FileManagment Plugins - FileCompress / FileDecompress

General 

To support use cases of migrating files from OnPrem to Google cloud there is a need to for comphrensive file handling capabilities. This includes FileList, FileCompression, FileDecompression, FileEncryption , FileDecryption etc.  There are few file level plugins available in CDAP like FileMove, FileDelete and this needs to be expanded. 

UseCase

Proposed Design 

1. FileList Plugin - BatchSource Plugin  - Implement a new FileList plugin ( Batchsource plugin) with similar capability of the current FileSource plugin but instead of actually reading the file contents  it would just pass the filenames with full URI to be used for processing the following actions in pipeline. 
2. FileCompress Plugin - Transform Plugin -  Implement a Compression plugin similar to the Field Compression Plugin which accepts an input file URI , reads the file compress this and stored it temporarily on teh same node and spits out the compressed file URI location that can be used by the next processing action.   
3. Invoke the current Google Cloud Storage plugin to persist the file to Cloud storage. 

FileList Plugin 

This is a Batchsource plugin similar to current FileSource Plugin but only list the filenames with full URI and not actually read the contents of the file. 

Plugin Properties

Implementation Logic 

Base the code off FileSource / AbstractFileSource Plugin  . Override the PathTrackingInputFormat  ( I have checked the PathTrackingBlobInputFormat, probably I can take that and just overwrite without a deligate and use that ).  

Use PathTrackingBlobInputFormat 

@Override
public StructuredRecord.Builder getCurrentValue() {
  String fieldName = schema.getFields().iterator().next().getName();
  return StructuredRecord.builder(schema).set(fieldName, val  Instead of actual record just pass in file path with name of the file ?);
}

FileCompress Plugin

This plugin will take input file name thats passed from the FileList Plugin, Get the fileInputStream using the URI and then using Gzip or Snappy libs compress the file and store them locally on the node. 

Plugin Properties

 Implement a new  Plugin for Compression,  use the exiting FieldCompression Plugin as starting point with following modification  

@Override
public void transform(StructuredRecord in, Emitter<StructuredRecord> emitter) throws Exception {
  StructuredRecord.Builder builder = StructuredRecord.builder(outSchema);

  Schema inSchema = in.getSchema();
  List<Field> inFields = inSchema.getFields();

  // Iterate through input fields. Check if field name is present 
 // in the fields that need to be compressed, if it's not then write
 // to output as it is. 
 for (Field field : inFields) {
    String FileName = field.getName();

Open the file , get the InputFile Stream and pass this to the GZIP compression API's and store the file locally on the node. 

Queries 

1. Should this plugin just store the file locally and pass the new compressed file name or should this actually read the file , compress contents and passed the compressed contents as a stream. The issue with stream is if the file size if larger then it might not be efficient or we might have to do everything in memory which is not ideal. 
2. If we store the files locally we need some way to clean it up after the processing.