1

我需要在 5 秒内创建 100mb 压缩文件，其中包含使用 java 的 CSV 文件。我创建了包含 CSV 文件的 test.zip，但生成 zip 文件需要太多时间（约 30 秒）。这是我到目前为止编写的代码：

ByteArrayOutputStream baos = new ByteArrayOutputStream();
/* Create instance of ZipOutputStream to create ZIP file. */
ZipOutputStream zipOutputStream = new ZipOutputStream(baos);

/* Create ZIP entry for file.The file which is created put into the
 * zip file.File is not on the disk, csvFileName indicates only the
 * file name to be put into the zip
 */
ZipEntry zipEntry = new ZipEntry("Test.zip");

zipOutputStream.putNextEntry(zipEntry);

/* Create OutputStreamWriter for CSV. There is no need for staging
 * the CSV on filesystem . Directly write bytes to the output stream.
 */
BufferedWriter bufferedWriter = new BufferedWriter(new OutputStreamWriter(zipOutputStream, "UTF-8"));

CsvListWriter csvListWriter = new CsvListWriter(bufferedWriter, CsvPreference.EXCEL_PREFERENCE);

/* Write the CSV header to the generated CSV file. */
csvListWriter.writeHeader(CSVGeneratorConstant.CSV_HEADERS);

/* Logic to Write the content to CSV */
long startTime = System.currentTimeMillis();

for (int rowIdx = 0; rowIdx < 7000000; rowIdx++) {
    final List<String> rowContent = new LinkedList<String>();
    for (int colIdx = 0; colIdx < 6; colIdx++) {
        String str = "R" + rowIdx + "C" + colIdx + " FieldContent";
        rowContent.add(str);
    }
    csvListWriter.write(rowContent);
}
long stopTime = System.currentTimeMillis();
long elapsedTime = stopTime - startTime;
System.out.println("time==" + elapsedTime / 1000f + "Seconds");

System.out.println("Size=====" + baos.size() / (Math.pow(1024, 2)) + "MB");

csvListWriter.close();
bufferedWriter.close();
zipOutputStream.close();
baos.close();

我正在使用超级 csv 库，但我也尝试在没有超级 csv 库的情况下在内存中创建 zip 文件但没有成功。你能帮我么？

4

1 回答 1

3

你的测试数据大约是 1GB，压缩到 100MB。根据您的硬件，可能无法实现 < 5s 的性能。

我整理了一个快速而肮脏的基准，突出了写入 zip 文件的性能影响。

  • 写入 CSV String.join()：9.6 秒
  • 使用 Super CSV 写入 CSV：12.7 秒
  • 在 zip 中写入 CSV String.join()：18.6 秒
  • 使用 Super CSV 在 zip 中写入 CSV：22.5 秒

使用 Super CSV (~122%) 似乎有一点开销，但无论是否使用 Super CSV，仅写入 zip 文件的时间几乎翻了一番 (~190%)。

这是 4 个场景的代码。

与您提供的代码不同，我直接写入文件（我没有注意到写入磁盘与写入内存之间有任何区别，即ByteArrayOutputStream）。我也跳过了BufferedWriterSuper CSV 示例，因为它已经在内部使用了它，并且我使用了 try-with-resources 来使事情变得更干净。

@Test
public void testWriteToCsvFileWithSuperCSV() throws Exception {
    long startTime = System.currentTimeMillis();

    try (FileOutputStream csvFile = new FileOutputStream(new File("supercsv.csv"));
         ICsvListWriter writer = new CsvListWriter(new OutputStreamWriter(csvFile, "UTF-8"), CsvPreference.EXCEL_PREFERENCE)
    ){
        for (int rowIdx = 0; rowIdx < 7000000; rowIdx++) {
            final List<String> rowContent = new LinkedList<>();
            for (int colIdx = 0; colIdx < 6; colIdx++) {
                String str = "R" + rowIdx + "C" + colIdx + " FieldContent";
                rowContent.add(str);
            }
            writer.write(rowContent);
        }
    }

    long stopTime = System.currentTimeMillis();
    long elapsedTime = stopTime - startTime;
    System.out.println("Writing to CSV with Super CSV took " + (elapsedTime / 1000f) + " seconds");
}

@Test
public void testWriteToCsvFileWithinZipWithSuperCSV() throws Exception {
    long startTime = System.currentTimeMillis();

    try (FileOutputStream zipFile = new FileOutputStream(new File("supercsv.zip"));
         ZipOutputStream zos = new ZipOutputStream(zipFile);
         ICsvListWriter writer = new CsvListWriter(new OutputStreamWriter(zos, "UTF-8"), CsvPreference.EXCEL_PREFERENCE)
    ){

        ZipEntry csvFile = new ZipEntry("supercsvwithinzip.csv");
        zos.putNextEntry(csvFile);

        for (int rowIdx = 0; rowIdx < 7000000; rowIdx++) {
            final List<String> rowContent = new LinkedList<>();
            for (int colIdx = 0; colIdx < 6; colIdx++) {
                String str = "R" + rowIdx + "C" + colIdx + " FieldContent";
                rowContent.add(str);
            }
            writer.write(rowContent);
        }
    }

    long stopTime = System.currentTimeMillis();
    long elapsedTime = stopTime - startTime;
    System.out.println("Writing to CSV within zip file with Super CSV took " + (elapsedTime / 1000f) + " seconds");
}

@Test
public void testWriteToCsvFileWithStringJoin() throws Exception {
    long startTime = System.currentTimeMillis();

    try (FileOutputStream textFile = new FileOutputStream(new File("join.csv"));
         BufferedWriter writer = new BufferedWriter(new OutputStreamWriter(textFile, "UTF-8"));
    ){

        for (int rowIdx = 0; rowIdx < 7000000; rowIdx++) {
            final List<String> rowContent = new LinkedList<>();
            for (int colIdx = 0; colIdx < 6; colIdx++) {
                String str = "R" + rowIdx + "C" + colIdx + " FieldContent";
                rowContent.add(str);
            }
            writer.append(String.join(",", rowContent) + "\n");
        }
    }

    long stopTime = System.currentTimeMillis();
    long elapsedTime = stopTime - startTime;
    System.out.println("Writing to CSV with String.join() took " + (elapsedTime / 1000f) + " seconds");
}

@Test
public void testWriteToCsvFileWithinZipWithStringJoin() throws Exception {
    long startTime = System.currentTimeMillis();

    try (FileOutputStream zipFile = new FileOutputStream(new File("join.zip"));
         ZipOutputStream zos = new ZipOutputStream(zipFile);
         BufferedWriter writer = new BufferedWriter(new OutputStreamWriter(zos, "UTF-8"));
    ){

        ZipEntry csvFile = new ZipEntry("joinwithinzip.csv");
        zos.putNextEntry(csvFile);

        for (int rowIdx = 0; rowIdx < 7000000; rowIdx++) {
            final List<String> rowContent = new LinkedList<>();
            for (int colIdx = 0; colIdx < 6; colIdx++) {
                String str = "R" + rowIdx + "C" + colIdx + " FieldContent";
                rowContent.add(str);
            }
            writer.append(String.join(",", rowContent) + "\n");
        }
    }

    long stopTime = System.currentTimeMillis();
    long elapsedTime = stopTime - startTime;
    System.out.println("Writing to CSV within zip with String.join() took " + (elapsedTime / 1000f) + " seconds");
}
于 2015-09-17T00:01:22.813 回答