JPEXS Free Flash Decompiler Issue Tracker

You made a great program here, thanks to it I managed to make a script that extracts some images from gameswf files (like attached above). But still not all of them, I'm still struggling with some values - can't comprehend what is DeltaX, where is starting drawing position which moved by moveDeltaX, how exactly rotateSkew works - official "SWF FILE FORMAT SPECIFICATION" document is kinda vague about these providing only theory without real examples. When there is only "scale" and "translate" values at play, there are no problems to match position of RECT (taking into account provided shapeBounds), but when there is rotation involved or shape is a complex one (consisting of two or more images) having shapeRecords in place - I fail miserably. I get any results except for the right ones. Take for example DefineShape2 (15) from attached file - I even created copy of the source image in graphics editor (I included it in the file, originally it is a separate image), scaled it according to scaleX/Y values, drawed RECT at shapeBounds, moved it to translateX/Y and tried to move it by moveDeltaX/Y - couldn't get proper cutting positions for neither of parts of that shape. If there is rotateSkew0/1 involved (DefineShape (212) for example) - things get even more complicated, I'm not even close to a proper result... Could you help me in this regard by any chance? Maybe provide links to some guides or something? I can't google anything useful and close to a desperation here.

To the original problem: in nightly 2772, the FB type values are displayed/edited as floats. To your second message: BitmapMatrix field is transformation matrix, you can find some methods regarding it in our source code: https://github.com/jindrapetrik/jpexs-decompiler/blob/master/libsrc/ffdec_lib/src/com/jpex s/decompiler/flash/exporters/commonshape/Matrix.java (methods transform for example) moveDeltaX in StyleChangeRecord will move cursor to desired position to stroke or fill path. The path is made of shape records. Converting the whole shape to actual bitmap with all parts correctly filled/stroked is very complex operation. I am not sure I can explain it properly. This is a class we use for it: https://github.com/jindrapetrik/jpexs-decompiler/blob/master/libsrc/ffdec_lib/src/com/jpex s/decompiler/flash/exporters/shape/BitmapExporter.java

That's exactly the reason I've asked - I just didn't get it. Yes, it's affine transformation of 3x3 matrix where third row is basically ignored. OK, there's quite simple formula for that. But it doesn't work for me somehow. Let's take attached file for example - it's the same picture which used in the file in initial message above. There are 15 sets of stars - all of them use RotateSkew. Let's take DefineShape 238 - it's the topmost of such parts here. I want to cut it out. Should get coordinates which lay somewhat in ranges: X0(1400,1440), Y0(200,240), X1(1805, 1820), Y1(565, 600) - based on surrounding objects, right? What do we have: shapeBounds Xmin = -6432, Xmax = 24000 Ymin = -7271, Ymax = 19439 matrix: sX = 37.15 rs1 = 38.93 tX = -66898 rs0 = -34.17 sY = 32.61 tY = 48262 Transformed coordinate: X' = X * ScaleX + Y * RotateSkew1 + TranslateX (swf spec. confirms it). Though we are looking for original coordinate, not for the transformed one. And as such: X * ScaleX = X' - Y * RotateSkew1 - TranslateX X = (X' - Y * RotateSkew1 - TranslateX) / ScaleX We don't know value of Y here, so we can get it from: Y' = X * RotateSkew0 + Y * ScaleY + TranslateY And then expand to this monster: X = (X'sY - Y'rs1 + tYrs1 - tXsY)/(sXsY - rs0rs1) Or just calculate inverse of the matrix and apply: X0/1 = Xmin/max * ScaleX_inv + Ymin/max * RotateSkew1_inv + TranslateX_inv Both methods will return same result. But here's the thing - result is not what I wanted. X is (1626, 1607), Y is (1, 800) which is far from boundaries above. I even tried to pick values by hand - and if one of boundaries will be set in place (for example, X0 would be 1403) then the other one will certainly be set off (X1 is 2203). Or vice versa (1550 and 1810) - just from changing one value (RotateSkew1) When there is no RotateSkew involved (is zero) - all coordinates are pretty much to the point from Scale and Translate. Suggested source code didn't help me - basically it contains same math which I found elsewhere in internet. Did I miss some conversion along the way or I'm just plainly dumb and can't see something obvious?