dispersiveXanes/xppl37_spectra.py

import numpy as np
import matplotlib.pyplot as plt
import copy
import argparse
import collections
import dispersiveXanes_alignment as alignment
import dispersiveXanes_utils as utils
import xanes_analyzeRun

parser = argparse.ArgumentParser(description='Process argv')

parser.add_argument('--run', type=int,default=82,help='which run to analyze')
parser.add_argument('--force', action="store_true",help='force calculation')
args = parser.parse_args()


profile_ret = collections.namedtuple("profile_ret",["run","p1","p2","calibs"])


nice_colors = ["#1b9e77", "#d95f02", "#7570b3"]
gradual_colors = ['#014636', '#016c59', '#02818a', '#3690c0', '#67a9cf', '#a6bddb', '#d0d1e6']#, '#ece2f0']


def calcSpectraForRun(run=82,calibs="all",realign=False,init="auto",alignCalib=0,force=False):
  """ Calculate spectra (spec1, spec2) for run based on alignment.

      Alignment can be 'forced' with realign=True.
      In such a case 
        init = run for initial alignment
        alignCalib = calibcycle for alignment
  """
  if init=="auto": init=run
  if isinstance(run,int):
    r = xanes_analyzeRun.AnalyzeRun(run,initAlign=init)
  else:
    r = run
  if realign:
    r.doShots(slice(20),calib=refCalib,doFit=True)
    r.saveTransform()
  # next line is used to force calculations in case of realignment
  fname = 'auto' if not realign else "thisfiledoesnotexists"
  if force: fname = "thisfiledoesnotexists"
  if len(r.results) == 0:
    try:
      r.load(fname)
      print("Loading previously saved results")
    except FileNotFoundError:
      r.analyzeScan(calibs=calibs,nImagesToFit=0,nSaveImg=4)
      r.save(overwrite=True)
  # cannot take the output from r.results because it might have been calculated for
  # a bigger range than asked for.
  if isinstance(calibs,int):
    calibsForOut = (calibs,)
  elif isinstance(calibs,slice):
    calibsForOut = list(range(r.nCalib))[calibs]
  elif calibs == "all":
    calibsForOut = r.results.keys()
  else:
    calibsForOut = calibs
  # focused data have one single calibcycle ...
  if len(r.results) > 1:
    p1 = [r.results[calib].p1 for calib in calibsForOut]
    p2 = [r.results[calib].p2 for calib in calibsForOut]
  else:
    idx = r.results[None].fom < 0.5
    p1 = [ r.results[None].p1[idx], r.results[None].p1[~idx] ]
    p2 = [ r.results[None].p2[idx], r.results[None].p2[~idx] ]
  return profile_ret( run =r, p1 =p1, p2=p2,calibs=calibsForOut)


def calcSpectraForRefAndSample(run=82,refCalibs=slice(None,None,2),forceSpectraCalculation=False):
  """ Function to calculate the Spectra with and without (ref) the sample.
      It can analyze two kinds of runs:
      * Single run with that alternates IN and OUT (like run 82)
        in this case use something like:
        calcSpectraForRefAndSample(82,refCalibs=slice(None,None,2)
        or
        calcSpectraForRefAndSample(82,refCalibs=(0,2,4,6))
      * Multiple runs (one with reference and one with sample) like run 155 and 156
        in this case use something like:
        calcSpectraForRefAndSample( run=(155,156) )
        where the first is the reference run and the second the one with the sample.
        in this second case the refCalibs does not play a role
      use forceSpectraCalculation = True, to re-read the images
  """
  if isinstance(run,int):
    run   = xanes_analyzeRun.AnalyzeRun(run)
    if isinstance(refCalibs,slice): refCalibs  = list(range(run.nCalib))[refCalibs]
    if isinstance(refCalibs,int):   refCalibs  = [refCalibs,]
    sampleCalibs = [c+1 for c in refCalibs]
    # need a single call (for sample and ref) to save all calibcycles
    data= calcSpectraForRun(run,calibs=refCalibs+sampleCalibs,\
                      force=forceSpectraCalculation);
    # for focused runs
    if len(data.run.results) == 1:
      ref    = profile_ret(run=data.run,p1=[data.p1[0],],p2=[data.p2[0],],calibs=[0,])
      sample = profile_ret(run=data.run,p1=[data.p1[1],],p2=[data.p2[1],],calibs=[0,])
    else:
      ref = calcSpectraForRun(run,calibs=refCalibs)
      sample  = calcSpectraForRun(run,calibs=sampleCalibs)
  elif isinstance(run,(list,tuple)):
    refRun     = xanes_analyzeRun.AnalyzeRun(run[0])
    sampleRun  = xanes_analyzeRun.AnalyzeRun(run[1],initAlign=run[0])
    refCalibs     = [0,]
    sampleCalibs  = [0,]
    ref = calcSpectraForRun(refRun,calibs=refCalibs,\
          force=forceSpectraCalculation)
    sample  = calcSpectraForRun(sampleRun,calibs=sampleCalibs,\
          force=forceSpectraCalculation)
  return ref,sample

def calcRef(r1,r2,calibs=None,threshold=0.05):
  """ r1 and r2 are list of 2d arrays (nShots,nPixels) for each calibcycle """
  if calibs is None: calibs = list(range(len(r1)))
  out = collections.OrderedDict()
  out["ratioOfAverage"] = dict()
  out["medianOfRatios"] = dict()
  for p1,p2,n in zip(r1,r2,calibs):
    out["ratioOfAverage"][n] = utils.ratioOfAverage(p1,p2,threshold=threshold)
    out["medianOfRatios"][n] = utils.medianRatio(p1,p2,threshold=threshold)
  # add curves with all calib together
  p1 = np.vstack(r1)
  p2 = np.vstack(r2)
  n = ",".join(map(str,calibs))
  ref1 = utils.ratioOfAverage(p1,p2,threshold=threshold)
  ref2 = utils.medianRatio(p1,p2,threshold=threshold)
  out["ratioOfAverage"][n] = utils.ratioOfAverage(p1,p2,threshold=threshold)
  out["medianOfRatios"][n] = utils.medianRatio(p1,p2,threshold=threshold)
  out["ratioOfAverage"]['all'] = out["ratioOfAverage"][n]
  out["medianOfRatios"]['all'] = out["medianOfRatios"][n]
  return out

def showDifferentRefs(run=82,refCalibs=slice(None,None,2),threshold=0.05):
  """ example plots showing how stable are the different ways of taking spectra """
  prof = calcSpectraForRun(run,calibs=refCalibs)
  refs = calcRef(prof.p1,prof.p2,calibs=prof.calibs)
  kind_of_av = list(refs.keys())
  fig,ax=plt.subplots(len(kind_of_av)+1,1,sharex=True,sharey=True)
  E = prof.run.E
  calibs = list(refs[kind_of_av[0]].keys())
  for ikind,kind in enumerate(kind_of_av):
    for calib in calibs:
      if isinstance(calib,int):
        ax[ikind].plot(E,refs[kind][calib],label="calib %s"%calib)
      else:
        if calibs == 'all': continue
        ax[ikind].plot(E,refs[kind][calib],label="calib %s"%calib,lw=2,color='k',alpha=0.7)
        ax[-1].plot(E,refs[kind][calib],label="calib all, %s"%kind,lw=1.5,color=nice_colors[ikind],alpha=0.8)
  for ikind,kind in enumerate(kind_of_av): ax[ikind].set_title("Run %d, %s"%(run,kind)) 
  ax[0].set_ylim(0.88,1.12)
  ax[0].set_ylim(0.88,1.12)
  ax[-2].legend()
  ax[-1].legend()
  for a in ax: a.grid()

def calcSampleAbs(run=82,refCalibs=slice(None,None,2),threshold=0.05,refKind="medianOfRatios",forceSpectraCalculation=False):
  """ example of use
      ratio = calcSampleAbs(82)
      ratio = calcSampleAbs( (155,156) )
  """
  ref,sample = calcSpectraForRefAndSample(run,refCalibs=refCalibs,forceSpectraCalculation=forceSpectraCalculation)
  temp  = calcRef(ref.p1,ref.p2,calibs=ref.calibs,threshold=threshold)
  ref   = temp[refKind]['all']
  p1 = np.vstack( sample.p1 )
  p2 = np.vstack( sample.p2 )
  print(p1.shape)
  p1,p2 = utils.maskLowIntensity(p1,p2,threshold=threshold)
  ratio = p2/p1
  ratio = ratio/ref
  return p1,p2,-np.log10(ratio)

def showSpectra(run=82,shots=slice(5),calibs=0,averageEachCalib=False,
                normalization="auto",shifty=1,xlim=(7060,7180),showAv=True):
  """ averageEachCalib: if True, plot only one (averaged) spectrum per calibcycle
      normalization: if "auto", the max of the spectra that will be plotted will be used
  """
  r = xanes_analyzeRun.AnalyzeRun(run=run)
  r.load()
  calibsSaved = list(r.results.keys()); calibsSaved.sort()
  res = [ r.results[c].p2 for c in calibsSaved ]
  if isinstance(calibs,slice): res=res[calibs]
  if isinstance(calibs,int):   res=[res[calibs],]
  avCalibs = [ np.nanmedian(spectra,axis=0) for spectra in res ]
  if averageEachCalib:
    res = avCalibs
    showAv = False; # it does not make sense to plot it twice !
  fig,ax = plt.subplots(len(res),1,sharex=True,sharey=True,squeeze=False)
  if normalization == "auto": normalization = np.nanmax( [temp[shots] for temp  in res] )
  for (av,spectra,a) in zip(avCalibs,res,ax[:,0]):
    spectra_norm = spectra[shots]/normalization
    av_norm      = av/normalization
    for i,spectrum in enumerate(spectra_norm):
      color = gradual_colors[i%len(gradual_colors)]
      a.axhline(i*shifty,ls='--',color=color)
      if showAv: a.fill_between(r.E,i*shifty,av_norm+i*shifty,color='#d95f0e',alpha=0.4)
      print(i)
      a.plot(r.E,spectrum+i*shifty,color=color,lw=2)
  ax[0][0].set_xlim(*xlim)
  ax[0][0].set_title("Run %d"%run)
  if not averageEachCalib: ax[0][0].set_ylim(0,shifty*(len(spectra_norm)))


color_ss = '#08519c'
color_av = '#238b45'
color_av_all = '#d95f0e'

def showAbs(run=82,shots=slice(5),normalization="auto",shifty=1,xlim=(7080,7180),showAv=True,showAvOverAll=True,smoothWidth=0,threshold=0.01,filterShot=0.1):
  """ normalization: if "auto", the max of the spectra that will be plotted will be used
      filterShot = means that it filters out the filterShot*100 percentile
  """
  E = alignment.defaultE
  p1,p2,abs = calcSampleAbs(run=run,threshold=threshold)
  p1_sum = p1.sum(-1)
  if filterShot>0:
    idx = p1_sum>np.percentile(p1_sum,filterShot*100)
    p1 = p1[idx]
    p2 = p2[idx]
    abs = abs[idx]
  p1_av  = np.nanmean(p1,axis=0)
  p2_av  = np.nanmean(p2,axis=0)
  # somehow nanmedian screws up when array is too big ... so using nanmean
  abs_av = np.nanmean(abs,axis=0)
  p1 = p1[shots]; p2=p2[shots]; abs = abs[shots]
  if smoothWidth > 0: abs = smoothSpectra(E,abs,res=smoothWidth)
  fig,ax = plt.subplots(1,2,sharex=True,sharey=True,squeeze=False)
  ax = ax[0]
  if normalization == "auto": normalization = np.nanmax( p1 )
  for ishot,(s1,s2,a) in enumerate(zip(p1,p2,abs)):
    s1_norm = s1/normalization
    s2_norm = s2/normalization
    color = gradual_colors[ishot%len(gradual_colors)]
    ax[0].axhline(ishot*shifty,ls='--',color=color)
    ax[1].axhline(ishot*shifty,ls='--',color=color)
    if showAvOverAll:
      if ishot == 0: ax[0].fill_between(E,ishot*shifty,p1_av/normalization+ishot*shifty,color=color_av_all,alpha=0.6)
      ax[1].plot(E,abs_av+ishot*shifty,color=color_av_all,lw=2,zorder=20)
    if showAv:
      ax[1].plot(E,np.nanmedian(abs,0)+ishot*shifty,color=color_av,lw=2,zorder=10)
    ax[0].plot(E,s1_norm+ishot*shifty,ls = '-' ,color='0.8',lw=2)
    ax[0].plot(E,s2_norm+ishot*shifty,ls = '-' ,color='0.3',lw=2)
    ax[1].plot(E,a+ishot*shifty,color=color_ss,lw=2)
  ax[0].set_xlim(*xlim)
  ax[0].set_title("Run %s"%str(run))
  ax[1].set_ylabel("Sample Absorption")
  ax[0].set_ylabel("Normalized Spectra")
  ax[0].set_ylim(0,shifty*(p1.shape[0]))
  print("STD of (average over shown shots) - (average over all): %.3f"%np.nanstd(np.nanmedian(abs,0)-abs_av))
  ax[1].set_title("STD of (average_{shown}) - (average_{all}): %.3f"%np.nanstd(np.nanmedian(abs,0)-abs_av))


def showAbsWithSweep(run=(155,156),first=0,period=150,nSpectra=10,**kwards):
  shots = slice(first,first+period,int(period/nSpectra))
  showAbs(run=run,shots=shots,**kwards)
 
def smoothSpectra(E,abs_spectra,res=0.5):
  from scipy import integrate
  if abs_spectra.ndim == 1: abs_spectra=abs_spectra[np.newaxis,:]
  out = np.empty_like(abs_spectra)
  for ispectrum in range(abs_spectra.shape[0]):
    idx = np.isfinite(abs_spectra[ispectrum])
    Eclean = E[idx]
    for i in range(len(E)):
      g = 1/np.sqrt(2*np.pi)/res*np.exp(-(E-E[i])**2/2/res**2)
      tointegrate = g*abs_spectra[ispectrum]
      # filter out nans
      tointegrate = tointegrate[idx]
      out[ispectrum,i] = integrate.simps(tointegrate,x=Eclean) 
  return out

def doLongCalc():
  #calcSpectraForRefAndSample(82,forceSpectraCalculation=True)

  # scanning requires a lower level call
  r = xanes_analyzeRun.AnalyzeRun(84)
  r.analyzeScan(nShotsPerCalib="all",calibs="all",nSaveImg=2,calibsToFit='even',nImagesToFit=3)
  r.save(overwrite=True)
  #calcSpectraForRefAndSample(84,forceSpectraCalculation=False)

  calcSpectraForRefAndSample(96,forceSpectraCalculation=True)
  calcSpectraForRefAndSample((155,156),forceSpectraCalculation=True)
  #r = calcSpectra(run,refCalibs=refCalib,force=force)


if __name__ == "__main__":
  pass
  #main(args.run,force=args.force)