Class SumMapper

src/com/winvector/logistic/mr/MapRedFn.java:134–263

	}
	
	public static final class SumMapper extends Mapper<LongWritable,Text,Text,DoubleWritable> {
		private LineBurster burster = null;
		// config
		private JobStateDescr config = null;
		// derived
		private VariableEncodings defs = null;
		private Log log = null;
		private String hostDescr = null;
		// scratch
		private double[] pscratch = null;
		// result
		private long nProcessed = 0;
		private VEval accum = null;
		
		@Override
		public void setup(final Context context) throws IOException {
			// read side-channel configuration
			log = LogFactory.getLog(SumMapper.class);
			hostDescr = WritableUtils.hostDescr();			
			log.info(".setup() " + hostDescr); 
			try {
				burster = SerialUtils.readSerialiazlabeFromString(context.getConfiguration().get(MapRedScan.BURSTERSERFIELD));
			} catch (ClassNotFoundException e) {
				throw new IOException(e.toString());
			}
			config = JobStateDescr.fromString(context.getConfiguration().get(MRFIELDNAME));
			defs = new VariableEncodings(config.defs,config.useIntercept,config.weightKey);
			accum = new VEval(config.x,config.wantGrad,config.wantHessian);
			pscratch = new double[defs.noutcomes()];
			nProcessed = 0;
		}

		@Override
		public void map(final LongWritable key, final Text value, final Context context) {
			final String origStr = value.toString();
			final BurstMap parsed = burster.parse(origStr);
			if(!parsed.isEmpty()) {
				final String resStr = parsed.getAsString(config.defs.resultColumn);
				if((resStr!=null)&&(resStr.length()>0)) {
					final Integer category = defs.category(resStr.trim());
					if((category!=null)&&(category>=0)) {
						final SparseSemiVec v = defs.vector(parsed);
						final double wt = defs.weight(parsed);
						if((wt>0.0)&&(v!=null)) {
							final ExampleRow r = new SparseExampleRow(v,wt,category);
							config.underlying.addTerm(config.x, config.wantGrad, config.wantHessian, r, accum, pscratch);
							++nProcessed;
						}
					}
				}
			}
		}
		
		/**
		 * only write summing terms (fx,gx,hx) into Text,WritableDouble format
		 * @param v
		 * @param context
		 * @throws InterruptedException