Java創建100MB壓縮的CSV文件性能問題

Question

我需要在5秒鍾內創建100mb的壓縮文件，其中包含使用java的CSV文件。 我創建了包含CSV文件的test.zip，但是生成zip文件花費了太多時間（〜30秒）。 這是我到目前為止編寫的代碼：

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();

我正在使用超級csv庫，但我也嘗試在沒有超級csv lib的情況下在內存中創建zip文件，但未成功。 你能幫我么？

Answer 1

您的測試數據約為1GB，壓縮到100MB。 根據您的硬件，可能無法達到<5s的性能。

我整理了一個快速而骯臟的基准，重點介紹了寫入zip文件對性能的影響。

• 用String.join()寫入CSV：9.6s
• 使用超級CSV寫入CSV：12.7秒
• 使用String.join()在zip中寫入CSV：18.6s
• 使用Super CSV在zip內寫入CSV：22.5秒

似乎使用Super CSV會有一點開銷（〜122％），但是無論是否使用Super CSV，僅寫入zip文件幾乎都會花費兩倍的時間（〜190％）。

這是這四個方案的代碼。

與您提供的代碼不同，我直接寫入文件（我注意到寫入磁盤與寫入內存（即ByteArrayOutputStream ）之間沒有任何區別）。 我還跳過了Super CSV示例中的BufferedWriter ，因為它已經在內部使用了它，並且我使用了try-with-resources使內容更整潔。

@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");
}