qgis - Processing algorithm fails from Python script

I'm trying to write a Python script performing a geoalgorithm. What is surprising me is the following:

1. I first test the algorithm by QGIS (2.8) interface. In my case, the GRASS interpolator v.surf.idw


2. I see that the solution is sufficiently good using a certain setting of parameters.


3. 
   
   

   Then, I run the same algorithm with the same parameters, froma Python script. In my case:

   
   

   out_ras = processing.runalg("grass:v.surf.idw", vl,12,2,"field_3" ,False, "%f , %f, %f, %f "% (xmin , xmax , ymin , ymax), 0.5, -1, 0.001, fileoutput)
   

where:

• vl is the point vector layer


• field_3 is the filed where vlaues to be interpolated


• fileoutput is the raster file in output


• (xmin, , xmax , ymin , ymax) are the Extent of my layer

This setting (perfectly working when launched from QGis interface) produices a Nodata value Raster (only 1 cell). It seems that the algorithm does not recognize the vector in input. I've also checked the CRS of the layer (with vl.crs().authid() ) and everything sounds good.

Any help? Any experience in detecting different behaviour of the SAME algorithm run by Python through processing instead of from QGIS UI ?