Java reads wrong values from binary structured file

Question

I have a file which consist of Delphi records. The record looks like:

TPoint = record
    X, Y: Double;
end;

As I know Delphi stores data as LittleEndian, and Java as BigEndian so the read method looks like:

public Point readPoint() throws IOException {
        double x = Double.longBitsToDouble(
                                Long.reverseBytes(
                                        Double.doubleToLongBits(in.readDouble()
                                        )
                                )
        );
        double y = Double.longBitsToDouble(
                Long.reverseBytes(
                        Double.doubleToLongBits(in.readDouble()
                        )
                )
        );
        return new Point(x, y);
}

Everything seemed to be fine but sometimes I take the corrupted data. I got

638 offset: 10256   Point{x=3.143E-319, y=48.47134}

But should to get

638 offset: 10256   Point{x=22.25315, y=48.47134}

When I opened the file and look the data with offset 10256 I see:

7F FB 3A 70 CE 40 36 40

which is 22.25315 in LittleEndian. So i wondered what is the problem? Is there any problem with Long.reverseBytes() method? Or it's impossible to convert any doubles from LE to BE and vice versa?

Why are you reading a double to start with? Surely you'd be better off using Long.reverseBytes(in.readLong()). — Jon Skeet
– Jon Skeet, Commented Aug 4, 2014 at 14:02
I have to agree. If you read the double with the bytes in the wrong order, it is possible that there is some normalization that occurs which is changing the bits before you get them to reverse them. — David Conrad
– David Conrad, Commented Aug 4, 2014 at 14:03
Added this as an answer, given that I suspect it's the problem... — Jon Skeet
– Jon Skeet, Commented Aug 4, 2014 at 14:04

Jon Skeet · Accepted Answer · 2014-08-04 14:03:45Z

4

I suspect the problem may be because you're reading 8 bytes as a big-endian double when they're not a big-endian double. Given that bits have specific meanings in double, that could cause problems - Java may be normalizing NaN values, for example. It's a bit like reading text data using an encoding that you know is incorrect, and then converting it back into bytes...

If you read the value as a long to start with, that should preserve all the bits, so you can reverse the bytes and then convert to a double:

double x = Double.longBitsToDouble(Long.reverseBytes(in.readLong()));
double y = Double.longBitsToDouble(Long.reverseBytes(in.readLong()));

answered Aug 4, 2014 at 14:03

Jon Skeet

1.5m893 gold badges9.3k silver badges9.3k bronze badges

Sign up to request clarification or add additional context in comments.

1 Comment

Durandal Over a year ago

doubleToLongBits definetly changes NaN to a canonical representation, its explicitly stated to do so in the javadocs: docs.oracle.com/javase/8/docs/api/java/lang/…

Rayz · Accepted Answer · 2014-08-04 14:10:10Z

1

Have found the bug. Instead of

Long.reverseBytes(Double.doubleToLongBits(in.readDouble())

just needed to read as follows:

Long.reverseBytes(in.readLong())

answered Aug 4, 2014 at 14:10

Rayz

5331 gold badge5 silver badges19 bronze badges

Collectives™ on Stack Overflow

Java reads wrong values from binary structured file

2 Answers 2

1 Comment

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

1 Comment

Comments

Your Answer

Sign up or log in

Post as a guest

Related