List of usage examples for org.apache.hadoop.io.compress SplitCompressionInputStream close
@Override public void close() throws IOException
From source file:org.hedera.io.input.WikiRevisionInputFormat.java
License:Apache License
/** * This code is copied from StreamWikiDumpNewInputFormat.java by Yusuke Matsubara. * Thanks to Tu Meteora for adjusting the code to the new mapreduce framework * @param job the job context/*from w w w . j ava 2 s . com*/ * @throws IOException */ public List<InputSplit> getSplits(JobContext jc, FileStatus file, long splitSize) throws IOException { List<InputSplit> splits = new ArrayList<InputSplit>(); Path path = file.getPath(); LOG.info("Splitting file " + path.getName()); Configuration conf = jc.getConfiguration(); configure(conf); long length = file.getLen(); FileSystem fs = file.getPath().getFileSystem(conf); BlockLocation[] blkLocations = fs.getFileBlockLocations(file, 0, length); if ((length != 0) && isSplitable(jc, path)) { long bytesRemaining = length; SeekableInputStream in = SeekableInputStream.getInstance(path, 0, length, fs, this.compressionCodecs); SplitCompressionInputStream is = in.getSplitCompressionInputStream(); long start = 0; long skip = 0; if (is != null) { start = is.getAdjustedStart(); length = is.getAdjustedEnd(); is.close(); in = null; } FileSplit split = null; Set<Long> processedPageEnds = new HashSet<Long>(); float factor = conf.getFloat(KEY_SKIP_FACTOR, 1.2F); READLOOP: while (((double) bytesRemaining) / splitSize > factor && bytesRemaining > 0) { // prepare matcher ByteMatcher matcher; { long st = Math.min(start + skip + splitSize, length - 1); split = makeSplit(path, st, Math.min(splitSize, length - st), blkLocations); if (in != null) in.close(); if (split.getLength() <= 1) { break; } in = SeekableInputStream.getInstance(split, fs, this.compressionCodecs); } matcher = new ByteMatcher(in); // read until the next page end in the look-ahead split while (!matcher.readUntilMatch(END_PAGE_TAG, null, split.getStart() + split.getLength(), null)) { if (matcher.getPos() >= length || split.getLength() == length - split.getStart()) break READLOOP; split = makeSplit(path, split.getStart(), Math.min(split.getLength() + splitSize, length - split.getStart()), blkLocations); } if (matcher.getLastUnmatchPos() > 0 && matcher.getPos() > matcher.getLastUnmatchPos() && !processedPageEnds.contains(matcher.getPos())) { splits.add(makeSplit(path, start, matcher.getPos() - start, blkLocations)); processedPageEnds.add(matcher.getPos()); long newstart = Math.max(matcher.getLastUnmatchPos(), start); bytesRemaining = length - newstart; start = newstart; skip = 0; } else { skip = matcher.getPos() - start; } } if (bytesRemaining > 0 && !processedPageEnds.contains(length)) { splits.add(makeSplit(path, length - bytesRemaining, bytesRemaining, blkLocations[blkLocations.length - 1].getHosts())); } if (in != null) in.close(); } else if (length != 0) { splits.add(makeSplit(path, 0, length, blkLocations)); } else { //Create empty hosts array for zero length files splits.add(makeSplit(path, 0, length, new String[0])); } return splits; }
From source file:org.wikimedia.wikihadoop.StreamWikiDumpInputFormat.java
License:Apache License
public List<InputSplit> getSplits(JobConf job, FileStatus file, String pattern, long splitSize) throws IOException { NetworkTopology clusterMap = new NetworkTopology(); List<InputSplit> splits = new ArrayList<InputSplit>(); Path path = file.getPath();//from w ww . j av a 2 s. c om long length = file.getLen(); FileSystem fs = file.getPath().getFileSystem(job); BlockLocation[] blkLocations = fs.getFileBlockLocations(file, 0, length); if ((length != 0) && isSplitable(fs, path)) { long bytesRemaining = length; SeekableInputStream in = SeekableInputStream.getInstance(path, 0, length, fs, this.compressionCodecs); SplitCompressionInputStream is = in.getSplitCompressionInputStream(); long start = 0; long skip = 0; if (is != null) { start = is.getAdjustedStart(); length = is.getAdjustedEnd(); is.close(); in = null; } LOG.info("locations=" + Arrays.asList(blkLocations)); FileSplit split = null; Set<Long> processedPageEnds = new HashSet<Long>(); float factor = job.getFloat(KEY_SKIP_FACTOR, 1.2F); READLOOP: while (((double) bytesRemaining) / splitSize > factor && bytesRemaining > 0) { // prepare matcher ByteMatcher matcher; { long st = Math.min(start + skip + splitSize, length - 1); split = makeSplit(path, st, Math.min(splitSize, length - st), clusterMap, blkLocations); System.err.println("split move to: " + split); if (in != null) in.close(); if (split.getLength() <= 1) { break; } in = SeekableInputStream.getInstance(split, fs, this.compressionCodecs); SplitCompressionInputStream cin = in.getSplitCompressionInputStream(); } matcher = new ByteMatcher(in); // read until the next page end in the look-ahead split boolean reach = false; while (!matcher.readUntilMatch(pageEndPattern, null, split.getStart() + split.getLength())) { if (matcher.getPos() >= length || split.getLength() == length - split.getStart()) break READLOOP; reach = false; split = makeSplit(path, split.getStart(), Math.min(split.getLength() + splitSize, length - split.getStart()), clusterMap, blkLocations); System.err.println("split extend to: " + split); } System.err.println( path + ": #" + splits.size() + " " + pageEndPattern + " found: pos=" + matcher.getPos() + " last=" + matcher.getLastUnmatchPos() + " read=" + matcher.getReadBytes() + " current=" + start + " remaining=" + bytesRemaining + " split=" + split); if (matcher.getLastUnmatchPos() > 0 && matcher.getPos() > matcher.getLastUnmatchPos() && !processedPageEnds.contains(matcher.getPos())) { splits.add(makeSplit(path, start, matcher.getPos() - start, clusterMap, blkLocations)); processedPageEnds.add(matcher.getPos()); long newstart = Math.max(matcher.getLastUnmatchPos(), start); bytesRemaining = length - newstart; start = newstart; skip = 0; } else { skip = matcher.getPos() - start; } } if (bytesRemaining > 0 && !processedPageEnds.contains(length)) { System.err.println( pageEndPattern + " remaining: pos=" + (length - bytesRemaining) + " end=" + length); splits.add(makeSplit(path, length - bytesRemaining, bytesRemaining, blkLocations[blkLocations.length - 1].getHosts())); } if (in != null) in.close(); } else if (length != 0) { splits.add(makeSplit(path, 0, length, clusterMap, blkLocations)); } else { //Create empty hosts array for zero length files splits.add(makeSplit(path, 0, length, new String[0])); } return splits; }
From source file:wiki.hadoop.mapred.lib.input.StreamWikiDumpInputFormat.java
License:Apache License
public List<InputSplit> getSplits(JobConf job, FileStatus file, String pattern, long splitSize) throws IOException { NetworkTopology clusterMap = new NetworkTopology(); List<InputSplit> splits = new ArrayList<InputSplit>(); Path path = file.getPath();/*from w ww . j a v a2s . c o m*/ long length = file.getLen(); FileSystem fs = file.getPath().getFileSystem(job); BlockLocation[] blkLocations = fs.getFileBlockLocations(file, 0, length); if ((length != 0) && isSplitable(fs, path)) { long bytesRemaining = length; SeekableInputStream in = SeekableInputStream.getInstance(path, 0, length, fs, this.compressionCodecs); SplitCompressionInputStream is = in.getSplitCompressionInputStream(); long start = 0; long skip = 0; if (is != null) { start = is.getAdjustedStart(); length = is.getAdjustedEnd(); is.close(); in = null; } LOG.info("locations=" + Arrays.asList(blkLocations)); FileSplit split = null; Set<Long> processedPageEnds = new HashSet<Long>(); float factor = job.getFloat(KEY_SKIP_FACTOR, 1.2F); READLOOP: while (((double) bytesRemaining) / splitSize > factor && bytesRemaining > 0) { // prepare matcher ByteMatcher matcher; { long st = Math.min(start + skip + splitSize, length - 1); split = makeSplit(path, st, Math.min(splitSize, length - st), clusterMap, blkLocations); System.err.println("split move to: " + split); if (in != null) in.close(); if (split.getLength() <= 1) { break; } in = SeekableInputStream.getInstance(split, fs, this.compressionCodecs); SplitCompressionInputStream cin = in.getSplitCompressionInputStream(); } matcher = new ByteMatcher(in, split.getStart()); // read until the next page end in the look-ahead split boolean reach = false; while (!matcher.readUntilMatch(pageEndPattern, null, split.getStart() + split.getLength())) { if (matcher.getPos() >= length || split.getLength() == length - split.getStart()) break READLOOP; reach = false; split = makeSplit(path, split.getStart(), Math.min(split.getLength() + splitSize, length - split.getStart()), clusterMap, blkLocations); System.err.println("split extend to: " + split); } System.err.println( path + ": #" + splits.size() + " " + pageEndPattern + " found: pos=" + matcher.getPos() + " last=" + matcher.getLastUnmatchPos() + " read=" + matcher.getReadBytes() + " current=" + start + " remaining=" + bytesRemaining + " split=" + split); if (matcher.getLastUnmatchPos() > 0 && matcher.getPos() > matcher.getLastUnmatchPos() && !processedPageEnds.contains(matcher.getPos())) { splits.add(makeSplit(path, start, matcher.getPos() - start, clusterMap, blkLocations)); processedPageEnds.add(matcher.getPos()); long newstart = Math.max(matcher.getLastUnmatchPos(), start); bytesRemaining = length - newstart; start = newstart; skip = 0; } else { skip = matcher.getPos() - start; } } if (bytesRemaining > 0 && !processedPageEnds.contains(length)) { System.err.println( pageEndPattern + " remaining: pos=" + (length - bytesRemaining) + " end=" + length); splits.add(new FileSplit(path, length - bytesRemaining, bytesRemaining, blkLocations[blkLocations.length - 1].getHosts())); } if (in != null) in.close(); } else if (length != 0) { splits.add(makeSplit(path, 0, length, clusterMap, blkLocations)); } else { //Create empty hosts array for zero length files splits.add(new FileSplit(path, 0, length, new String[0])); } return splits; }