Apache CSV parser is not working on tab delimited data with quotation marks

Question

I would like to parse a Google eBook transaction report. I opened it in Notepad++ to see precisely the filed and record delimiters. It is a tab delimited file, every header field and data field are surrounded by quotation marks. The first two lines of the CSV file are:

"Transaction Date" "Id"    "Product"   "Type"  "Preorder"  "Qty"   "Primary ISBN"  "Imprint Name"  "Title" "Author"    "Original List Price Currency"  "Original List Price"   "List Price Currency"   "List Price [tax inclusive]"    "List Price [tax exclusive]"    "Country of Sale"   "Publisher Revenue %"   "Publisher Revenue" "Payment Currency"  "Payment Amount"    "Currency Conversion Rate"
"2016. 09. 01." "ID:1166315449551685"   "Single Purchase"   "Sale"  "None"  "1" "9789633780664" "Book and Walk Kft" "Bánk bán"  "József Katona" "HUF"   "0,00"  "HUF"   "0,00"  "0,00"  "HU"    "52,0%" "0,00"  ""  ""  ""

I use the following code to parse the CSV file:

private List<Sales> parseCsv(File csv) {
    Calendar max = Calendar.getInstance();
    Calendar current = Calendar.getInstance();
    boolean firstRound = true;

    List<Sales> sales = new ArrayList<>();
    Sales currentRecord;
    Reader in;
    try {
        in = new FileReader(csv);
        Iterable<CSVRecord> records;

        try {

            records = CSVFormat.TDF.withQuote('\"').withFirstRecordAsHeader().parse(in);
            for (CSVRecord record : records) {
                currentRecord = new Sales();
                currentRecord.setAuthor(record.get("Author"));
                currentRecord.setTitle(record.get("Title"));
                currentRecord.setPublisher(record.get("Imprint Name"));
                currentRecord.setIsbn(record.get("Primary ISBN"));
                currentRecord.setChannel("Google");
                currentRecord.setBookId(record.get("Id"));
                currentRecord.setCountry(record.get("Country of Sale"));
                currentRecord.setUnits(Integer.parseInt(record.get("Qty")));
                currentRecord.setUnitPrice(Float.parseFloat(record.get("List Price [tax exclusive]")));

                Date transDate;
                try {
                    transDate = sourceDateFormat.parse(record.get("Transaction Date"));
                    if (firstRound) {
                        max.setTime(transDate);
                    };
                    current.setTime(transDate);
                    if (current.after(max)) {
                        max.setTime(current.getTime());
                    }
                    currentRecord.setDatum(transDate);
                } catch (ParseException e) {
                    // TODO Auto-generated catch block
                    LOG.log(Level.SEVERE,"Nem megfeelő formátumú a dátum a {0} file-ban",csv.getAbsolutePath());
                }

                currentRecord.setCurrencyCustomer(record.get("List Price Currency"));
                currentRecord.setCurrencyProceeds(record.get("Payment Amount"));
                currentRecord.setCurrencyProceeds(record.get("Payment Currency"));
                sales.add(currentRecord);
            }
            LOG.log(Level.INFO, "Daily sales transactions of {0} were successfully parsed from ",
                    csv.getAbsolutePath());
            return sales;
        } catch (IOException e1) {
            // TODO Auto-generated catch block
            LOG.log(Level.SEVERE, "Valami nem stimmel a {0} file szerkezetével",csv.getAbsolutePath());
        }
    } catch (FileNotFoundException e1) {
        // TODO Auto-generated catch block
        LOG.log(Level.SEVERE,"A {0} file-t nem találom.",csv.getAbsolutePath());
    }
    return null;
};

When I debug the parsing proccess than I can see that record.get("Author") threw runtime exception:

java.lang.IllegalArgumentException: Mapping for Author not found, expected one of [��"

Obviously I have column named Author. Any ide what goes wrong?

Answer 1

When converting this to a unit-test and running with the current commons-csv version 1.4 this works fine for me, therefore:

• Check with the latest version of commons-csv
• Make sure that there are actually TABs in the file, not blanks around the author-entries for some reason
• Specify the actual encoding of the file when calling parse() to correctly handle non-ASCII characters (thanks to comments from @tonakai)

The following unit-test works fine with commons-csv 1.4

private final static String DATA = "\"Transaction Date\"\t\"Id\"\t\"Product\"\t\"Type\"\t\"Preorder\"\t\"Qty\"\t\"Primary ISBN\"\t\"Imprint Name\"\t\"Title\"\t\"Author\"\t\"Original List Price Currency\"\t\"Original List Price\"\t\"List Price Currency\"\t\"List Price [tax inclusive]\"\t\"List Price [tax exclusive]\"\t\"Country of Sale\"\t\"Publisher Revenue %\"\t\"Publisher Revenue\"\t\"Payment Currency\"\t\"Payment Amount\"\t\"Currency Conversion Rate\"\n" +
        "\"2016. 09. 01.\"\t\"ID:1166315449551685\"\t\"Single Purchase\"\t\"Sale\"\t\"None\"\t\"1\"\t\"9789633780664\"\t\"Book and Walk Kft\"\t\"Bánk bán\"\t\"József Katona\"\t\"HUF\"\t\"0,00\"\t\"HUF\"\t\"0,00\"\t\"0,00\"\t\"HU\"\t\"52,0%\"\t\"0,00\"\t\"\"\t\"\"\t\"\"";

@Test
public void parseCsv() throws IOException {
    final CSVFormat format = CSVFormat.TDF.withQuote('\"').withFirstRecordAsHeader();
    Iterable<CSVRecord> records = format.parse(new StringReader(DATA));

    System.out.println("Headers: " + Arrays.toString(format.getHeader()));

    for (CSVRecord record : records) {
        assertNotNull(record.get("Author"));
        assertNotNull(record.get("Title"));
        assertNotNull(record.get("Imprint Name"));
        assertNotNull(record.get("Primary ISBN"));
        assertNotNull(record.get("Id"));
        assertNotNull(record.get("Country of Sale"));
        assertNotNull(record.get("Qty"));
        assertNotNull(record.get("List Price [tax exclusive]"));

        assertNotNull(record.get("Transaction Date"));

        assertNotNull(record.get("List Price Currency"));
        assertNotNull(record.get("Payment Amount"));
        assertNotNull(record.get("Payment Currency"));

        System.out.println("Record: " + record.toString());
    }
}

Answer 2

It turned out that the encoding was the source of the problem. Based on @tonakai comment I started to analyze the encoding of the Google csv report. It was UTF-16 Little Endian. As my file included byte-order mark, I had to use 'BOMInputStream' and refactor my code a little bit.

Reader r = newReader(csv);
CSVParser csvParser= CSVFormat.TDF.withFirstRecordAsHeader().withQuoteMode(QuoteMode.ALL).parse(r);

.....

private InputStreamReader newReader(final File csv) throws FileNotFoundException {
        return new InputStreamReader(new BOMInputStream(new FileInputStream(csv),ByteOrderMark.UTF_16LE), StandardCharsets.UTF_16LE);
    }

It is working now