I feel this adds a lot of boilerplate code.

My understanding this that this PR takes this approach
```
beamKvCoder -> new NemoKVEncoderFactory(beamKvCoder.getKeyCoder(), 
beamKvCoder.getValueCoder())
```

such that in the IR we can access the key and value coders while bypassing 
Beam-specific code.

What do you think about this approach? This also bypasses Beam-specific code, 
perhaps with less code.

```
NemoKeyCoderExtractor -> beamKvCoder.getKeyCoder() OR 
rddTupleCoder.getKeyCoder()
```

[ Full content available at: https://github.com/apache/incubator-nemo/pull/115 ]
This message was relayed via gitbox.apache.org for [email protected]

Reply via email to