java - 从 CSV 文件读取速度慢

Question

我正在尝试读取 csv 文件，但速度很慢。下面是代码的粗略解释:

private static Film[] readMoviesFromCSV() {
    // Regex to split by comma without splitting in double quotes.
    // https://regexr.com/3s3me <- example on this data
    var pattern = Pattern.compile(",(?=(?:[^\\\"]*\\\"[^\\\"]*\\\")*[^\\\"]*$)");
    Film[] films = null;
    try (var br = new BufferedReader(new FileReader(FILENAME))) {
        var start = System.currentTimeMillis();
        var temparr = br.lines().skip(1).collect(Collectors.toList());  // skip first line and read into List
        films = temparr.stream().parallel()
                .map(pattern::split)
                .filter(x -> x.length == 24 && x[7].equals("en")) // all fields(total 24) and english speaking movies
                .filter(x -> (x[14].length() > 0)) // check if it has x[14] (date)
                .map(movieData -> new Film(movieData[8], movieData[9], movieData[14], movieData[22], movieData[23], movieData[7]))
                // movieData[8] = String title, movieData[9] = String overview
                // movieData[14] = String date (constructor parses it to LocalDate object)
                // movieData[22] = String avgRating
                .toArray(Film[]::new);
        System.out.println(MessageFormat.format("Execution time: {0}", (System.currentTimeMillis() - start)));
        System.out.println(films.length);
    } catch (FileNotFoundException e) {
        e.printStackTrace();
    } catch (IOException e) {
        e.printStackTrace();
    }
    return films;
}

文件大约 30 MB 大，平均需要大约 3-4 秒。我正在使用流，但它仍然很慢。是因为每次的 split 吗？

编辑:我已经成功使用 uniVocity-parsers 库将读取和处理时间加快了 3 倍。平均需要 950 毫秒才能完成。这真是令人印象深刻。

private static Film[] readMoviesWithLib() {
    Film[] films = null;
    CsvParserSettings parserSettings = new CsvParserSettings();
    parserSettings.setLineSeparatorDetectionEnabled(true);
    RowListProcessor rowProcessor = new RowListProcessor();
    parserSettings.setProcessor(rowProcessor);
    parserSettings.setHeaderExtractionEnabled(true);
    CsvParser parser = new CsvParser(parserSettings);
    var start = System.currentTimeMillis();
    try {
        parser.parse(new BufferedReader(new FileReader(FILENAME)));
    } catch (FileNotFoundException e) {
        e.printStackTrace();
    }
    List<String[]> rows = rowProcessor.getRows();
    films = rows.stream()
            .filter(Objects::nonNull)
            .filter(x -> x.length == 24 && x[14] != null && x[7] != null)
            .filter(x -> x[7].equals("en"))
            .map(movieData -> new Film(movieData[8], movieData[9], movieData[14], movieData[22], movieData[23], movieData[7]))
            .toArray(Film[]::new);
    System.out.printf(MessageFormat.format("Time: {0}",(System.currentTimeMillis()-start)));
    return films;
}

Answer 1

这里是 univocity-parsers 库的作者。您可以通过像这样重写来进一步加快您在编辑中发布的代码的速度:

    //initialize an arraylist with a good size to avoid reallocation
    final ArrayList<Film> films = new ArrayList<Film>(20000);
    CsvParserSettings parserSettings = new CsvParserSettings();
    parserSettings.setLineSeparatorDetectionEnabled(true);
    parserSettings.setHeaderExtractionEnabled(true);

    //don't generate strings for columns you don't want
    parserSettings.selectIndexes(7, 8, 9, 14, 22, 23);

    //keep generating rows with the same number of columns found in the input
    //indexes not selected will have nulls as they are not processed.
    parserSettings.setColumnReorderingEnabled(false); 

    parserSettings.setProcessor(new AbstractRowProcessor(){
        @Override
        public void rowProcessed(String[] row, ParsingContext context) {
            if(row.length == 24 && "en".equals(row[7]) && row[14] != null){
                films.add(new Film(row[8], row[9], row[14], row[22], row[23], row[7]));
            }
        }
    });

    CsvParser parser = new CsvParser(parserSettings);
    long start = System.currentTimeMillis();
    try {
        parser.parse(new File(FILENAME), "UTF-8"); 
    } catch (FileNotFoundException e) {
        e.printStackTrace();
    }

    System.out.printf(MessageFormat.format("Time: {0}",(System.currentTimeMillis()-start)));
    return films.toArray(new Film[0]);

为了方便起见，如果您必须将内容处理到不同的类中，您也可以 use annotations在您的 Film 类(class)中。

希望这有帮助。