Ce site est optimisé pour être consulté depuis un navigateur moderne dans lequel JavaScript est activé.

Binary format loader for C Runtime

vblack

I've seen the C# loader for binary format.

Here the port for C++ of the SkeletonBinary format. I don't have access to Github, but if you want to use it.

To use it
skeletonData = spSkeletonJson_readSkeletonBinaryFile(json, skeletonDataFile); instead of spSkeletonJson_readSkeletonDataFile for .skel files

Tested with animation and with mesh as well. It's 10x faster at least for loading.

If you can considers to adapt it / add to github.

Cheers;

Nate

Nice! Unfortunately to go in spine-c it has to be ANSI C rather than C++. If you don't want to convert, I will give it a go when I find some time.

vblack

Yes it was a quick port for proof of concept (from the C# code). It won't be too difficult to adapt for C ANSI . I will try to adapt as well during the week.

Loomind

I have Visual Studio 2010 that compile well you're ANSI C code to C++. Unfortunately there's some unsecure functions(). Did you plan to fix them ? As I know if VS2013 compile it fine, it will works alone with Unreal engine, not only with SFML.

jpoag

vblack a écrit
Yes it was a quick port for proof of concept (from the C# code). It won't be too difficult to adapt for C ANSI . I will try to adapt as well during the week.

For anyone who hasn't already guessed it, ANSI-C is harder to code for coming from a C++ background/direction.

Loomind a écrit
I have Visual Studio 2010 that compile well you're ANSI C code to C++. Unfortunately there's some unsecure functions(). Did you plan to fix them ? As I know if VS2013 compile it fine, it will works alone with Unreal engine, not only with SFML.

You sound confused. Code is C++ and needs to be C to go in official github. No one has trouble compiling code.

choigawoon

FFD bug fixed in ReadAttachment function.

from
TO_ARRAY(&mesh->regionUVs, uvs);
to
TO_ARRAY(&mesh->regionUVs, &mesh->uvscount, uvs);

and there is another issue with this implement with ffd animation.
in my project, the interpolation of a animation between timelines is not applied.
it seems the animation stepped.
I'm trying to compare the implement of this post with the implement of libgdx.

Pharan

Thanks for letting us know.

When you say "interpolation of a animation between timelines is not applied" and "the animation stepped", do you mean FFD keys are not being interpolated?
Comparing with the libGDX implementation is a good idea. Check Animation.java.

choigawoon

first of all, thanks to reply.
i'm just feeling chaos because of this problem. it's almost time to ship my project, but i'm just wandering now.
one of my exports takes too much time to initialize, almost 5 mins or more,
and the only solution is using the binary loader in my thinking.

When you say "interpolation of a animation between timelines is not applied" and "the animation stepped", do you mean FFD keys are not being interpolated?

>> No, It's not when using this binary loader but It was good when using the json loader. You correctly understand what i'm telling.

Comparing with the libGDX implementation is a good idea. Check Animation.java.

>> It might be ok there. i'm using the ANSI C implementation of spine-runtime, and it was good when using json loader.

I tried to find out what happened in the binary loader when loading timelines of ffd.
I assumed that this binary loader screwed up the timelines of ffd( especially, skinned mesh. but region attachment, boundingbox attachment, events works well )
now i have a mind to make one of my own binary loader as referencing the libgdx, but it's not easy. and i cannot find it yet.
i'm beginner in the area of skeleton animation and also spine.
sometimes i can't get a right term for describing the problem. so if you cannot understand, or if you have a any idea about this, reply to me, plz. it can be helpful for me.
anyway, thanks a lot.

Nate

You know SkeletonJson is correct, so load the same data as JSON and also as binary and print it out. It may be easiest to edit SkeletonJson/Binary and add print statements. Put the data side by side and look for major differences. There will be small differences with floats that aren't important. Probably you'll see something significantly different, then you'll know where to look to fix it.

choigawoon

i've just found the bug with all values logging. and i found that all timelines are messed up.

in libgdx

private void readCurve (DataInput input, int frameIndex, CurveTimeline timeline) throws IOException {
      switch (input.readByte()) {
      case CURVE_STEPPED:
         timeline.setStepped(frameIndex);
         break;
      case CURVE_BEZIER:
         setCurve(timeline, frameIndex, input.readFloat(), input.readFloat(), input.readFloat(), input.readFloat());
         break;
      }
   }

in this implement

void ReadCurve (BufferedStream& input, int frameIndex, spCurveTimeline* timeline) {
            switch (input.ReadByte()) {
                case CURVE_STEPPED:
                    spCurveTimeline_setStepped(timeline, frameIndex);
                    break;
            

            case CURVE_BEZIER:
                spCurveTimeline_setCurve(timeline, frameIndex, ReadFloat(input), ReadFloat(input), ReadFloat(input), ReadFloat(input));
                break;
        }
    }

both implementation look same but, it's not because of calling convention.
c/c++ normally works as stdcall. and it makes pass the values from right to left.
so, the last 4 arugments were reversed. everything in timelines was messed up.
I did manually call function before calling spCurveTimeline_setCurve and assigned the values.

now i fixed it. thanks for all helping. have a nice day!

Pharan

Good to know! Thanks for sharing!

Nate

Yikes. Actually it seems that the order the ReadFloat calls are executed is unspecified! Good find!

vblack

Hello;

In V3, it seems that boneData::flipX,flipY,inheritScale, inheritRotation are no more written to the file (this is breaking the loader, because it reads the 4 boolean and it offsets the whole file.).

Is it intentional ?

Nate

Sorry, this was a bug. It is now fixed in the latest version.

vblack

Seems better in 3.0.04 but still two bytes missing instead of four bytes (inheritScale and rotation maybe).

Nate

Ah, those 2 bytes are gone because flip has been removed. Only spine-libgdx has been updated to support v3 so far. The other runtimes have not yet been updated.

vblack

Hi;
I've adapter my C++ loader (which also backward compatible to 2.0 to 3.1.05).

However I've still have an issue with reading the draw order timeline. I have the correct number of drawOrderCount, but in line 633 (from the Java version), I have mismatched offsetCount version, like 61 for 7 slotCount.

Can you check that ?
Cheers;;

Pharan

Just making sure, are you reading the count as a varint and not just an int?

vblack

Yes it is using for a while.

I have attached the latest code of my C++ loader . It has

Backward support from 2.0 to 3.1.05 (parsing the version).
3.1.05, works excepted for keyframed drawOrder : I suspected a bug in the exporter.
Include latest fix, including the call order different from C# and C++

Note: I have a .skel, maybe Nate can check it out.

OK found why, thanks to the documented format in http://esotericsoftware.com/spine-binary-format
The bug is present in C#, and Java as well

https://github.com/EsotericSoftware/spine-runtimes/blob/master/spine-csharp/src/SkeletonBinary.cs#L628
https://github.com/EsotericSoftware/spine-runtimes/blob/master/spine-libgdx/spine-libgdx/src/com/esotericsoftware/spine/SkeletonBinary.java#L636

timeline.SetFrame(i, ReadFloat(input), drawOrder);

ReadFloat should be at the beginning of the loop

"
For each draw order keyframe:
float time: The time in seconds for the keyframe.
"

In the 3.1.05, the time is read before the offsetCount (I renamed to changeCount);
I updated the file, it is now working


for (int i = 0; i < drawOrderCount; i++) {
        float time;
        int changeCount;
        if (version >= 3105)
        {
                 time = ReadFloat(input); // The time in seconds for the keyframe. Now read at first
        }
        changeCount = ReadVarint(input, true);

Page suivante »