TraceWin.py

class dst:
    '''
    Simple class to read in a 
    TraceWin distribution file

    Class afterwards hold the following
    dictionary items:
      - x [m]
      - xp [rad]
      - y [m]
      - yp [rad]
      - phi [rad]
      - E [MeV]
    '''
    def __init__(self, filename):
        # easy storage..
        self.filename=filename
        # used to create dict behaviour..
        self._columns=['x','xp','y','yp','phi','E']
        # read in the file..
        self._readBinaryFile()

    def _readBinaryFile(self):
        # Thanks Emma!

        import numpy

        fin=file(self.filename,'r')

        # dummy, Np, Ib, freq, dummy
        Header_type = numpy.dtype([
            ('dummy12', numpy.int16),
            ('Np',      numpy.int32),
            ('Ib',      numpy.float64),
            ('freq',    numpy.float64),
            ('dummy3',  numpy.int8)
            ])
        Header=numpy.fromfile(fin, dtype=Header_type, count=1)
        self.Np=Header['Np'][0]
        self.Ib=Header['Ib'][0]
        self.freq=Header['freq'][0]

        # Some toutatis distributions has an undocumented 7th line of 0's
        Table=numpy.fromfile(fin, dtype=numpy.float64, count=self.Np*7+1)
        if len(Table)==self.Np*7+1:
            self._data=Table[:-1].reshape(self.Np,7)
        elif len(Table)==self.Np*6+1: # this is true in most cases
            self._data=Table[:-1].reshape(self.Np,6)
        else:
            raise ValueError("Incorrect table dimensions found:",len(Table))

        # convert x,y from cm to m:
        self._data[:,0]*=1e-2
        self._data[:,2]*=1e-2

        self.mass=Table[-1]

    def keys(self):
        return self._columns[:]

    def __getitem__(self, key):
        # makes the class function as a dictionary
        # e.g. dst['x'] returns the x array..
        try:
            i=self._columns.index(key)
            return self._data[:,i]
        except:
            raise ValueError("Available keys: "+str(self._columns))

    def __setitem__(self, key, value):
        try:
            i=self._columns.index(key)
            self._data[:,i]=value
        except:
            raise ValueError("Available keys: "+str(self._columns))
    
    def save(self, filename):
        '''
        Save the distribution file
        so it can be read by TraceWin again

        Stolen from Ryoichi's func.py (with permission)
        '''

        from struct import pack

        fout=open(filename,'w')
        out =pack('b',125)
        out+=pack('b',100)
        out+=pack('i',self.Np)
        out+=pack('d',self.Ib)
        out+=pack('d',self.freq)
        out+=pack('b',125)

        data=self._data.copy()

        # convert x,y from m to cm:
        data[:,0]*=1e2
        data[:,2]*=1e2

        data=self._data.reshape(self.Np*6,1)

        for x in data:
            out+=pack('d',x)

        out+=pack('d',self.mass)
        print >>fout, out
        #data.tofile(fout)
        fout.close()


class plt:
    '''
    Simple class to read in a 
    TraceWin plot file

    Class afterwards hold the following
    dictionary items:
      - Ne (number of locations)
      - Np (number of particles)
      - Ib [A] (beam current)
      - freq [MHz]
      - mc2  [MeV] 
      - Nelp [m] (locations)

    each plt[i], where i is element number, holds:
      - Zgen [cm] (location)
      - phase0 [deg] (ref phase)
      - wgen [MeV] (ref energy)
      - x [array, m]
      - xp [array, rad]
      - y [array, m]
      - yp [array, rad]
      - phi [array, rad]
      - E [array, MeV]
      - l [array] (is lost)

      Example::
        plt=ess.TraceWin.plt('calc/dtl1.plt')
        for i in [97,98]:
          data=plt[i]$
          if data:
            print data['x']
    '''
    def __init__(self, filename):
        # easy storage..
        self.filename=filename
        # used to create dict behaviour..
        self._columns=['x','xp','y','yp','phi','E', 'l']
        # read in the file..
        self._readBinaryFile()

    def _readBinaryFile(self):
        # Thanks Emma!

        import numpy

        fin=file(self.filename,'r')

        # dummy, Np, Ib, freq, dummy
        Header_type = numpy.dtype([
            ('dummy12', numpy.int16),
            ('Ne',      numpy.int32),
            ('Np',      numpy.int32),
            ('Ib',      numpy.float64),
            ('freq',    numpy.float64),
            ('mc2',     numpy.float64),
            ])
        SubHeader_type = numpy.dtype([
             ('dummy12', numpy.int8),
             ('Nelp',      numpy.int32),
             ('Zgen',     numpy.float64),
             ('phase0',     numpy.float64),
             ('wgen',     numpy.float64),
             ])


        Header=numpy.fromfile(fin, dtype=Header_type, count=1)
        self.Np=Header['Np'][0]
        self.Ne=Header['Ne'][0]
        self.Ib=Header['Ib'][0]
        self.freq=Header['freq'][0]
        self.mc2=Header['mc2'][0]

        self._data=[]
        self.Nelp=[]

        i=0
        while i<self.Ne:
            SubHeader=numpy.fromfile(fin, dtype=SubHeader_type, count=1)
            i=SubHeader['Nelp'][0]

            self.Nelp.append(i)

            Table=numpy.fromfile(fin, dtype=numpy.float32, count=self.Np*7)
            Table=Table.reshape(self.Np,7)
            data={}
            for key in ['Zgen','phase0','wgen']:
                data[key]=SubHeader[key][0]
            for j in xrange(7):
                    c=self._columns[j]
                    data[c]=Table[:,j]
                    # convert x,y from cm to m
                    if c in ['x', 'y']:
                        data[c]*=1e-2
            self._data.append(data)

    def __getitem__(self, key):
        if key in self.Nelp:
            import numpy

            i=self.Nelp.index(key)

            ret={}
            # some particles are lost, exclude those:
            lost_mask=self._data[i]['l']==0
            for key in self._data[i]:
                if isinstance(self._data[i][key], numpy.ndarray):
                    ret[key]=self._data[i][key][lost_mask]
                else:
                    ret[key]=self._data[i][key]
            return ret
        else:
            print "No data to plot at element",key

    def calc_s(self):
        '''
        Generates self.s which holds
        the position of each element
        in metres
        '''
        import numpy

        self.s=[]
        for i in self.Nelp:
            self.s.append(self[i]['Zgen']/100.0)
        self.s=numpy.array(self.s)

    def calc_avg(self):
        '''
        Calculates averages of 6D coordinates at each
        element, such that e.g.
        self.avg["x"] gives average X at each location.

        Units: m, rad, MeV
        '''
        import numpy

        self.avg=dict(x=[], xp=[], y=[], yp=[], E=[], phi=[])

        vals=self._columns[:-1]

        for i in self.Nelp:
            data=self[i]
            for v in vals:
                self.avg[v].append(numpy.average(data[v]))

    def calc_rel(self):
        '''
        Calculates relativistic gamma/beta
        at each position, based on 
        AVERAGE beam energy
        (NOT necessarily reference)
        '''
        import numpy

        if not hasattr(self,'avg'):
            self.calc_avg()
        self.gamma=[]
        self.beta=[]
        for i,j in zip(self.Nelp,xrange(len(self.Nelp))):
            Eavg=self.avg['E'][j]
            self.gamma.append((self.mc2+Eavg)/self.mc2)
            self.beta.append(numpy.sqrt(1.-1./self.gamma[-1]**2))
        self.gamma=numpy.array(self.gamma)
        self.beta=numpy.array(self.beta)

    def calc_minmax(self,pmin=5,pmax=95):
        '''
        Calculates min/max values of beam coordinates
        in percentile, pmin is lower and pmax upper.

        Units: cm
        '''
        import numpy

        self.min=dict(x=[], xp=[], y=[], yp=[], E=[])
        self.max=dict(x=[], xp=[], y=[], yp=[], E=[])

        for i in self.Nelp:
            data=self[i]
            for v in self.min.keys():
                self.min[v].append(numpy.percentile(data[v],pmin))
                self.max[v].append(numpy.percentile(data[v],pmax))

        for v in self.min.keys():
            self.min[v]=numpy.array(self.min[v])
            self.max[v]=numpy.array(self.max[v])

    def calc_sigma(self):
        '''
        Calculates the sigma matrix

        Creates self.sigma such that self.sigma[i,j]
        returns the sigma matrix for value i,j.

        The numbering is:
        0: x
        1: xp
        2: y
        3: yp
        4: E
        5: phi
        '''

        import numpy

        if not hasattr(self,'avg'):
               self.calc_avg()

        vals=self._columns[:-1]

        self.sigma=[]
        for j in xrange(len(self.Nelp)):
            i=self.Nelp[j]
            data=self[i]

            self.sigma.append([[numpy.mean( (data[n]-self.avg[n][j]) * (data[m] - self.avg[m][j]) ) for n in vals] for m in vals])

        self.sigma=numpy.array(self.sigma)

    def calc_std(self):
        '''
        Calculates the beam sizes

        '''

        import numpy

        if not hasattr(self,'sigma'):
               self.calc_sigma()

        vals=self._columns[:-1]

        self.std={}

        for j in xrange(len(vals)):
            v=vals[j]
            self.std[v]=numpy.sqrt(self.sigma[:,j,j])

    def calc_twiss(self):
        '''
        Calculates emittance, beta, alfa, gamma
        for each plane, x-xp, y-yp, and E-phi
        '''

        import numpy

        if not hasattr(self,'sigma'):
               self.calc_sigma()
        if not hasattr(self,'gamma'):
               self.calc_rel()

        self.twiss_eps=[]
        for j in xrange(len(self.Nelp)):
            self.twiss_eps.append([numpy.sqrt(numpy.linalg.det(self.sigma[j][i:i+2][:,i:i+2])) for i in (0,2,4)])
        self.twiss_eps=numpy.array(self.twiss_eps)

        # Calculate normalized emittance:
        # TODO: this is NOT correct normalization for longitudinal
        self.twiss_eps_normed=self.twiss_eps.copy()
        for i in xrange(3):
            self.twiss_eps_normed[:,i]*=self.gamma*self.beta

        # Calculate beta:
        # This is a factor 10 different from what TraceWin plots
        self.twiss_beta = [[self.sigma[j][i][i]/self.twiss_eps[j,i/2] for i in (0,2,4)] for j in xrange(len(self.Nelp))]
        self.twiss_beta = numpy.array(self.twiss_beta)

        # Calculate alpha:
        self.twiss_alpha = [[-self.sigma[j][i][i+1]/self.twiss_eps[j,i/2] for i in (0,2,4)] for j in xrange(len(self.Nelp))]
        self.twiss_alpha = numpy.array(self.twiss_alpha)




class density_file:
    '''
    Simple class to read a TraceWin density file
    into a pythonized object
    '''

    def __init__(self, filename):
        import numpy, sys


        self.filename=filename
        self.fin=file(self.filename, 'r')

        # currently unknown:
        self.version=0

        # first we simply count how many elements we have:
        counter=0
        while True:
            try:
                self._skipAndCount()
                counter+=1
            except IndexError: # EOF reached..
                break
        if sys.flags.debug:
            print "Number of steps found:", counter
        self.fin.seek(0)

        # set up the arrays..
        self.i=0
        # z position [m] :
        self.z=numpy.zeros(counter)
        # element index number
        self.nelp=numpy.zeros(counter)
        # current [mA] :
        self.ib=numpy.zeros(counter)
        # number of lost particles:
        self.Np=numpy.zeros(counter)

        self.Xouv=numpy.zeros(counter)
        self.Youv=numpy.zeros(counter)


        self.moy=numpy.zeros((counter,7))
        self.moy2=numpy.zeros((counter,7))

        self._max=numpy.zeros((counter,7))
        self._min=numpy.zeros((counter,7))

        if self.version>=5:
            self.rms_size=numpy.zeros((counter,7))
            self.rms_size2=numpy.zeros((counter,7))

        if self.version>=6:
            self.min_pos_moy=numpy.zeros((counter,7))
            self.max_pos_moy=numpy.zeros((counter,7))

        if self.version>=7:
            self.rms_emit=numpy.zeros((counter,3))
            self.rms_emit2=numpy.zeros((counter,3))

        if self.version>=8:
            self.energy_accept=numpy.zeros(counter)
            self.phase_ouv_pos=numpy.zeros(counter)
            self.phase_ouv_neg=numpy.zeros(counter)

        self.lost=numpy.zeros((counter,self.Nrun))
        self.powlost=numpy.zeros((counter,self.Nrun))

        self.lost2=numpy.zeros(counter)
        self.Minlost=numpy.zeros(counter)
        self.Maxlost=numpy.zeros(counter)

        self.powlost2=numpy.zeros(counter)
        self.Minpowlost=numpy.zeros(counter)
        self.Maxpowlost=numpy.zeros(counter)


        while self.i<counter:
            self._getFullContent()
            self.i+=1


    def _getHeader(self):
        import numpy


        # header..
        version=numpy.fromfile(self.fin, dtype=numpy.int16, count=1)[0]
        year=numpy.fromfile(self.fin, dtype=numpy.int16, count=1)[0]

        # there is much more data written, but it is undocumented. Our trick to get back "on line":
        shift=0
        while year!=2011 or version!=8:
            shift+=1
            version=year
            year=numpy.fromfile(self.fin, dtype=numpy.int16, count=1)[0]

        self.vlong=numpy.fromfile(self.fin, dtype=numpy.int16, count=1)[0]
        self.Nrun=numpy.fromfile(self.fin, dtype=numpy.int32, count=1)[0]

        if shift:
            raise ValueError("ERROR, shifted "+str(shift*2)+" bytes")

        self.version=version
        self.year=year

    def _skipAndCount(self):

        import numpy

        self._getHeader()
        
        if self.version==8:
            numpy.fromfile(self.fin, dtype=numpy.int16, count=8344/2)
        if self.Nrun>1:
            #WARN not 100% sure if this is correct..
            numpy.fromfile(self.fin, dtype=numpy.int16, count=((5588+self.Nrun*12)/2))

    def _get_7dim_array(array):
        '''
        Unused?
        '''
        return dict(x=array[0],
                    y=array[1],
                    phase=array[2],
                    energy=array[3],
                    r=array[4],
                    z=array[5],
                    dpp=array[6],
                    )

    def _getFullContent(self):

        import numpy

        #self._getHeader()
        # no need to read the header again:
        # (though only if we are SURE about content!)
        numpy.fromfile(self.fin, dtype=numpy.int16, count=5)

        self.nelp[self.i]=numpy.fromfile(self.fin, dtype=numpy.int32, count=1)[0]
        self.ib[self.i]=numpy.fromfile(self.fin, dtype=numpy.float32, count=1)[0]
        self.z[self.i]=numpy.fromfile(self.fin, dtype=numpy.float32, count=1)[0]
        # Aperture
        self.Xouv[self.i]=numpy.fromfile(self.fin, dtype=numpy.float32, count=1)[0]
        self.Youv[self.i]=numpy.fromfile(self.fin, dtype=numpy.float32, count=1)[0]
        step=numpy.fromfile(self.fin, dtype=numpy.int32, count=1)[0]

        n=7 # x [m], y[m], Phase [deg], Energy [MeV], R[m], Z[m], dp/p

        self.moy[self.i]=numpy.fromfile(self.fin, dtype=numpy.float32, count=n)[:]
        self.moy2[self.i]=numpy.fromfile(self.fin, dtype=numpy.float32, count=n)[:]
        
        self._max[self.i]=numpy.fromfile(self.fin, dtype=numpy.float32, count=n)[:]
        self._min[self.i]=numpy.fromfile(self.fin, dtype=numpy.float32, count=n)[:]

        if self.version>=5:
            self.rms_size[self.i]=numpy.fromfile(self.fin, dtype=numpy.float32, count=n)
            self.rms_size2[self.i]=numpy.fromfile(self.fin, dtype=numpy.float32, count=n)
        if self.version>=6:
            self.min_pos_moy[self.i]=numpy.fromfile(self.fin, dtype=numpy.float32, count=n)
            self.max_pos_moy[self.i]=numpy.fromfile(self.fin, dtype=numpy.float32, count=n)
        if self.version>=7:
            self.rms_emit[self.i]=numpy.fromfile(self.fin, dtype=numpy.float32, count=3)[:]
            self.rms_emit2[self.i]=numpy.fromfile(self.fin, dtype=numpy.float32, count=3)[:]
        if self.version>=8:
            self.energy_accept[self.i]=numpy.fromfile(self.fin, dtype=numpy.float32, count=1)
            self.phase_ouv_pos[self.i]=numpy.fromfile(self.fin, dtype=numpy.float32, count=1)
            self.phase_ouv_neg[self.i]=numpy.fromfile(self.fin, dtype=numpy.float32, count=1)

        self.Np[self.i]=numpy.fromfile(self.fin, dtype=numpy.int64, count=1)[0]

        if self.Np[self.i]:
            for i in xrange(self.Nrun):
                self.lost[self.i,i]=numpy.fromfile(self.fin, dtype=numpy.int64, count=1)[0]
                self.powlost[self.i,i]=numpy.fromfile(self.fin, dtype=numpy.float32, count=1)[0]
            self.lost2[self.i]=numpy.fromfile(self.fin, dtype=numpy.int64, count=1)[0]
            self.Minlost[self.i]=numpy.fromfile(self.fin, dtype=numpy.int64, count=1)[0]
            self.Maxlost[self.i]=numpy.fromfile(self.fin, dtype=numpy.int64, count=1)[0]
            self.powlost2[self.i]=numpy.fromfile(self.fin, dtype=numpy.float64, count=1)[0]
            self.Minpowlost[self.i]=numpy.fromfile(self.fin, dtype=numpy.float32, count=1)[0]
            self.Maxpowlost[self.i]=numpy.fromfile(self.fin, dtype=numpy.float32, count=1)[0]

            if self.vlong==1:
                tab=numpy.fromfile(self.fin, dtype=numpy.uint64, count=n*step)
            else:
                tab=numpy.fromfile(self.fin, dtype=numpy.uint32, count=n*step)
    
            if self.ib[self.i]>0:
                tabp=numpy.fromfile(self.fin, dtype=numpy.uint32, count=3*step)

    def _avg_merge(self,other,param):
        '''
        returns the average of the parameter
        weighted by how many Nruns in self and other object

        This allows for different lengths of the two arrays..
        '''
        mine=getattr(self,param)
        new=getattr(other,param)
        if len(mine)>len(new):
            ret = mine.copy()
            ret[:len(new)] = (mine[:len(new)]*self.Nrun+new*other.Nrun)/(self.Nrun+other.Nrun)
        elif len(mine)<len(new):
            ret = new.copy()
            ret[:len(mine)] = (mine*self.Nrun+new[:len(mine)]*other.Nrun)/(self.Nrun+other.Nrun)
        else:
            ret = (mine*self.Nrun+new*other.Nrun)/(self.Nrun+other.Nrun)
        return ret

    def _sum_merge(self,other,param):
        '''
        returns the sum of the parameter

        This allows for different lengths of the two arrays..
        '''
        mine=getattr(self,param)
        new=getattr(other,param)
        if len(mine)>len(new):
            ret = mine.copy()
            ret[:len(new)] += new
        elif len(mine)<len(new):
            ret = new.copy()
            ret[:len(mine)] += mine
        else:
            ret = mine+new
        return ret

    def _concatenate_merge(self,other,param):
        '''
        returns the concatenation of the two matrices

        This allows for different lengths of the two arrays/matrices..
        '''
        import numpy

        mine=getattr(self,param)
        new=getattr(other,param)
        ret=numpy.zeros(( max([len(mine),len(new)]), len(mine[0])+len(new[0]) ))
        ret[:len(mine),:len(mine[0])]=mine
        ret[:len(new),len(mine[0]):]=new
        return ret

    def _fun_merge(self,other,function,param):
        '''
        returns the function applied on the parameter

        This allows for different lengths of the two arrays..
        '''
        mine=getattr(self,param)
        new=getattr(other,param)
        if len(mine)>len(new):
            ret = mine.copy()
            ret[:len(new)] = function(mine[:len(new)],new)
        elif len(mine)<len(new):
            ret = new.copy()
            ret[:len(mine)] = function(mine,new[:len(mine)])
        else:
            ret = function(mine,new)
        return ret

    def merge(self,objects):
        '''
        Merge with list of objects
        '''
        import numpy

        if not isinstance(objects,list):
            raise TypeError("You tried to merge a non-list")

        # for now we only allow objects with same version..
        for o in objects:
            if self.version != o.version:
                raise ValueError("Cannot merge files with differing version")

        # merge info..
        for o in objects:
            if len(self.ib)<len(o.ib):
                raise ValueError("Sorry, not implemented yet. Complain to Yngve")

            self.ib=self._avg_merge(o,'ib')

            # this looks strange to me, but it is what TraceWin does..
            self.moy=self._sum_merge(o,'moy')
            self.moy2=self._sum_merge(o,'moy')

            self._max=self._fun_merge(o,numpy.maximum,'_max')
            self._min=self._fun_merge(o,numpy.minimum,'_min')

            if self.version>=5:
                # this looks strange to me, but it is what TraceWin does..
                self.rms_size=self._sum_merge(o,'rms_size')
                self.rms_size2=self._sum_merge(o,'rms_size2')

            if self.version>=6:
                self.max_pos_moy=self._fun_merge(o,numpy.maximum,'max_pos_moy')
                self.min_pos_moy=self._fun_merge(o,numpy.minimum,'min_pos_moy')

            if self.version>=7:
                # this looks strange to me, but it is what TraceWin does..
                self.rms_emit=self._sum_merge(o,'rms_emit')
                self.rms_emit2=self._sum_merge(o,'rms_emit2')

            if self.version>=8:
                # Warning: TraceWin does NOT merge these data in any way
                self.energy_accept=self._avg_merge(o,'energy_accept')
                self.phase_ouv_pos=self._avg_merge(o,'phase_ouv_pos')
                self.phase_ouv_neg=self._avg_merge(o,'phase_ouv_neg')

            # Note, we don't get into the problem of differing table sizes
            # particles are lost, because we have written zeroes for
            # the rest of the tables

            self.lost=self._concatenate_merge(o,'lost')
            self.powlost=self._concatenate_merge(o,'powlost')

            self.lost2=self._sum_merge(o,'lost2')
            self.powlost2=self._sum_merge(o,'powlost2')

            self.Minlost=self._fun_merge(o,numpy.minimum,'Minlost')
            self.Maxlost=self._fun_merge(o,numpy.maximum,'Maxlost')
            self.Minpowlost=self._fun_merge(o,numpy.minimum,'Minpowlost')
            self.Maxpowlost=self._fun_merge(o,numpy.maximum,'Maxpowlost')

            # Note: We are ignoring tab/tabp data...

            # merge final info (make sure to do this last!)
            self.Np=self._sum_merge(o,'Np')
            self.Nrun+=o.Nrun

    def savetohdf(self,filename='Density.h5',group='TraceWin'):
        '''
        Saves data to HDF5
        '''
        import h5py

        fout = h5py.File(filename,'a')
        if group in fout:
            del fout[group]
        group = fout.create_group(group)

        # header attributes..
        group.attrs['version']=self.version
        group.attrs['year']=self.year
        group.attrs['Nrun']=self.Nrun
        group.attrs['vlong']=self.vlong

        length=len(self.z)

        partran = sum(self.Np)>0

        # one number per location
        arrays=     ['z', 'nelp', 'ib', 'Np', 'Xouv', 'Youv']
        array_units=['m',     '', 'mA',   '',    'm',    'm']
        if self.version>=8:
            arrays +=      ['energy_accept', 'phase_ouv_pos', 'phase_ouv_neg']
            array_units += [           'eV',           'deg',           'deg']
        if partran:
            arrays +=      [ 'lost2', 'Minlost', 'Maxlost', 'powlost2', 'Minpowlost', 'Maxpowlost']
            array_units += [      '',        '',        '',      'W*w',          'W',          'W']

        # 7 numbers per location..
        coordinates=   ['moy', 'moy2', '_max', '_min']
        coordinate_units=['m',  'm*m',   'm',   'm']
        if self.version>=5 and partran:
            coordinates += ['rms_size','rms_size2']
            coordinate_units += [  'm',      'm*m']
        if self.version>=6 and partran:
            coordinates += ['min_pos_moy', 'max_pos_moy']
            coordinate_units += [     'm',           'm']


        for val,unit in zip(arrays,array_units):
            data_set=group.create_dataset(val, (length,), dtype='f')
            data_set[...]=getattr(self,val)
            if unit:
                data_set.attrs['unit']=unit

        for val,unit in zip(coordinates,coordinate_units):
            data_set=group.create_dataset(val, (length,7), dtype='f')
            data_set[...]=getattr(self,val)
            if unit:
                data_set.attrs['unit']=unit

        if self.version>=7 and partran:
            # 3 numbers per location..
            emit_data=['rms_emit', 'rms_emit2']
            emit_units=['m*rad', 'm*m*rad*rad']
            for val,unit in zip(emit_data,emit_units):
                data_set=group.create_dataset(val, (length,3), dtype='f')
                data_set[...]=getattr(self,val)
                if unit:
                    data_set.attrs['unit']=unit
        if partran:
            # 1 numbers per location and per run..
            data=['lost', 'powlost']
            units=['', 'W']
            for val,unit in zip(data, units):
                data_set=group.create_dataset(val, (length,self.Nrun), dtype='f')
                data_set[...]=getattr(self,val)
                if unit:
                    data_set.attrs['unit']=unit

        fout.close()


class remote_data_merger:
    def __init__(self, base='.'):
        self._base=base
        self._files=[]

    def add_file(self,filepath):
        import os

        if os.path.exists(filepath):
            fname=filepath
        else:
            fullpath=os.path.join(self._base,filepath)
            if os.path.exists(fullpath):
                fname=fullpath
            else:
                raise ValueError("Could not find file "+filepath)
        if fname not in self._files:
            self._files.append(fname)

    def generate_partran_out(self,filename=None):
        '''
        Creates a string to be written to file
        each line is a list.

        If filename is given, writes directly to output file.

        '''

        import numpy as np

        h1=[]
        h2=[]

        d1=[]
        d2=[]
        d3=[]

        if self._files:
            for f in self._files:
                string=file(f,'r').read()
                split=string.split('$$$')
                if split[9]!='Data_Error':
                    raise ValueError("Magic problem, please complain to Yngve")

                thisdata=split[10].strip().split('\n')

                if not h1:
                    h1=[thisdata[0]+" (std in paranthesis)"]
                    h2=thisdata[2:10]
                d1.append(thisdata[1].split())
                d2.append(thisdata[10])
                d3.append(thisdata[11])

            # fix d1:
            for i in xrange(len(d1)):
                for j in xrange(len(d1[0])):
                    d1[i][j]=float(d1[i][j])
            d1=np.array(d1)
            means=d1.mean(axis=0)
            stds=d1.std(axis=0)
            d1=[]
            for i in xrange(len(stds)):
                if stds[i]/means[i]<1e-10:
                    stds[i]=0.0
            for i in xrange(len(stds)):
                # some small std are removed..
                if stds[i]/means[i]>1e-8:
                    d1.append('%f(%f)' % (means[i],stds[i]))
                else: #error is 0
                    d1.append(str(means[i]))
            d1=[' '.join(d1)]

            # create data:
            data=h1+d1+h2+d2+d3

            if filename:
                file(filename,'w').write('\n'.join(data))

            return data

class partran(dict):
    '''
    Read partran1.out files..
    '''
    def __init__(self,filename):
        self.filename=filename
        self._readAsciiFile()

    def _readAsciiFile(self):

        import numpy

        stream=file(self.filename,'r')
        for i in xrange(10):
            line=stream.readline()
            if line.strip()[0]=='#':
                break
        self.columns=['NUM']+line.split()[1:]
        self.data=numpy.loadtxt(stream)

        self._dict={}
        for i in xrange(len(self.columns)):
            self[self.columns[i]]=self.data[:,i]