Examples of org.apache.lucene.benchmark.byTask.Benchmark

Package org.apache.lucene.benchmark.byTask

Examples of org.apache.lucene.benchmark.byTask.Benchmark

org.apache.lucene.benchmark.byTask.Benchmark
Run the benchmark algorithm.
Usage: java Benchmark algorithm-file
1. Read algorithm.
2. Run the algorithm.
Things to be added/fixed in "Benchmarking by tasks":
1. TODO - report into Excel and/or graphed view.
2. TODO - perf comparison between Lucene releases over the years.
3. TODO - perf report adequate to include in Lucene nightly build site? (so we can easily track performance changes.)
4. TODO - add overall time control for repeated execution (vs. current by-count only).
5. TODO - query maker that is based on index statistics.

        "[ { AddDoc } : * ] : 4 ",
        "CloseIndex",
    };
    
    // 2. execute the algorithm  (required in every "logic" test)
    Benchmark benchmark = execBenchmark(algLines);


    // 3. test number of docs in the index
    IndexReader ir = IndexReader.open(benchmark.getRunData().getDirectory());
    int ndocsExpected = 20; // Reuters20DocMaker exhausts after 20 docs.
    assertEquals("wrong number of docs in the index!", ndocsExpected, ir.numDocs());
    ir.close();
  }

View Full Code Here

      "# ----- alg ",
      "{WriteLineDoc()}:" + NUM_TRY_DOCS,
    };


    // Run algo
    Benchmark benchmark = execBenchmark(algLines1);


    // Verify we got somewhere between 1-500 lines (some
    // Reuters docs have no body, which WriteLineDoc task
    // skips).
    BufferedReader r = new BufferedReader(new FileReader(lineFile));
    int numLines = 0;
    while(r.readLine() != null)
      numLines++;
    r.close();
    assertTrue("did not see the right number of docs; should be > 0 and <= " + NUM_TRY_DOCS + " but was " + numLines, numLines > 0 && numLines <= NUM_TRY_DOCS);
    
    // Index the line docs
    String algLines2[] = {
      "# ----- properties ",
      "analyzer=org.apache.lucene.analysis.SimpleAnalyzer",
      "doc.maker=org.apache.lucene.benchmark.byTask.feeds.LineDocMaker",
      "docs.file=" + lineFile.getAbsolutePath().replace('\\', '/'),
      "doc.maker.forever=false",
      "doc.reuse.fields=false",
      "autocommit=false",
      "ram.flush.mb=4",
      "# ----- alg ",
      "ResetSystemErase",
      "CreateIndex",
      "{AddDoc}: *",
      "CloseIndex",
    };
    
    // Run algo
    benchmark = execBenchmark(algLines2);


    // now we should be able to open the index for write. 
    IndexWriter iw = new IndexWriter(benchmark.getRunData().getDirectory(),null,false,IndexWriter.MaxFieldLength.UNLIMITED);
    iw.close();


    IndexReader ir = IndexReader.open(benchmark.getRunData().getDirectory());
    assertEquals(numLines + " lines were were created but " + ir.numDocs() + " docs are in the index", numLines, ir.numDocs());
    ir.close();


    lineFile.delete();
  }

View Full Code Here

      "{AddDoc}: " + NUM_DOCS,
      "CloseIndex",
    };


    // Run algo
    Benchmark benchmark = execBenchmark(algLines1);


    List stats = benchmark.getRunData().getPoints().taskStats();


    // Count how many tokens all ReadTokens saw
    int totalTokenCount1 = 0;
    for (Iterator it = stats.iterator(); it.hasNext();) {
      TaskStats stat = (TaskStats) it.next();
      if (stat.getTask().getName().equals("ReadTokens")) {
        totalTokenCount1 += stat.getCount();
      }
    }


    // Separately count how many tokens are actually in the index:
    IndexReader reader = IndexReader.open(benchmark.getRunData().getDirectory());
    assertEquals(NUM_DOCS, reader.numDocs());


    TermEnum terms = reader.terms();
    TermDocs termDocs = reader.termDocs();
    int totalTokenCount2 = 0;

View Full Code Here

        "{ [ AddDoc]: 4} : * ",
        "CloseIndex",
    };
    
    // 2. execute the algorithm  (required in every "logic" test)
    Benchmark benchmark = execBenchmark(algLines);


    // 3. test number of docs in the index
    IndexReader ir = IndexReader.open(benchmark.getRunData().getDirectory());
    int ndocsExpected = 2 * 20; // Reuters20DocMaker exhausts after 20 docs.
    assertEquals("wrong number of docs in the index!", ndocsExpected, ir.numDocs());
    ir.close();
  }

View Full Code Here


  // create the benchmark and execute it. 
  public static Benchmark execBenchmark(String[] algLines) throws Exception {
    String algText = algLinesToText(algLines);
    logTstLogic(algText);
    Benchmark benchmark = new Benchmark(new StringReader(algText));
    benchmark.execute();
    return benchmark;
  }

View Full Code Here

        "  CloseIndex",
        "} : 2",
    };
    
    // 2. execute the algorithm  (required in every "logic" test)
    Benchmark benchmark = execBenchmark(algLines);


    // 3. test number of docs in the index
    IndexReader ir = IndexReader.open(benchmark.getRunData().getDirectory());
    int ndocsExpected = 20; // Reuters20DocMaker exhausts after 20 docs.
    assertEquals("wrong number of docs in the index!", ndocsExpected, ir.numDocs());
    ir.close();
  }

View Full Code Here

        "  CloseIndex(false)",
        "} : 2",
    };
    
    // 2. execute the algorithm  (required in every "logic" test)
    Benchmark benchmark = execBenchmark(algLines);


    // 3. test number of docs in the index
    IndexReader ir = IndexReader.open(benchmark.getRunData().getDirectory());
    int ndocsExpected = 20; // Reuters20DocMaker exhausts after 20 docs.
    assertEquals("wrong number of docs in the index!", ndocsExpected, ir.numDocs());
    ir.close();
  }

View Full Code Here

        "  CreateIndex",
        "  { \"AddDocs\"  AddDoc > : * ",
        "} : 2",
    };
    // 2. execute the algorithm  (required in every "logic" test)
    Benchmark benchmark = execBenchmark(algLines);


    assertTrue("did not use the specified MergeScheduler", ((MyMergeScheduler) benchmark.getRunData().getIndexWriter().getMergeScheduler()).called);
    benchmark.getRunData().getIndexWriter().close();


    // 3. test number of docs in the index
    IndexReader ir = IndexReader.open(benchmark.getRunData().getDirectory());
    int ndocsExpected = 20; // Reuters20DocMaker exhausts after 20 docs.
    assertEquals("wrong number of docs in the index!", ndocsExpected, ir.numDocs());
    ir.close();
  }

View Full Code Here

        "  { \"AddDocs\"  AddDoc > : * ",
        "} : 2",
    };


    // 2. execute the algorithm  (required in every "logic" test)
    Benchmark benchmark = execBenchmark(algLines);
    assertTrue("did not use the specified MergeScheduler", ((MyMergePolicy) benchmark.getRunData().getIndexWriter().getMergePolicy()).called);
    benchmark.getRunData().getIndexWriter().close();
    
    // 3. test number of docs in the index
    IndexReader ir = IndexReader.open(benchmark.getRunData().getDirectory());
    int ndocsExpected = 20; // Reuters20DocMaker exhausts after 20 docs.
    assertEquals("wrong number of docs in the index!", ndocsExpected, ir.numDocs());
    ir.close();
  }

View Full Code Here

        "  NewRound",
        "} : 2",
    };


    // 2. execute the algorithm  (required in every "logic" test)
    Benchmark benchmark = execBenchmark(algLines);
    final IndexWriter writer = benchmark.getRunData().getIndexWriter();
    assertEquals(2, writer.getMaxBufferedDocs());
    assertEquals(IndexWriter.DISABLE_AUTO_FLUSH, (int) writer.getRAMBufferSizeMB());
    assertEquals(3, writer.getMergeFactor());
    assertFalse(writer.getUseCompoundFile());
    writer.close();
    Directory dir = benchmark.getRunData().getDirectory();
    IndexReader reader = IndexReader.open(dir);
    TermFreqVector [] tfv = reader.getTermFreqVectors(0);
    assertNotNull(tfv);
    assertTrue(tfv.length > 0);
    reader.close();

View Full Code Here

0 1 2 3

TOP

Related Classes of org.apache.lucene.benchmark.byTask.Benchmark

org.apache.lucene.benchmark.BenchmarkTestCase

org.apache.lucene.benchmark.byTask.tasks.alt.AltPackageTaskTest

org.apache.lucene.benchmark.byTask.TestPerfTasksLogic

org.apache.lucene.benchmark.byTask.utils.Algorithm

org.apache.lucene.benchmark.byTask.utils.Config

java.io.FileReader

java.io.File

All source code are property of their respective owners. Java is a trademark of Sun Microsystems, Inc and owned by ORACLE Inc. Contact coftware#gmail.com.