jit.expr issue

    Sep 11 2009 | 2:15 pm
    I am working with a a 2 plane char matrix and trying to convert pix to world range inside matrix land.
    I can't get results with this expr: [jit.expr @expr "(in[0].p[0] / 320.)*2. -1." "(in[0].p[1] / 240.)*2. -1."]
    any idea?

    • Sep 11 2009 | 6:19 pm
      not sure exactly what you are trying to do. the char matrix is spitting out values between 0 and 255. if you want to convert that to float, between 0. and 1., simply convert the type of the matrix to float32:
      jit.expr 2 float32 4 1 @adapt 0 @expr in[0] in[0]
      looks like maybe you want 0 255 to -1 to 1:
      jit.expr 2 float32 4 1 @adapt 0 @expr (in[0]*2)-1 (in[0]*2)-1
      not sure why you have 320 and 240 so maybe you can explain more what you're trying to achieve.
    • Sep 11 2009 | 7:31 pm
      Sorry, there was an error in my statement. Yes, of course a char matrix goes up to 255.
      I was trying to convert float32 matrix filled with values between 0-320. for plane 0 and 0-240 for plane 1 to world coordinates. That is good now.
      However I'd like now to tell my expr to convert plane 1 from 0. 240. to 1. -1. not -1. 1.
      My patch allows to sort elmts from cv.jit.sort and fill a new matrix without "cell-jump".
      I tryed (-in[0].p[1] / 240.)*2. -1. (in[0].p[1] / -240.)*2. -1.
      But don't really know how to do this ..
      Thank you.
    • Sep 11 2009 | 7:45 pm
    • Sep 11 2009 | 8:01 pm
      Yes, Thanks.
      But is there a reason for [jit.expr] does'nt accept the "" in my second expression?
    • Jan 06 2010 | 10:08 am
      Yes, thanks ej.
      Other, I discovered that exprfill message to [jit.matrix] do not work as expr message do with [jit.expr]. I thought that they both would work the same way? But no?
      It would be nice to have clarifications with expression syntaxe and features. A special tutorial somewhere would be nice.
    • Jan 06 2010 | 5:38 pm
      Writing an in-depth jit.expr/exprfill tutorial has been on my "Todo Someday" list for some time now. Some of the later Jitter Recipes in Book 3 make use of it to draw geometry.
      For your particular problem, exprfill can't understand the "in[0]" because, well the exprfill is a one time action on a jit.matrix. If you need a dynamic expression with inputs, you are better off using jit.expr.
      Also, if you want to fill different planes with different expressions, you need to use separate exprfill messages with a plane argument before the expression, i.e. (exprfill 0 snorm[0],exprfill 1 snorm[1]).
      Andrew B.
    • Jan 06 2010 | 8:15 pm
      indeed, it would be fantastic to have a jit.expr tutorial as it is one of the most powerful objects within jitter. Please keep us informed if it happens!