zenfenan commented on a change in pull request #11307: [FLINK-16371]
[BulkWriter] Fix Hadoop Compression BulkWriter
URL: https://github.com/apache/flink/pull/11307#discussion_r389804308
##########
File path:
flink-formats/flink-compress/src/main/java/org/apache/flink/formats/compress/CompressWriterFactory.java
##########
@@ -42,39 +47,57 @@
private Extractor<IN> extractor;
private CompressionCodec hadoopCodec;
+ private String hadoopCodecName;
+ private Map<String, String> hadoopConfigurationMap;
+ private String codecExtension;
public CompressWriterFactory(Extractor<IN> extractor) {
- this.extractor = Preconditions.checkNotNull(extractor,
"extractor cannot be null");
+ this.extractor = checkNotNull(extractor, "Extractor cannot be
null");
+ this.hadoopConfigurationMap = new HashMap<>();
}
public CompressWriterFactory<IN> withHadoopCompression(String
hadoopCodecName) {
return withHadoopCompression(hadoopCodecName, new
Configuration());
}
public CompressWriterFactory<IN> withHadoopCompression(String
hadoopCodecName, Configuration hadoopConfiguration) {
- return withHadoopCompression(new
CompressionCodecFactory(hadoopConfiguration).getCodecByName(hadoopCodecName));
- }
+ CompressionCodec codec = new
CompressionCodecFactory(hadoopConfiguration).getCodecByName(hadoopCodecName);
+ this.codecExtension = checkNotNull(codec, "Unable to load the
provided Hadoop codec [" + hadoopCodecName + "]")
+ .getDefaultExtension();
+
+ this.hadoopCodecName = hadoopCodecName;
+
+ for (Map.Entry<String, String> entry : hadoopConfiguration) {
+ hadoopConfigurationMap.put(entry.getKey(),
entry.getValue());
+ }
- public CompressWriterFactory<IN> withHadoopCompression(CompressionCodec
hadoopCodec) {
- this.hadoopCodec = Preconditions.checkNotNull(hadoopCodec,
"hadoopCodec cannot be null");
return this;
}
@Override
public BulkWriter<IN> create(FSDataOutputStream out) throws IOException
{
- try {
- return (hadoopCodec != null)
- ? new HadoopCompressionBulkWriter<>(out,
extractor, hadoopCodec)
- : new NoCompressionBulkWriter<>(out, extractor);
- } catch (Exception e) {
- throw new IOException(e.getLocalizedMessage(), e);
+ if (hadoopCodecName == null || hadoopCodecName.length() == 0) {
+ return new NoCompressionBulkWriter<>(out, extractor);
}
+
+ initializeCompressionCodec();
+
+ return new
HadoopCompressionBulkWriter<>(hadoopCodec.createOutputStream(out), extractor);
}
- public String codecExtension() {
- return (hadoopCodec != null)
- ? hadoopCodec.getDefaultExtension()
- : "";
+ public String getExtension() {
+ return (hadoopCodecName != null) ? this.codecExtension : "";
}
+ private void initializeCompressionCodec() {
+ if (hadoopCodec == null) {
+ Configuration conf = new Configuration();
+
+ for (Map.Entry<String, String> entry :
hadoopConfigurationMap.entrySet()) {
+ conf.set(entry.getKey(), entry.getValue());
+ }
+
+ hadoopCodec = new
CompressionCodecFactory(conf).getCodecByName(this.hadoopCodecName);
+ }
+ }
Review comment:
I'm all in for making the code more concise and clear. However, I thought of
not spending much time on the Map to Configuration, and vice versa since I
thought of getting rid of this Map based approach altogether in a further patch
(based on a decision from the ML thread on SerializableHadoopConfiguration). Do
you still want me to go ahead with these separate static methods ?
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
