can a filter return a value to the script?

[hanfrunz]

Hello,

is it possible, that a filter returns a value to a script?
I'd like to code a keyframe-filter (with GUI) so you can set marks in the video (keyframes=KF). For every KF you can choose a filtersetting (=the value will be returned)

example:
i set two KF at frame 200 and 500 (clip is 1000f long):

KF1: frame=200, filtersetting 1
KF2: frame=500, filtersetting 2

my script:

clip=AVISource("movie.avi")
setting=Keyframefilter(clip,"Keyframelist.txt",s0="do nothing",s1="blur",s2="sharpen")
if (setting=00) then v=clip
if (setting=01) then v=clip.blur(1.0)
if (setting=02) then v=clip.sharpen()
return v

So:
frame [000 - 199] stay untouched
frame [200 - 499] get blured
frame [500 - end] get sharper

hanfrunz

[sh0dan]

ConditionalReader should do half of what you want.

ConditionalFilter should do the other half.

v = Xsource()
blurred = blur(v,1.0)
sharpened = blur(v,-1.0)
conditionalfilter(v,blurred,v,"blurmode", "=", "1")
conditionalfilter(last,sharpened,last,"blurmode", "=", "2")
conditionalreader("blursettings.txt", "blurmode")

blursettings.txt:

type int
default 0
R 200 499 1
R 500 1499 2

[hanfrunz]

okay i have to do the textfile thing...
mmh then my filter could generate a complete script.

maybe possible in a new version ?

hanfrunz

[sh0dan]

... but doesn't it do exactly what you describe?

[hanfrunz]

yes of course it does, but will i see the changes immediately, when i shuttle through the video? Mmh is this done with my first idea? Or is it possible to put a avisynth-script in a filter (like in ffdshow)?

That would be interesting...

[Fizick]

I have similar question.
Can a one filter (my own source) get some info about every frame from other my filter (or other instance of first filter)?
Not through a file.

[Bidoche]

@Fizick

You can code filters so one will directly (ie without script knowledge) get info from another one you made.

[Fizick]

I can... Good.
But how? any example or MORE detailed info, please.

[Si]

Have a look here

neuron2 invented a system for his Decomb package and I've also copied it for PFR to pass info to KernelDeInt.

All code is available at our sites.

regards

Simon

[Bidoche]

@Fizick

Neuron2 uses extra space in frame buffer to hold his own data, ... as a workaround for the lack of frame properties.
It works fine as long as there is no filter between which may alter your metadata. (Filters are NOT required to leave the extra space alone, and in fact they often write on it)

If that restriction is unacceptable to you, you can do it like this : (STL code)

#include <map>     //for map
#include <string>  //for string
#include <utility> //for pair

class InterrogatedFilter
{

  typedef std::map<pair<IScriptEnvironment *, std::string>, Clip *> InstanceMap;

  static InstanceMap map_;

  InterrogatedFilter(std::string const& identif, IScriptEnvironment *env) 
  {
     map_[ std::make_pair(env, identif) ] = this;  //reg self into map
  }

  ~InterrogatedFilter()
  {
     for(InstanceMap::iterator it = map_.begin(); it != map_.end(); ++it )
       if ( it->second == this )
       {
         map_.erase(it);
         break;
       }
     //remove the entry corresponding to this from the instance map
  } 

  static PClip FindInstance(std::string const& identif, IScriptEnvironment *env)
  {
    InstanceMap::iterator it = map_.find( std::make_pair(env, identif) );
    return it != map.end() ? PClip(it->second) : PClip();
  }

  //Clip methods and interrogation methods ....

};//InterrogatedFilter


class InterrogatingFilter
{
  //use InterrogatedFilter::FindInstance somewhere to fetch the (right) clip to interrogate...
  //that means you must have the same identif value here to use
};

NB: this is not perfect (not thread safe for example)

[Fizick]

Thanks!
I shall try that strange method.

Corrected: methodS.

[Bidoche]

Well, I would have called using an instance map a classic solution.

Anyway, if you have trouble doing that, feel free to ask.