.0]
	at org.apache.flink.core.memory.HybridMemorySegment.<init>(HybridMemorySegment.java:86) ~[flink-core-1.12.0.jar:1.12.0]
	at org.apache.flink.core.memory.MemorySegmentFactory.wrap(MemorySegmentFactory.java:55) ~[flink-core-1.12.0.jar:1.12.0]
	at org.apache.flink.table.data.binary.BinaryStringData.fromBytes(BinaryStringData.java:98) ~[flink-table-common-1.12.0.jar:1.12.0]
	at org.apache.flink.table.data.StringData.fromBytes(StringData.java:73) ~[classes/:1.12.0]
	at org.apache.flink.table.data.ColumnarRowData.getString(ColumnarRowData.java:114) ~[flink-table-runtime-blink_2.11-1.12.0.jar:1.12.0]
	at org.apache.flink.table.data.RowData.lambda$createFieldGetter$245ca7d1$1(RowData.java:243) ~[flink-table-common-1.12.0.jar:1.12.0]
	at org.apache.flink.table.data.RowData.lambda$createFieldGetter$25774257$1(RowData.java:317) ~[flink-table-common-1.12.0.jar:1.12.0]
	at org.apache.flink.table.filesystem.FileSystemLookupFunction.extractLookupKey(FileSystemLookupFunction.java:161) ~[flink-table-runtime-blink_2.11-1.12.0.jar:1.12.0]
	at org.apache.flink.table.filesystem.FileSystemLookupFunction.checkCacheReload(FileSystemLookupFunction.java:132) ~[flink-table-runtime-blink_2.11-1.12.0.jar:1.12.0]
	at org.apache.flink.table.filesystem.FileSystemLookupFunction.eval(FileSystemLookupFunction.java:103) ~[flink-table-runtime-blink_2.11-1.12.0.jar:1.12.0]
	at LookupFunction$8.flatMap(Unknown Source) ~[?:?]

查看堆栈，发现MemorySegment类中的抛出此异常

	/**
	 * Creates a new memory segment that represents the memory of the byte array.
	 *
	 * <p>Since the byte array is backed by on-heap memory, this memory segment holds its
	 * data on heap. The buffer must be at least of size 8 bytes.
	 *
	 * @param buffer The byte array whose memory is represented by this memory segment.
	 */
	MemorySegment(byte[] buffer, Object owner) {
		if (buffer == null) {
			throw new NullPointerException("buffer");
		}

		this.heapMemory = buffer;
		this.address = BYTE_ARRAY_BASE_OFFSET;
		this.size = buffer.length;
		this.addressLimit = this.address + this.size;
		this.owner = owner;
	}

在Flink的Jira里已经有人提了issue，具体可查看https://issues.apache.org/jira/browse/FLINK-20576，但还没有官方回复

我们可以先通过修改源码的方式在上游StringData类中处理掉buffer = null的情况，避免下游抛出异常

修改 org.apache.flink.table.data.StringData类的 fromBytes(byte[] bytes, int offset, int numBytes)方法

/**
	 * Creates an instance of {@link StringData} from the given UTF-8 byte array with offset and number
	 * of bytes.
	 */
	BinaryStringData NULL_STRING_DATA = BinaryStringData.fromBytes(new byte[]{});
	static StringData fromBytes(byte[] bytes, int offset, int numBytes) {
		if (bytes == null) {
			System.out.println("bytes is null !!!!!!!!!!!!!!!!!!!!!!!!!");
			return NULL_STRING_DATA;
		}
		return BinaryStringData.fromBytes(bytes, offset, numBytes);
	}

注意要放在 org.apache.flink.table.data 目录下，在自己的IDE里建一个相同路径的目录，这时在IDE里执行程序已经不会抛出该异常了

在服务器端执行我们还需要替换 lib/flink-table_2.11-1.12.0.jar 包中的该class，记得替换前把该jar包备份一份

在服务器上创建 org.apache.flink.table.data 目录，将编译好的StringData.class上传到该目录下: org/apache/flink/table/data/StringData.class

执行如下命令 jar -uvf flink-table_2.11-1.12.0.jar org/ 就可以把修改后的StringData.class覆盖原jar中的文件了，之后在服务器上执行flink job也避免该异常出现了