diff --git a/java/src/main/java/org/softwareheritage/graph/maps/LabelMapBuilder.java b/java/src/main/java/org/softwareheritage/graph/maps/LabelMapBuilder.java index 3f97a82..47905cd 100644 --- a/java/src/main/java/org/softwareheritage/graph/maps/LabelMapBuilder.java +++ b/java/src/main/java/org/softwareheritage/graph/maps/LabelMapBuilder.java @@ -1,425 +1,541 @@ package org.softwareheritage.graph.maps; import com.martiansoftware.jsap.*; import it.unimi.dsi.big.webgraph.LazyLongIterator; import it.unimi.dsi.big.webgraph.labelling.ArcLabelledImmutableGraph; import it.unimi.dsi.big.webgraph.labelling.BitStreamArcLabelledImmutableGraph; import it.unimi.dsi.fastutil.BigArrays; import it.unimi.dsi.fastutil.Size64; import it.unimi.dsi.fastutil.bytes.ByteArrays; import it.unimi.dsi.fastutil.io.BinIO; import it.unimi.dsi.fastutil.io.FastBufferedInputStream; import it.unimi.dsi.fastutil.longs.LongBigArrays; import it.unimi.dsi.fastutil.longs.LongIterator; import it.unimi.dsi.fastutil.objects.Object2LongFunction; import it.unimi.dsi.io.OutputBitStream; import it.unimi.dsi.logging.ProgressLogger; import it.unimi.dsi.big.webgraph.BVGraph; import it.unimi.dsi.big.webgraph.ImmutableGraph; import it.unimi.dsi.big.webgraph.NodeIterator; import org.slf4j.Logger; import org.slf4j.LoggerFactory; import org.softwareheritage.graph.labels.DirEntry; import org.softwareheritage.graph.labels.SwhLabel; import java.io.*; import java.lang.reflect.Array; +import java.math.BigInteger; +import java.nio.ByteBuffer; +import java.nio.charset.Charset; import java.nio.charset.StandardCharsets; import java.util.*; import java.util.concurrent.TimeUnit; public class LabelMapBuilder { final static String SORT_BUFFER_SIZE = "40%"; final static Logger logger = LoggerFactory.getLogger(LabelMapBuilder.class); String graphPath; String debugPath; String tmpDir; ImmutableGraph graph; Object2LongFunction swhIdMph; long[][] orderMap; Object2LongFunction filenameMph; long numFilenames; int totalLabelWidth; public LabelMapBuilder(String graphPath, String debugPath, String tmpDir) { this.graphPath = graphPath; this.debugPath = debugPath; this.tmpDir = tmpDir; } private static JSAPResult parse_args(String[] args) { JSAPResult config = null; try { SimpleJSAP jsap = new SimpleJSAP(LabelMapBuilder.class.getName(), "", new Parameter[]{ new FlaggedOption("graphPath", JSAP.STRING_PARSER, JSAP.NO_DEFAULT, JSAP.REQUIRED, 'g', "graph", "Basename of the compressed graph"), new FlaggedOption("debugPath", JSAP.STRING_PARSER, JSAP.NO_DEFAULT, JSAP.NOT_REQUIRED, 'd', "debug-path", "Store the intermediate representation here for debug"), new FlaggedOption("tmpDir", JSAP.STRING_PARSER, "tmp", JSAP.NOT_REQUIRED, 't', "tmp", "Temporary directory path"),}); config = jsap.parse(args); if (jsap.messagePrinted()) { System.exit(1); } } catch (JSAPException e) { e.printStackTrace(); } return config; } public static void main(String[] args) throws IOException, InterruptedException { JSAPResult config = parse_args(args); String graphPath = config.getString("graphPath"); String tmpDir = config.getString("tmpDir"); String debugPath = config.getString("debugPath"); LabelMapBuilder builder = new LabelMapBuilder(graphPath, debugPath, tmpDir); logger.info("Loading graph and MPH functions..."); builder.loadGraph(); - builder.computeLabelMapSort(); + // builder.computeLabelMapSort(); + builder.computeLabelMapBsort(); } @SuppressWarnings("unchecked") // Suppress warning for Object2LongFunction cast static Object2LongFunction loadMPH(String mphBasename) throws IOException { Object2LongFunction mphMap = null; try { mphMap = (Object2LongFunction) BinIO.loadObject(mphBasename + ".mph"); } catch (ClassNotFoundException e) { logger.error("unknown class object in .mph file: " + e); System.exit(2); } return mphMap; } static long getMPHSize(Object2LongFunction mph) { return (mph instanceof Size64) ? ((Size64) mph).size64() : mph.size(); } void computeLabelMapSort() throws IOException { // Pass the intermediate representation to sort(1) so that we see the labels in the order they will // appear in the label file. ProcessBuilder processBuilder = new ProcessBuilder(); processBuilder.command("sort", "-k1,1n", "-k2,2n", // Numerical sort "--numeric-sort", "--buffer-size", SORT_BUFFER_SIZE, "--temporary-directory", tmpDir); Process sort = processBuilder.start(); BufferedOutputStream sort_stdin = new BufferedOutputStream(sort.getOutputStream()); - BufferedInputStream sort_stdout = new BufferedInputStream(sort.getInputStream()); + // BufferedInputStream sort_stdout = new BufferedInputStream(sort.getInputStream()); + FastBufferedInputStream sort_stdout = new FastBufferedInputStream(sort.getInputStream()); final FastBufferedInputStream fbis = new FastBufferedInputStream(System.in); hashLabelStream(fbis, new EdgeLabelLineWriter() { @Override public void writeLine(long src, long dst, long filenameId, int permission) throws IOException { sort_stdin.write((src + "\t" + dst + "\t" + filenameId + "\t" + permission + "\n") .getBytes(StandardCharsets.US_ASCII)); } }); sort_stdin.close(); EdgeLabelLineIterator mapLines = new TextualEdgeLabelLineIterator(sort_stdout); writeLabels(mapLines); - logger.info("Done"); } void computeLabelMapBsort() throws IOException, InterruptedException { // Pass the intermediate representation to bsort(1) so that we see the labels in the order they will // appear in the label file. + String tmpFile = tmpDir + "/labelsToSort.bin"; final FastBufferedInputStream fbis = new FastBufferedInputStream(System.in); final DataOutputStream out = new DataOutputStream(new BufferedOutputStream(new FileOutputStream(tmpFile))); + + // Number of bytes to represent a node. + final int nodeBytes = (Long.SIZE - Long.numberOfLeadingZeros(graph.numNodes())) / 8 + 1; + ByteBuffer buffer = ByteBuffer.allocate(Long.BYTES); + logger.info("Writing labels to a packed binary files (node bytes: {})", nodeBytes); + hashLabelStream(fbis, new EdgeLabelLineWriter() { @Override public void writeLine(long src, long dst, long filenameId, int permission) throws IOException { - out.writeLong(src); - out.writeLong(dst); + buffer.putLong(0, src); + out.write(buffer.array(), Long.BYTES - nodeBytes, nodeBytes); + buffer.putLong(0, dst); + out.write(buffer.array(), Long.BYTES - nodeBytes, nodeBytes); out.writeLong(filenameId); out.writeInt(permission); } }); ProcessBuilder processBuilder = new ProcessBuilder(); processBuilder.command( "/home/seirl/bsort/src/bsort", "-v", - "-r", String.valueOf(8 + 8 + 8 + 4), - "-k", String.valueOf(8 + 8), + "-r", String.valueOf(nodeBytes * 2 + Long.BYTES + Integer.BYTES), + "-k", String.valueOf(nodeBytes * 2), tmpFile ); Process sort = processBuilder.start(); sort.waitFor(); - final DataInputStream sortedLabels = new DataInputStream(new DataInputStream(new FileInputStream(tmpFile))); - BinaryEdgeLabelLineIterator mapLines = new BinaryEdgeLabelLineIterator(sortedLabels); + final DataInputStream sortedLabels = new DataInputStream(new BufferedInputStream(new FileInputStream(tmpFile))); + BinaryEdgeLabelLineIterator mapLines = new BinaryEdgeLabelLineIterator(sortedLabels, nodeBytes); writeLabels(mapLines); logger.info("Done"); } void loadGraph() throws IOException { graph = BVGraph.loadMapped(graphPath); swhIdMph = loadMPH(graphPath); orderMap = LongBigArrays.newBigArray(getMPHSize(swhIdMph)); BinIO.loadLongs(graphPath + ".order", orderMap); filenameMph = loadMPH(graphPath + "-labels"); numFilenames = getMPHSize(filenameMph); totalLabelWidth = DirEntry.labelWidth(numFilenames); } void hashLabelStream(FastBufferedInputStream input, EdgeLabelLineWriter writer) throws IOException { // Compute intermediate representation and write it on : // "