tut/Set2_NewLinearOperator/SConstruct

Go to the documentation of this file.

"""/*!
@page ExampleSet2 Integrate a new Linear Operator

This examples is for a SLIMpy user who is familiar with the SLIMpy 
vector and linear operator API.
In this example I will go through the steps to create a new linear operator in SLIMpy.

\par Learing objectives:
 - Add a new Linear Operator to SLIMpy  
 - Gain more understanding about SLIMpy's AST


\par Inputs:
 - None so far
\par Python Modules:
 - new_fft_slimpy.py contains a slimpy fft linear operator 
 - new_fft_rsf_integration.py contains the rsf definition of the fft operator

\par Outputs:
 - None so far
\par Prerequisite:
 - SLIMpy and ContribSLIMpy
 - Madagascar
 - SCons

@section Walkthrough
There are two parts to createing a linear operator in SLIMpy: 
 - A Python class that represents the operator in SLIMpy, found in new_fft_slimpy.py
 - A Python class that defines how a plugin may use the operator, found in new_fft_rsf_integration.py

In this example the operator class will be an fft transfrom and the plugin definition will be RSF.
We need to import both of those python modules to get a working linear operator in RSF. if we neglected to
import new_fft_rsf_integration.py the code would give us an error when it tries to execute the fft1X executable
which does not exist. Since fft1X is not defined in SLIMpy's RSF plugin, SLIMpy uses its best 
guess on how to use fft1X with RSF. 

Try it yourself in interactive python mode with the fft1X do not import new_fft_rsf_integration. You will
see that fft1X is a fully formed linear operotor. Without the rsf information SLIMpy will not give you error 
checking when the operator is created or applied, and the domain is known but the range is not. the range of the 
operator will become an instance of the voidSpace class.
*/"""

import new_fft_rsf_integration
from new_fft_slimpy import fft1X  
from slimproj import *

from rsfproj import *

@slim_builder_simple
def new_fft_bldr( vectors ):
    """
    use our new fft operator
    """
    z = vectors[0] 

    
    F = fft1X( z.space, sym=True, opt=False )
    
    x = F*z
    
    print x
    print repr(x.space)
    import pdb
    pdb.set_trace()
    nrm = x.norm( 1 )
    
    tmp = x - nrm
    
    res = F.adj( ) * tmp
    
    return res

Flow( 'newfoo', None, 'sfsigmoid n1=128 n2=128' )

new_fft_bldr( 'res', 'newfoo' )
    

---