private ExecMapperContext setupExecContext(Operator operator, List<Path> paths) { ExecMapperContext context = null; if (hasVC || work.getSplitSample() != null) { context = new ExecMapperContext(job); if (operator != null) { operator.passExecContext(context); } } setFetchOperatorContext(job, paths); return context; }
private ExecMapperContext setupExecContext(Operator operator, List<Path> paths) { ExecMapperContext context = null; if (hasVC || work.getSplitSample() != null) { context = new ExecMapperContext(job); if (operator != null) { operator.passExecContext(context); } } setFetchOperatorContext(job, paths); return context; }
if (work.getSplitSample() != null) { inputSplits = splitSampling(work.getSplitSample(), inputSplits);
protected FetchInputFormatSplit[] getNextSplits() throws Exception { while (getNextPath()) { // not using FileInputFormat.setInputPaths() here because it forces a connection to the // default file system - which may or may not be online during pure metadata operations job.set("mapred.input.dir", StringUtils.escapeString(currPath.toString())); // Fetch operator is not vectorized and as such turn vectorization flag off so that // non-vectorized record reader is created below. HiveConf.setBoolVar(job, HiveConf.ConfVars.HIVE_VECTORIZATION_ENABLED, false); Class<? extends InputFormat> formatter = currDesc.getInputFileFormatClass(); Utilities.copyTableJobPropertiesToConf(currDesc.getTableDesc(), job); InputFormat inputFormat = getInputFormatFromCache(formatter, job); InputSplit[] splits = inputFormat.getSplits(job, 1); FetchInputFormatSplit[] inputSplits = new FetchInputFormatSplit[splits.length]; for (int i = 0; i < splits.length; i++) { inputSplits[i] = new FetchInputFormatSplit(splits[i], inputFormat); } if (work.getSplitSample() != null) { inputSplits = splitSampling(work.getSplitSample(), inputSplits); } if (inputSplits.length > 0) { return inputSplits; } } return null; }
if (hasVC || work.getSplitSample() != null) { currRecReader = new HiveRecordReader<WritableComparable, Writable>(reader, job) { @Override
if (hasVC || work.getSplitSample() != null) { currRecReader = new HiveRecordReader<WritableComparable, Writable>(reader, job) { @Override
private ExecMapperContext setupExecContext(Operator operator, List<Path> paths) { ExecMapperContext context = null; if (hasVC || work.getSplitSample() != null) { context = new ExecMapperContext(job); if (operator != null) { operator.passExecContext(context); } } setFetchOperatorContext(job, paths); return context; }
protected FetchInputFormatSplit[] getNextSplits() throws Exception { while (getNextPath()) { // not using FileInputFormat.setInputPaths() here because it forces a connection to the // default file system - which may or may not be online during pure metadata operations job.set("mapred.input.dir", StringUtils.escapeString(currPath.toString())); // Fetch operator is not vectorized and as such turn vectorization flag off so that // non-vectorized record reader is created below. HiveConf.setBoolVar(job, HiveConf.ConfVars.HIVE_VECTORIZATION_ENABLED, false); Class<? extends InputFormat> formatter = currDesc.getInputFileFormatClass(); Utilities.copyTableJobPropertiesToConf(currDesc.getTableDesc(), job); InputFormat inputFormat = getInputFormatFromCache(formatter, job); InputSplit[] splits = inputFormat.getSplits(job, 1); FetchInputFormatSplit[] inputSplits = new FetchInputFormatSplit[splits.length]; for (int i = 0; i < splits.length; i++) { inputSplits[i] = new FetchInputFormatSplit(splits[i], inputFormat); } if (work.getSplitSample() != null) { inputSplits = splitSampling(work.getSplitSample(), inputSplits); } if (inputSplits.length > 0) { return inputSplits; } } return null; }
if (hasVC || work.getSplitSample() != null) { currRecReader = new HiveRecordReader<WritableComparable, Writable>(reader, job) { @Override