Apache CSV Commons: IOException: (line 5) invalid char between encapsulated token and delimiter

Apache CSV Commons library interrupted with exception:

IOException: (line 5) invalid char between encapsulated token and delimiter

It tries to read following .csv file:

"id", "category_id", "brand_id", "catalog_number"
"6427146", "4045", "764\"13", "A26-30-01"
"6425052", "4058", "764\"13", "P9B02VN"
"6424406", "4054", "764\"13", "A40-30-10-80"
"6152302", "4046", "764\"13", "1.75\\" center distance"
"6152301", "4046", "764\"13", "ZL110"
"6152300", "4046", "764\"13", "ZAF460-70"
"6152299", "4046", "764\"13", "ZA75-84"
"6152298", "4046", "764\"13", "ZA75-80"
"6152297", "4046", "764\"13", "ZA75-55-1SBN153510R5506"

Library cannot read line #5.

Code:

@Test
public void testReadCsvFile() throws IOException {
    Reader reader = new FileReader("products-with-escaped-escape-symbol.csv");

    Iterable<CSVRecord> records = CSVFormat.DEFAULT
            .withHeader(HEADERS)
            .withFirstRecordAsHeader()
            .withIgnoreHeaderCase()
            .withIgnoreSurroundingSpaces()
            .withEscape('\\')
            .parse(reader);

    for (CSVRecord record : records) {
        String brandId = record.get("brand_id");
        assertThat(brandId, is("764\"13"));
    }
}

Is there a way to adjust Apache CSV Commons library to read a .csv file with such structure?

The best thing you can do is let the source of this data know that they have povided you with invalid input. They are supposed to send a CSV file and they did failed to do that. This input is not a CSV file, it merely somewhat resembles a CSV file.

If that option is not available, you can create your own Reader which filters each line before passing it to the CSVParser:

List<CSVRecord> records;

CSVFormat format = CSVFormat.DEFAULT
        .withHeader(HEADERS)
        .withFirstRecordAsHeader()
        .withIgnoreHeaderCase()
        .withIgnoreSurroundingSpaces()
        .withEscape('\\');

try (PipedReader reader = new PipedReader();
     PipedWriter writer = new PipedWriter(reader)) {

    Runnable filterTask = () -> {
        try (BufferedReader fileReader = Files.newBufferedReader(
                Path.of("products-with-escaped-escape-symbol.csv"));
             PipedWriter filteredWriter = writer) {

            String line;
            while ((line = fileReader.readLine()) != null) {
                line = line.replaceAll(
                    "^(" +
                    "(?:\\s*\"(?:[^\"\\\\]|\\\\.)*\"\\s*,)*" +
                    "\\s*\"(?:[^\"\\\\]|\\\\[^\\\\])*" +
                    ")" +
                    "\\\\\\\\\"($|\\s*[^,])", "$1\\\\\"$2");
                filteredWriter.write(line);
                filteredWriter.write('\n');
            }
        } catch (IOException e) {
            throw new UncheckedIOException(e);
        }
    };

    CompletableFuture<?> filter = CompletableFuture.runAsync(filterTask);

    try (CSVParser parser = format.parse(reader)) {
        records = parser.getRecords();
        filter.get();
    } catch (ExecutionException e) {
        Throwable cause = e.getCause();
        if (cause instanceof RuntimeException re) {
            throw re;
        } else {
            throw new RuntimeException(cause);
        }
    } catch (InterruptedException e) {
        throw new RuntimeException(e);
    }
} catch (IOException e) {
    throw new UncheckedIOException(e);
}

This is not a completely reliable solution. If a comma followed \\" and that comma was intended to be part of the value, rather than a value separator, writing code to recognize that case would be a lot more complicated.