############################################### ## PLANETOPLOT ## ## --> PPCLASS ## ## A generic and versatile Python module ## ## ... to read netCDF files and plot ## ############################################### ## Author: Aymeric Spiga. 02-03/2013 ## ############################################### # python built-in librairies import os import time as timelib import pickle # added librairies import numpy as np import netCDF4 import matplotlib.pyplot as mpl # personal librairies import ppplot import ppcompute ############################################### ################################### #### HEADER ## #### ... executed when imported ## ################################### # where settings files are located... whereset = None whereset = ppcompute.findset(whereset) # ... we load user-defined automatic settings from set_ppclass.txt zefile = "set_ppclass.txt" glob_listx = [] ; glob_listy = [] ; glob_listz = [] ; glob_listt = [] glob_listarea = [] try: f = open(whereset+zefile, 'r') ; lines = f.readlines() for stuff in lines[5].strip().split(';'): glob_listx.append(stuff) for stuff in lines[8].strip().split(';'): glob_listy.append(stuff) for stuff in lines[11].strip().split(';'): glob_listz.append(stuff) for stuff in lines[14].strip().split(';'): glob_listt.append(stuff) for stuff in lines[17].strip().split(';'): glob_listarea.append(stuff) except IOError: print "warning: "+zefile+" not in "+whereset+" ; no presets." ################################## #### USEFUL GENERIC FUNCTIONS #### ################################## # inspect variables and dimensions in a netCDF file def inspect(filename): print "**** INSPECT FILE",filename test = netCDF4.Dataset(filename) print "**** VARIABLES: ",test.variables.keys() for dim in test.dimensions.keys(): output = "**** DIMENSION: "+str(dim)+" "+str(len(test.dimensions[dim])) try: output = output + " ----> "+str(test.variables[dim][0])+" "+str(test.variables[dim][-1]) except: pass print output ; output = "" # check a tab and exit if wrong. if just one string make it a list. # (if allownumber, convert this into a string). def checktab(tab,mess="",allownone=False,allownumber=False): if tab is None: if not allownone: print "pp.define: no "+mess ; exit() else: pass else: if not isinstance(tab, list): if isinstance(tab, str): tab = [tab] elif (isinstance(tab, int) or isinstance(tab, float)) and allownumber: tab = [str(tab)] else: print "pp.define: "+mess+" should be either a string or a list of strings!" ; exit() elif isinstance(tab, list): if isinstance(tab[0],str): pass elif (isinstance(tab[0], int) or isinstance(tab[0], float)) and allownumber: for iii in range(len(tab)): tab[iii] = str(tab[iii]) else: print "pp.define: "+mess+" should be either a string or a list of strings!" ; exit() return tab # determine which method is to be applied to a given dimension def findmethod(tab): if tab is None: output = "free" elif tab[0,0] != tab[0,1]: output = "comp" else: output = "fixed" return output # read what is given by the user (version of T. Navarro) def readslices(saxis): if saxis == None: zesaxis = None else: zesaxis = np.empty((len(saxis),2)) for i in range(len(saxis)): a = separatenames(saxis[i]) if len(a) == 1: zesaxis[i,:] = float(a[0]) else: zesaxis[i,0] = float(a[0]) zesaxis[i,1] = float(a[1]) return zesaxis # look for comas in the input name to separate different names (files, variables,etc ..) # (needed by readslices) def separatenames (name): if name is None: names = None else: names = [] ; stop = 0 ; currentname = name while stop == 0: indexvir = currentname.find(',') if indexvir == -1: stop = 1 ; name1 = currentname else: name1 = currentname[0:indexvir] names = np.concatenate((names,[name1])) currentname = currentname[indexvir+1:len(currentname)] return names ####################### ### THE MAIN OBJECT ### ####################### class pp(): # print out a help string when help is invoked on the object def __repr__(self): whatprint = 'pp object. \"help(pp)\" for more information\n' return whatprint # default settings # -- user can define settings by two methods. # -- 1. yeah = pp(file="file.nc") # -- 2. yeah = pp() ; yeah.file = "file.nc" def __init__(self,file=None,var="notset",\ filegoal=None,vargoal=None,\ x=None,y=None,z=None,t=None,\ stridex=1,stridey=1,\ stridez=1,stridet=1,\ compute="mean",\ verbose=False,noproj=False,\ superpose=False,\ plotin=None,\ forcedimplot=-1,\ out="gui",\ filename="myplot",\ folder="./",\ includedate=True,\ xlabel=None,ylabel=None,\ xcoeff=None,ycoeff=None,\ proj=None,\ vmin=None,vmax=None,\ div=None,\ colorb=None,\ lstyle=None,\ marker=None,\ color=None,\ label=None,\ title=None): self.request = None self.nfin = 0 ; self.nvin = 0 self.nplotx = None ; self.nploty = None self.nplotz = None ; self.nplott = None self.status = "init" self.fig = None ; self.subv = None ; self.subh = None self.n = 0 ; self.howmanyplots = 0 self.nplot = 0 self.p = None self.customplot = False self.allfield = None ## what could be defined by the user self.file = file self.var = var self.filegoal = filegoal self.vargoal = vargoal self.x = x ; self.y = y ## if None, free dimension self.z = z ; self.t = t ## if None, free dimension self.stridex = stridex ; self.stridey = stridey self.stridez = stridez ; self.stridet = stridet self.compute = compute self.verbose = verbose self.noproj = noproj self.plotin = plotin self.superpose = superpose self.forcedimplot = forcedimplot self.out = out self.filename = filename self.folder = folder self.includedate = includedate ## here are user-defined plot settings ## -- if not None, valid on all plots in the pp() objects self.xlabel = xlabel ; self.xcoeff = xcoeff self.ylabel = ylabel ; self.ycoeff = ycoeff self.proj = proj self.vmin = vmin ; self.vmax = vmax self.div = div self.colorb = colorb self.lstyle = lstyle self.marker = marker self.color = color self.label = label self.title = title # print status def printstatus(self): if self.filename == "THIS_IS_A_CLONE": pass else: print "**** PPCLASS. Done step: " + self.status ##################################################### # EMULATE OPERATORS + - * / ** << FOR PP() OBJECTS # ##################################################### # define the operation << # ... e.g. obj2 << obj1 # ... means: get init for pp object obj2 from another pp object obj1 # ... (this should solve the affectation trap obj2 = obj1) def __lshift__(self,other): if other.__class__.__name__ == "pp": self.file = other.file self.var = other.var self.filegoal = other.filegoal self.vargoal = other.vargoal self.x = other.x ; self.y = other.y ## if None, free dimension self.z = other.z ; self.t = other.t ## if None, free dimension self.stridex = other.stridex ; self.stridey = other.stridey self.stridez = other.stridez ; self.stridet = other.stridet self.verbose = other.verbose self.noproj = other.noproj self.plotin = other.plotin self.superpose = other.superpose self.forcedimplot = other.forcedimplot self.out = other.out self.filename = other.filename self.folder = other.folder self.xlabel = other.xlabel ; self.xcoeff = other.xcoeff self.ylabel = other.ylabel ; self.ycoeff = other.ycoeff self.proj = other.proj self.vmin = other.vmin ; self.vmax = other.vmax self.div = other.div self.colorb = other.colorb self.lstyle = other.lstyle self.marker = other.marker self.color = other.color self.label = other.label self.title = other.title self.includedate = other.includedate else: print "!! ERROR !! argument must be a pp object." ; exit() # check the compatibility of two objects for operations # --> if other is a pp class, test sizes and return isnum = False # --> if other is an int or a float, return isnum = True # --> otherwise, just print an error and exit def checktwo(self,other): if other.__class__.__name__ == "pp": isnum = False if self.status in ["init","defined"] or other.status in ["init","define"]: print "!! ERROR !! Please use .retrieve to get fields for plots with one of your pp operands." ; exit() if self.nfin != other.nfin or \ self.nvin != other.nvin or \ self.nplott != other.nplott or \ self.nplotz != other.nploty or \ self.nploty != other.nploty or \ self.nplotx != other.nplotx : print "!! ERROR !! The two operands do not have the same number of files, variables, t z y x requests." exit() elif isinstance(other,int) or isinstance(other,float): isnum = True else: print "!! ERROR !! The operand is neither a pp class nor an integer or a float." ; exit() return isnum # define a selective copy of a pp() object for operations # ... copy.copy() is not conservative (still acts like a pointer) # ... copy.deepcopy() does not work with netCDF objects # so what is done here is a copy of everything except # (to avoid sharing with self and therefore modifying self through operations) # - request attribute of pp() object # - field attribute of the onerequest() objects def selective_copy(self): if self.status in ["init","defined"]: print "!! ERROR !! Please use .retrieve to get fields for the object you want to copy from." ; exit() the_clone = pp() for k, v in vars(self).items(): if k != "request": setattr(the_clone,k,v) the_clone.verbose = False the_clone.filename = "THIS_IS_A_CLONE" # trick to avoid additional outputs the_clone.define() for i in range(self.nfin): for j in range(self.nvin): for t in range(self.nplott): for z in range(self.nplotz): for y in range(self.nploty): for x in range(self.nplotx): obj_ref = self.request[i][j][t][z][y][x] obj = the_clone.request[i][j][t][z][y][x] for k, v in vars(obj_ref).items(): if k != "field": setattr(obj,k,v) the_clone.status = "retrieved" the_clone.filename = self.filename return the_clone # define the operation + on two objects. or with an int/float. # ... with selective_copy the self object is not modified. def __add__(self,other): isnum = self.checktwo(other) the_clone = self.selective_copy() for i in range(self.nfin): for j in range(self.nvin): for t in range(self.nplott): for z in range(self.nplotz): for y in range(self.nploty): for x in range(self.nplotx): obj = the_clone.request[i][j][t][z][y][x] obj_ref = self.request[i][j][t][z][y][x] if not isnum: ope = other.request[i][j][t][z][y][x].field if obj_ref.field.shape != ope.shape: print "!! ERROR !! The two fields for operation do not have the same shape.",self.field.shape,other.field.shape exit() else: ope = other goal = self.vargoal[j] + self.filegoal[i] if ("vector" in goal) or ("contour" in goal): if self.verbose: print "!! WARNING !! No operation was made on contours and vectors. This can be debatted actually." obj.field = obj_ref.field else: obj.field = obj_ref.field + ope return the_clone # define the operation - on two objects. or with an int/float. # ... with selective_copy the self object is not modified. def __sub__(self,other): isnum = self.checktwo(other) the_clone = self.selective_copy() for i in range(self.nfin): for j in range(self.nvin): for t in range(self.nplott): for z in range(self.nplotz): for y in range(self.nploty): for x in range(self.nplotx): obj = the_clone.request[i][j][t][z][y][x] obj_ref = self.request[i][j][t][z][y][x] if not isnum: ope = other.request[i][j][t][z][y][x].field if obj_ref.field.shape != ope.shape: print "!! ERROR !! The two fields for operation do not have the same shape.",self.field.shape,other.field.shape exit() else: ope = other goal = self.vargoal[j] + self.filegoal[i] if ("vector" in goal) or ("contour" in goal): if self.verbose: print "!! WARNING !! No operation was made on contours and vectors. This can be debatted actually." obj.field = obj_ref.field else: obj.field = obj_ref.field - ope return the_clone # define the operation * on two objects. or with an int/float. # ... with selective_copy the self object is not modified. def __mul__(self,other): isnum = self.checktwo(other) the_clone = self.selective_copy() for i in range(self.nfin): for j in range(self.nvin): for t in range(self.nplott): for z in range(self.nplotz): for y in range(self.nploty): for x in range(self.nplotx): obj = the_clone.request[i][j][t][z][y][x] obj_ref = self.request[i][j][t][z][y][x] if not isnum: ope = other.request[i][j][t][z][y][x].field if obj_ref.field.shape != ope.shape: print "!! ERROR !! The two fields for operation do not have the same shape.",self.field.shape,other.field.shape exit() else: ope = other goal = self.vargoal[j] + self.filegoal[i] if ("vector" in goal) or ("contour" in goal): if self.verbose: print "!! WARNING !! No operation was made on contours and vectors. This can be debatted actually." obj.field = obj_ref.field else: obj.field = obj_ref.field * ope return the_clone # define the operation / on two objects. or with an int/float. # ... with selective_copy the self object is not modified. def __div__(self,other): isnum = self.checktwo(other) the_clone = self.selective_copy() for i in range(self.nfin): for j in range(self.nvin): for t in range(self.nplott): for z in range(self.nplotz): for y in range(self.nploty): for x in range(self.nplotx): obj = the_clone.request[i][j][t][z][y][x] obj_ref = self.request[i][j][t][z][y][x] if not isnum: ope = other.request[i][j][t][z][y][x].field if obj_ref.field.shape != ope.shape: print "!! ERROR !! The two fields for operation do not have the same shape.",self.field.shape,other.field.shape exit() else: ope = other goal = self.vargoal[j] + self.filegoal[i] if ("vector" in goal) or ("contour" in goal): if self.verbose: print "!! WARNING !! No operation was made on contours and vectors. This can be debatted actually." obj.field = obj_ref.field else: obj.field = obj_ref.field / ope return the_clone # define the reverse operation float/int + object def __radd__(self,other): isnum = self.checktwo(other) if not isnum: print "!! ERROR !! Operand should be a number" ; exit() return self.__add__(other) # define the reverse operation float/int - object def __rsub__(self,other): isnum = self.checktwo(other) if not isnum: print "!! ERROR !! Operand should be a number" ; exit() return self.__sub__(other) # define the reverse operation float/int * object def __rmul__(self,other): isnum = self.checktwo(other) if not isnum: print "!! ERROR !! Operand should be a number" ; exit() return self.__mul__(other) # define the reverse operation float/int / object def __rdiv__(self,other): isnum = self.checktwo(other) if not isnum: print "!! ERROR !! Operand should be a number" ; exit() return self.__div__(other) # define the operation ** on one object. # ... with selective_copy the self object is not modified. def __pow__(self,num): the_clone = self.selective_copy() if isinstance(num,int) or isinstance(num,float): for i in range(self.nfin): for j in range(self.nvin): for t in range(self.nplott): for z in range(self.nplotz): for y in range(self.nploty): for x in range(self.nplotx): obj = the_clone.request[i][j][t][z][y][x] obj_ref = self.request[i][j][t][z][y][x] goal = self.vargoal[j] + self.filegoal[i] if ("vector" in goal) or ("contour" in goal): if self.verbose: print "!! WARNING !! No operation was made on contours and vectors. This can be debatted actually." obj.field = obj_ref.field else: obj.field = obj_ref.field ** num else: print "!! ERROR !! To define a power, either an int or a float is needed." ; exit() return the_clone ### TBD: reverse power? for exponentials? ############################################################################################## # define method # --------- # ... (file and var are either one string or a vector of strings) # ... the goal of define is to define a 2D array of onerequest() objects (see class below) # given the number of file, var, x, y, z, t asked by the user # ... objectives for file or var are given through filegoal and vargoal # --> possible values: main contour vector # --------- # ... then onerequest() objects are being defined more precisely # by getting index_x index_y index_z index_t # and setting method_x method_y method_z method_t to either # - "free" for free dimensions (plot dimensions) # - "comp" for averages, max, min # - "fixed" for fixed dimensions (possibly several i.e. multislice) ############################################################################################## def define(self): self.printstatus() # initial check and get dimensions self.file = checktab(self.file,mess="file") self.nfin = len(self.file) if self.verbose: for i in range(self.nfin): inspect(self.file[i]) self.var = checktab(self.var,mess="var") self.nvin = len(self.var) # check goal tabs for files and variables # ... default is to plot everything if self.filegoal is None: self.filegoal = ["main"]*self.nfin if self.vargoal is None: self.vargoal = ["main"]*self.nvin self.filegoal = checktab(self.filegoal, mess="filegoal") self.vargoal = checktab(self.vargoal, mess="vargoal") if len(self.filegoal) != self.nfin: print "!! ERROR !! filegoal must be the same size as file." ; exit() if len(self.vargoal) != self.nvin: print "!! ERROR !! vargoal must be the same size as var." ; exit() # variables: initial check self.x = checktab(self.x,mess="x",allownone=True,allownumber=True) self.y = checktab(self.y,mess="y",allownone=True,allownumber=True) self.z = checktab(self.z,mess="z",allownone=True,allownumber=True) self.t = checktab(self.t,mess="t",allownone=True,allownumber=True) # for the moment not var- nor file- dependent. # but this could be the case. sx = readslices(self.x) ; sy = readslices(self.y) sz = readslices(self.z) ; st = readslices(self.t) # get methods mx = findmethod(sx) ; my = findmethod(sy) mz = findmethod(sz) ; mt = findmethod(st) # get number of plots to be done if mx in ["fixed","comp"]: self.nplotx = sx.size/2 else: self.nplotx = 1 if my in ["fixed","comp"]: self.nploty = sy.size/2 else: self.nploty = 1 if mz in ["fixed","comp"]: self.nplotz = sz.size/2 else: self.nplotz = 1 if mt in ["fixed","comp"]: self.nplott = st.size/2 else: self.nplott = 1 if self.verbose: print "**** OK. Plots over x,y,z,t -->",self.nplotx,self.nploty,self.nplotz,self.nplott # create the list of onerequest() objects self.request = [[[[[[ \ onerequest() \ for x in range(self.nplotx)] for y in range(self.nploty)] \ for z in range(self.nplotz)] for t in range(self.nplott)] \ for j in range(self.nvin)] for i in range(self.nfin)] # loop on onerequest() objects for i in range(self.nfin): for j in range(self.nvin): for t in range(self.nplott): for z in range(self.nplotz): for y in range(self.nploty): for x in range(self.nplotx): obj = self.request[i][j][t][z][y][x] # fill in names for files and variables obj.verbose = self.verbose obj.file = self.file[i] obj.var = self.var[j] # indicate the computation method obj.compute = self.compute # open the files (the same file might be opened several times but this is cheap) obj.openfile() ### get x,y,z,t dimensions from file obj.getdim() ### get methods obj.method_x = mx ; obj.method_y = my obj.method_z = mz ; obj.method_t = mt ### get index obj.getindextime(dalist=st,ind=t,stride=self.stridet) obj.getindexvert(dalist=sz,ind=z,stride=self.stridez) obj.getindexhori(dalistx=sx,dalisty=sy,indx=x,indy=y,stridex=self.stridex,stridey=self.stridey) # change status self.status = "defined" return self ############################################################################################## # retrieve method # --> for each element onerequest() in the array, get field .var from .f file # --> see below the onerequest() class: # - only get what is needed for computing and plotting # - averages etc... are computed here # --> RESULT: each onerequest() object has now its attribute .field filled # --> if one wants to perform operations on fields, this should be done after retrieve() ############################################################################################## def retrieve(self): self.printstatus() # check if things were done OK before if self.status != "defined": print "!! ERROR !! Please use .define() to define your pp object." ; exit() # create the list of allfield() objects # --> so that the user can easily access values self.allfield = [[[[[[ \ [] \ for x in range(self.nplotx)] for y in range(self.nploty)] \ for z in range(self.nplotz)] for t in range(self.nplott)] \ for j in range(self.nvin)] for i in range(self.nfin)] ## first get fields ## ... only what is needed is extracted from the files ## ... and computations are performed for i in range(self.nfin): for j in range(self.nvin): for t in range(self.nplott): for z in range(self.nplotz): for y in range(self.nploty): for x in range(self.nplotx): obj = self.request[i][j][t][z][y][x] obj.getfield() obj.computations() self.allfield[i][j][t][z][y][x] = obj.field # change status self.status = "retrieved" return self ########################################################## # get: a shortcut method for the define + retrieve chain # ########################################################## def get(self): self.define() self.retrieve() return self ######################################## # smooth: smooth the field in 1D or 2D # ######################################## ## TBD: smooth not OK with masked array in the end of retrieve() def smooth(self,window): if self.verbose: print "!! WARNING !! Performing a smoothing with a window size",window print "!! WARNING !! To come back to unsmoothed file, use .get() again" for i in range(self.nfin): for j in range(self.nvin): for t in range(self.nplott): for z in range(self.nplotz): for y in range(self.nploty): for x in range(self.nplotx): obj = self.request[i][j][t][z][y][x] if obj.field.ndim == 1: print "!! ERROR !! 1D smoothing not supported yet because reduces array sizes." exit() # TBD TBD TBD #obj.field = ppcompute.smooth1d(obj.field,window=window) elif obj.field.ndim == 2: obj.field = ppcompute.smooth2d(obj.field,window=window) ############################################################################################## # defineplot method # --> defineplot first defines arrays of plot objects and set each of them # ... simple looping except cases where goal is not main (e.g. contour or vector) # --> principle: each onerequest() object gives birth to a subplot # --> defineplot vs. makeplot: defining plot and actually plotting it are clearly separated # --> THE KEY OUPUT OF defineplot IS AN ARRAY self.p OF PLOT OBJECTS # optional arguments # --> extraplot: to indicate a number of plots to be added afterwards (use self.plotin) # --> loadfile: to use self.p from a previously saved file ############################################################################################## def defineplot(self,extraplot=0,loadfile=None): # ----------------------------------------------------- # LOAD MODE: load a self.p object. count plots from it. # ----------------------------------------------------- if loadfile is not None: try: filehandler = open(loadfile, 'r') ; self.p = pickle.load(filehandler) except IOError: print "!! ERROR !! Cannot find object file to load." ; exit() self.status = "definedplot" ; self.plotin = None self.nplot = len(self.p) ; self.howmanyplots = self.nplot return # ----------------------------------------------------- # REGULAR MODE # ----------------------------------------------------- self.printstatus() # check if things were done OK before if self.status in ["init","defined"]: print "!! ERROR !! Please use .retrieve() to get fields for plots with your pp object." ; exit() # check self.plotin (an existing fig on which to add plots afterwards) if self.plotin.__class__.__name__ == "pp": if self.plotin.fig is None: self.plotin = None # this is an additional security in case # a pp object is given without figure opened yet. elif self.plotin is not None: print "!! ERROR !! plotin argument must be a pp object." ; exit() # initialize the array of subplot objects # either something new or attributes coming from plotin object if self.plotin is None: self.p = [ ] else: self.p = self.plotin.p # create an array of subplot objects # ... in theory the order of looping can be changed without any harm # ... the only important thing is to keep i,j,t,z,y,x resp. for file,var,t,z,y,x count = 0 for i in range(self.nfin): if self.filegoal[i] == "main": for j in range(self.nvin): if self.vargoal[j] == "main": for t in range(self.nplott): for z in range(self.nplotz): for y in range(self.nploty): for x in range(self.nplotx): # look at dimension and append the right plot object obj = self.request[i][j][t][z][y][x] dp = obj.dimplot if dp == 1 or self.forcedimplot == 1: plobj = ppplot.plot1d() elif dp == 2 or self.forcedimplot == 2: plobj = ppplot.plot2d() elif dp == 0: print "**** OK. VALUES VALUES VALUES",obj.field else: print "!! ERROR !! 3D or 4D plots not supported" ; exit() # load abscissa and ordinate in obj obj.definecoord() # we start to define things here before appending # (convenient: could be overridden by the user before makeplot) # ... the if loop is necessary so that we can loop above on the dp=0 case if dp in [1,2]: # and define what to do in plobj plobj.invert = obj.invert_axes plobj.swap = obj.swap_axes # axis labels plobj.xlabel = obj.absclab ; plobj.ylabel = obj.ordilab # superpose or not (this is mostly for saving purpose) plobj.superpose = self.superpose # get title, colormaps, labels, etc.. from var plobj.var = obj.var plobj.define_from_var() # generic 1D/2D: load field and coord in plot object plobj.field = obj.field # field to be plotted plobj.absc = obj.absc # abscissa (or longitude) plobj.ordi = obj.ordi # ordinate (or latitude) # -- useless in 1D but not used anyway # specific 1D plot stuff if dp == 1: # -- a default label plobj.label = "" if self.nfin > 1: plobj.label = plobj.label + " file #"+str(i+1) if self.nvin > 1: plobj.label = plobj.label + " var #"+str(j+1) if self.nplott > 1: plobj.label = plobj.label + " t #"+str(t+1) if self.nplotz > 1: plobj.label = plobj.label + " z #"+str(z+1) if self.nploty > 1: plobj.label = plobj.label + " y #"+str(y+1) if self.nplotx > 1: plobj.label = plobj.label + " x #"+str(x+1) # specific 2d plot stuff if dp == 2: # -- light grey background for missing values if type(plobj.field).__name__ in 'MaskedArray': plobj.axisbg = '0.75' # -- define if it is a map or a plot plobj.mapmode = ( obj.method_x+obj.method_y == "freefree" \ and "grid points" not in obj.name_x \ and not self.noproj ) # possible user-defined plot settings shared by all plots if self.div is not None: plobj.div = self.div if self.xlabel is not None: plobj.xlabel = self.xlabel if self.xcoeff is not None: plobj.xcoeff = self.xcoeff if self.ylabel is not None: plobj.ylabel = self.ylabel if self.ycoeff is not None: plobj.ycoeff = self.ycoeff if self.title is not None: plobj.title = self.title # -- 1D specific if dp == 1: if self.lstyle is not None: plobj.lstyle = self.lstyle if self.marker is not None: plobj.marker = self.marker if self.color is not None: plobj.color = self.color if self.label is not None: plobj.label = self.label # -- 2D specific elif dp == 2: if self.proj is not None and not self.noproj: plobj.proj = self.proj if self.vmin is not None: plobj.vmin = self.vmin if self.vmax is not None: plobj.vmax = self.vmax if self.colorb is not None: plobj.colorb = self.colorb # finally append plot object self.p.append(plobj) count = count + 1 # self.nplot is number of plot to be defined in this call to defineplot() # (because of self.plotin this might less than length of self.p) self.nplot = count # --- superimposed contours and vectors --- # we have to start another loop because we need forward information # TBD: there is probably a more flexible way to do that count = 0 for i in range(self.nfin): for j in range(self.nvin): for t in range(self.nplott): for z in range(self.nplotz): for y in range(self.nploty): for x in range(self.nplotx): goal = self.vargoal[j] + self.filegoal[i] obj = self.request[i][j][t][z][y][x] if "mainmain" in goal and obj.dimplot == 2: # the plot object we consider in the loop pl = self.p[count] # -- see if there is a contour requested... # (we use try because we might be at the end of the list) found = 0 try: condvar = self.vargoal[j+1] except: condvar = "itisok" try: condfile = self.filegoal[i+1] except: condfile = "itisok" # ... get contour ########################################## # NB: contour is expected to be right after main otherwise it is not displayed ########################################## if condvar == "contour": plobj.addcontour = self.request[i][j+1][t][z][y][x].field ; found += 1 if condfile == "contour": plobj.addcontour = self.request[i+1][j][t][z][y][x].field ; found += 1 # see if there is a vector requested... # (we use try because we might be at the end of the list) try: condvar = self.vargoal[j+found+1]+self.vargoal[j+found+2] except: condvar = "itisok" try: condfile = self.filegoal[i+found+1]+self.filegoal[i+found+2] except: condfile = "itisok" # ... get vector and go directly to the next iteration # (in some cases we would do this twice but this is cheap) if "vector" in condvar: plobj.addvecx = self.request[i][j+found+1][t][z][y][x].field plobj.addvecy = self.request[i][j+found+2][t][z][y][x].field if "vector" in condfile: plobj.addvecx = self.request[i+found+1][j][t][z][y][x].field plobj.addvecy = self.request[i+found+2][j][t][z][y][x].field count = count + 1 # COUNT PLOTS. if 0 just exit. # self.howmanyplots is self.nplot + possible extraplots self.howmanyplots = self.nplot + extraplot if self.howmanyplots > 0: if self.verbose: print "**** OK. expect %i plots" % (self.howmanyplots) else: exit() # because this means that we only had 0D values ! # final status self.status = "definedplot" return self ############################################################################################## # makeplot method # --> after defineplot and before makeplot, user-defined plot settings can be easily given # simply by modifying the attributes of each elements of self.p # --> to change only one plot setting, no need to call defineplot again before makeplot # --> in the end, the array self.p of plot objects is saved for easy and convenient replotting # --> see practical examples in the folder 'examples' ############################################################################################## def makeplot(self): self.printstatus() # a few initial operations # ------------------------ if "definedplot" not in self.status: print "!! ERROR !! Please use .defineplot() before .makeplot() can be used with your pp object." ; exit() # open a figure and define subplots # --------------------------------- if self.plotin is None: # start from scratch self.fig = mpl.figure(figsize=(16,8)) self.subv,self.subh = ppplot.definesubplot(self.howmanyplots,self.fig) self.n = 0 ## adapted space for labels etc ## ... except for ortho because there is no label anyway self.customplot = self.p[0].field.ndim == 2 \ and self.p[0].mapmode == True \ and self.p[0].proj not in ["ortho"] if self.customplot: margin = 0.07 self.fig.subplots_adjust(left=margin,right=1-margin,bottom=margin,top=1-margin) else: # start from an existing figure. # extraplot must have been set in the call to the previous figure. self.fig = self.plotin.fig self.subv,self.subh = self.plotin.subv,self.plotin.subh self.n = self.plotin.n self.howmanyplots = self.plotin.howmanyplots self.customplot = self.plotin.customplot # LOOP on all subplots # NB: cannot use 'for pl in self.p' if self.plotin not None # -------------------- for count in range(self.nplot): # the plot object we consider in the loop pl = self.p[self.n] # before making the plot, create a subplot. the first one is numbered 1 not 0. # ... if pl.superpose, we use only one and only figure # ... (and we have to be careful with not doing things several times) if pl.superpose: if self.n == 0: self.fig.add_subplot(1,1,1,axisbg=pl.axisbg) # define one subplot (still needed for user-defined font sizes) sav = pl.xlabel,pl.ylabel,pl.xcoeff,pl.ycoeff,pl.title,pl.swaplab # save titles and labels else: pl.invert = False ; pl.lstyle = None # don't invert again axis # set saved titles and labels if self.plotin is None: pl.xlabel,pl.ylabel,pl.xcoeff,pl.ycoeff,pl.title,pl.swaplab = sav else: prev_plot = self.plotin.p[self.n-1] pl.xlabel = prev_plot.xlabel pl.ylabel = prev_plot.ylabel pl.xcoeff = prev_plot.xcoeff pl.ycoeff = prev_plot.ycoeff pl.title = prev_plot.title pl.swaplab = prev_plot.swaplab else: self.fig.add_subplot(self.subv,self.subh,self.n+1,axisbg=pl.axisbg) if self.verbose: print "**** Done subplot %i / %i " %( self.n+1,self.howmanyplots ) # finally make the plot pl.make() # increment plot count (and propagate this in plotin) self.n = self.n+1 if self.plotin is not None: self.plotin.n = self.n # once completed show the plot (cannot show intermediate plotin) # ... added a fix (customplot=True) for the label problem in basemap print "**** PPCLASS. Done step: makeplot" if (self.n == self.howmanyplots): ppplot.save(mode=self.out,filename=self.filename,folder=self.folder,custom=self.customplot,includedate=self.includedate) mpl.close() # SAVE A PICKLE FILE WITH THE self.p ARRAY OF OBJECTS if self.verbose: print "**** Saving session in "+self.filename + ".ppobj" savfile = self.folder + "/" + self.filename + ".ppobj" try: filehandler = open(savfile, 'w') pickle.dump(self.p, filehandler) except IOError: print "!! WARNING !! Saved object file not written. Probably do not have permission to write here." return self ########################################################### # plot: a shortcut method for the defineplot + plot chain # ########################################################### def plot(self,extraplot=0): self.defineplot(extraplot=extraplot) self.makeplot() return self ####################################################### # getplot: a shortcut method for the get + plot chain # ####################################################### def getplot(self,extraplot=0): self.get() self.plot(extraplot=extraplot) return self ################################################################### # getdefineplot: a shortcut method for the get + defineplot chain # ################################################################### def getdefineplot(self,extraplot=0): self.get() self.defineplot(extraplot=extraplot) return self ############################################################## # f: operation on two pp objects being on status 'definedplot' # this allows for one field being function of another one # e.g. u.f(v) means u will be displayed as a function of v # ... no need to do defineplot after u.f(v), makeplot directly ############################################################## def f(self,other): # preamble: for this operation to work, defineplot() must have been done if self.status != "definedplot": if self.verbose: print "!! WARNING !! performing defineplot on operand" self.defineplot() if other.status != "definedplot": if self.verbose: print "!! WARNING !! performing defineplot on operand" other.defineplot() # check total number of plots if self.howmanyplots != other.howmanyplots: print "!! ERROR !! The two operands do not have the same number of subplots." exit() # and now operation. count = 0 while count < self.howmanyplots: sobj = self.p[count] ; oobj = other.p[count] if sobj.field.ndim !=1 or oobj.field.ndim !=1: if self.verbose: print "!! WARNING !! Flattening arrays because more than one-dimensional." sobj.field = np.ravel(sobj.field) oobj.field = np.ravel(oobj.field) sobj.absc = oobj.field sobj.xlabel = oobj.ylabel if sobj.absc.size > sobj.field.size: if self.verbose: print "!! WARNING !! Trying to define y=f(x) with x and y not at the same size.",sobj.absc.size,sobj.field.size print "!! WARNING !! Modifying x to fit y size but please check." sobj.absc = sobj.absc[0:sobj.field.size] count = count + 1 return self ########################################################### # copyopt: get options from e.g. a parser # ... allow for simple scripting and user-defined settings # ... must be called between defineplot and makeplot # REQUIRED: attributes of opt must be the same as in the pp object ########################################################### def getopt(self,opt): # -- if only one, or less than the number of plots --> we take the first one # -- if as many as number of plots --> OK, each plot has its own setting # (except a few cases such as trans) for iii in range(self.howmanyplots): ### try: self.p[iii].trans = opt.trans except: pass ### try: self.p[iii].div = opt.div except: pass ### try: self.p[iii].logy = opt.logy except: pass ### try: self.p[iii].colorb = opt.colorb[iii] except: try: self.p[iii].colorb = opt.colorb[0] except: pass ### try: self.p[iii].title = opt.title[iii] except: try: self.p[iii].title = opt.title[0] except: pass ### try: self.p[iii].xlabel = opt.xlabel[iii] except: try: self.p[iii].xlabel = opt.xlabel[0] except: pass ### try: self.p[iii].ylabel = opt.ylabel[iii] except: try: self.p[iii].ylabel = opt.ylabel[0] except: pass ### try: self.p[iii].lstyle = opt.lstyle[iii] except: try: self.p[iii].lstyle = opt.lstyle[0] except: pass ### try: self.p[iii].color = opt.color[iii] except: try: self.p[iii].color = opt.color[0] except: pass ### try: self.p[iii].marker = opt.marker[iii] except: try: self.p[iii].marker = opt.marker[0] except: pass ### try: self.p[iii].label = opt.label[iii] except: try: self.p[iii].label = opt.label[0] except: pass ### try: self.p[iii].proj = opt.proj[iii] except: try: self.p[iii].proj = opt.proj[0] except: pass ### try: self.p[iii].back = opt.back[iii] except: try: self.p[iii].back = opt.back[0] except: pass ### try: self.p[iii].area = opt.area[iii] except: try: self.p[iii].area = opt.area[0] except: pass ### try: self.p[iii].blon = opt.blon[iii] except: try: self.p[iii].blon = opt.blon[0] except: pass ### try: self.p[iii].blat = opt.blat[iii] except: try: self.p[iii].blat = opt.blat[0] except: pass ### try: self.p[iii].vmin = opt.vmin[iii] except: try: self.p[iii].vmin = opt.vmin[0] except: pass ### try: self.p[iii].vmax = opt.vmax[iii] except: try: self.p[iii].vmax = opt.vmax[0] except: pass ########################################################## ### THE ONEREQUEST SUBOBJECT TO PP (ON WHICH IT LOOPS) ### ########################################################## class onerequest(): # default settings. mostly initialized to diagnose problem, except dimplot, nplot, verbose, swap_axes, invert_axes # ------------------------------- def __init__(self): self.file = '!! file: I am not set, damned !!' self.f = None self.dim = None self.var = '!! var: I am not set, damned !!' self.index_x = [] ; self.index_y = [] ; self.index_z = [] ; self.index_t = [] self.index_x2d = [] ; self.index_y2d = [] self.method_x = '!! method_x: I am not set, damned !!' self.method_y = '!! method_y: I am not set, damned !!' self.method_z = '!! method_z: I am not set, damned !!' self.method_t = '!! method_t: I am not set, damned !!' self.field = None self.name_x = None ; self.name_y = None ; self.name_z = None ; self.name_t = None self.dim_x = None ; self.dim_y = None ; self.dim_z = None ; self.dim_t = None self.field_x = None ; self.field_y = None ; self.field_z = None ; self.field_t = None self.dimplot = 0 self.nplot = 1 self.absc = None ; self.ordi = None ; self.absclab = None ; self.ordilab = None self.verbose = True self.swap_axes = False ; self.invert_axes = False self.compute = None # open a file. for now it is netcdf. TBD for other formats. # check that self.var is inside. # ------------------------------- def openfile(self): if not os.path.exists(self.file): print '!! ERROR !! I could not find the following file: '+self.file ; exit() if not os.path.isfile(self.file): print '!! ERROR !! This does not appear to be a file: '+self.file ; exit() self.f = netCDF4.Dataset(self.file) if self.verbose: print "**** OK. Opened file "+self.file if self.var not in self.f.variables.keys(): print '!! ERROR !! File '+self.file+' does not contain variable: '+self.var print '..... try instead with ',self.f.variables.keys() ; exit() # copy attributes from another existing object # -------------------------------------------- def copy(self,source): for k, v in vars(source).items(): setattr(self,k,v) # get x,y,z,t dimensions from NETCDF file # TBD: user could request for a specific altitude dimension # TBD: staggered variables could request specific dimensions # ------------------------------- def getdim(self): # GET SIZES OF EACH DIMENSION if self.verbose: print "**** OK. Found variable "+self.var shape = self.f.variables[self.var].shape self.dim = len(shape) if self.dim == 1: if self.verbose: print "**** OK. 1D field. I assume this varies with time." self.dim_x = 1 ; self.dim_y = 1 ; self.dim_z = 1 ; self.dim_t = shape[0] elif self.dim == 2: if self.verbose: print "**** OK. 2D field. I assume this is not-time-varying lat-lon map." self.dim_x = shape[1] ; self.dim_y = shape[0] ; self.dim_z = 1 ; self.dim_t = 1 elif self.dim == 3: if self.verbose: print "**** OK. 3D field. I assume this is time-varying lat-lon map." self.dim_x = shape[2] ; self.dim_y = shape[1] ; self.dim_z = 1 ; self.dim_t = shape[0] elif self.dim == 4: if self.verbose: print "**** OK. 4D field." self.dim_x = shape[3] ; self.dim_y = shape[2] ; self.dim_z = shape[1] ; self.dim_t = shape[0] # LONGITUDE. Try preset fields. If not present set grid points axis. self.name_x = "nothing" for c in glob_listx: if c in self.f.variables.keys(): self.name_x = c if self.name_x == "nothing": self.field_x = np.array(range(self.dim_x)) self.name_x = "x grid points" else: self.field_x = self.f.variables[self.name_x] # LATITUDE. Try preset fields. If not present set grid points axis. self.name_y = "nothing" for c in glob_listy: if c in self.f.variables.keys(): self.name_y = c if self.name_y == "nothing": self.field_y = np.array(range(self.dim_y)) self.name_y = "y grid points" else: self.field_y = self.f.variables[self.name_y] # ensure that lon and lat are 2D fields # 1. simple 1D case (not time-varying) if len(self.field_x.shape)*len(self.field_y.shape) == 1: if self.verbose: print "**** OK. recasting lon and lat as 2D fields." [self.field_x,self.field_y] = np.meshgrid(self.field_x,self.field_y) # 2. complex 3D case (time-varying, actually just copied over time axis) elif len(self.field_x.shape)*len(self.field_y.shape) == 9: if self.verbose: print "**** OK. reducing lon and lat as 2D fields. get rid of time." self.field_x = self.field_x[0,:,:] self.field_y = self.field_y[0,:,:] # if xy axis are apparently undefined, set 2D grid points axis. if "grid points" not in self.name_x: if self.field_x.all() == self.field_x[0,0]: print "!! WARNING !! xy axis look undefined. creating non-dummy ones." self.field_x = np.array(range(self.dim_x)) ; self.name_x = "x grid points" self.field_y = np.array(range(self.dim_y)) ; self.name_y = "y grid points" [self.field_x,self.field_y] = np.meshgrid(self.field_x,self.field_y) if self.dim_x > 1: if self.verbose: print "**** OK. x axis %4.0f values [%5.1f,%5.1f]" % (self.dim_x,self.field_x.min(),self.field_x.max()) if self.dim_y > 1: if self.verbose: print "**** OK. y axis %4.0f values [%5.1f,%5.1f]" % (self.dim_y,self.field_y.min(),self.field_y.max()) # ALTITUDE. Try preset fields. If not present set grid points axis. # WARNING: how do we do if several are available? self.name_z = "nothing" for c in glob_listz: if c in self.f.variables.keys(): self.name_z = c if self.name_z == "nothing": self.field_z = np.array(range(self.dim_z)) self.name_z = "z grid points" else: self.field_z = self.f.variables[self.name_z][:] # specify dimension # TBD: have to check that this is not a 3D field if self.dim_z > 1: if self.verbose: print "**** OK. z axis %4.0f values [%5.1f,%5.1f]" % (self.dim_z,self.field_z.min(),self.field_z.max()) # TIME. Try preset fields. self.name_t = "nothing" for c in glob_listt: if c in self.f.dimensions.keys(): self.name_t = c try: # speed up: only get first value, last one. dafirst = self.f.variables[self.name_t][0] dalast = self.f.variables[self.name_t][self.dim_t-1] self.field_t = np.linspace(dafirst,dalast,num=self.dim_t) if dafirst == dalast: self.field_t = np.array([dafirst]) except: # ... or if a problem encountered, define a simple time axis self.field_t = np.array(range(self.dim_t)) self.name_t = "t grid points" if self.dim_t > 1: if self.verbose: print "**** OK. t axis %4.0f values [%5.1f,%5.1f]" % (self.dim_t,self.field_t.min(),self.field_t.max()) # get list of index to be retrieved for time axis ### TBD: il faudrait ne prendre que les indices qui correspondent a l interieur d un plot (dans all) # ------------------------------- def getindextime(self,dalist=None,ind=None,stride=1): if self.method_t == "free": self.index_t = np.arange(0,self.dim_t,stride) if self.dim_t > 1: self.dimplot = self.dimplot + 1 if self.verbose: print "**** OK. t values. all." else: self.method_t = "fixed" if self.verbose: print "**** OK. no t dimension." elif self.method_t == "comp": start = np.argmin( np.abs( self.field_t - dalist[ind][0] ) ) stop = np.argmin( np.abs( self.field_t - dalist[ind][1] ) ) self.index_t = np.arange(start,stop,stride) if self.verbose: print "**** OK. t values. comp over interval ",self.field_t[start],self.field_t[stop]," nvalues=",self.index_t.size elif self.method_t == "fixed": self.index_t.append( np.argmin( np.abs( self.field_t - dalist[ind][0] ) )) if self.verbose: print "**** OK. t values",self.field_t[self.index_t] else: print "!! ERROR !! method "+self.method_t+" not supported" self.index_t = np.array(self.index_t) # get list of index to be retrieved for vertical axis ### TBD: il faudrait ne prendre que les indices qui correspondent a l interieur d un plot (dans all) # ------------------------------- def getindexvert(self,dalist=None,ind=None,stride=1): if self.method_z == "free": self.index_z = np.arange(0,self.dim_z,stride) if self.dim_z > 1: self.dimplot = self.dimplot + 1 if self.verbose: print "**** OK. z values. all." else: self.method_z = "fixed" if self.verbose: print "**** OK. no z dimension." elif self.method_z == "comp": start = np.argmin( np.abs( self.field_z - dalist[ind][0] ) ) stop = np.argmin( np.abs( self.field_z - dalist[ind][1] ) ) self.index_z = np.arange(start,stop,stride) if self.verbose: print "**** OK. z values. comp over interval",self.field_z[start],self.field_z[stop]," nvalues=",self.index_z.size elif self.method_z == "fixed": self.index_z.append( np.argmin( np.abs( self.field_z - dalist[ind][0] ) )) if self.verbose: print "**** OK. z values",self.field_z[self.index_z] else: if self.verbose: print "!! ERROR !! method "+self.method_z+" not supported" self.index_z = np.array(self.index_z) # get list of index to be retrieved for horizontal grid # --> index_x and index_y are slices to be retrieved from NETCDF files # --> index_x2d and index_y2d are the actual (x,y) coordinates corresponding to each relevant point # [this is slightly more complicated because 2D arrays for lat-lon projection possibly irregular] # NB: to append index we use lists (the most convenient) then we convert into a numpy.array ### TBD: il faudrait ne prendre que les indices qui correspondent a l interieur d un plot (dans all) # ------------------------------- def getindexhori(self,dalistx=None,dalisty=None,indx=None,indy=None,stridex=1,stridey=1): ## get what is the method over x and y axis test = self.method_x+self.method_y ## CASE 0, EASY CASES: ## - LAT IS FREE (we do here what must be done whatever LON case is) ## - LON IS FREE (we do here what must be done whatever LAT case is) ## - LAT IS COMP AND LON IS FREE ## - LON IS COMP AND LAT IS FREE if self.method_x == "free" or test in ["compfree","compcomp"]: self.index_x = range(0,self.dim_x,stridex) if self.dim_x > 1: if self.method_x == "free": self.dimplot = self.dimplot + 1 if self.verbose: print "**** OK. x values. all." else: self.method_x = "fixed" if self.verbose: print "**** OK. no x dimension." if self.method_y == "free" or test in ["freecomp","compcomp"]: self.index_y = range(0,self.dim_y,stridey) if self.dim_y > 1: if self.method_y == "free": self.dimplot = self.dimplot + 1 if self.verbose: print "**** OK. y values. all." else: self.method_y = "fixed" if self.verbose: print "**** OK. no y dimension." ## CASE 0 above, this is just for continuity for free. ## ... for comp we have to select bounds. ## ... TBD: take the bool array strategy for what follows! if self.method_x in ["free","comp"] and self.method_y in ["free","comp"]: ### ref1_dirty_hack ### ... for the moment this is a hack. but actually this is more powerful. if self.method_x == "comp": yeah = (self.field_x >= dalistx[indx][0])*(self.field_x <= dalistx[indx][1]) self.index_x = yeah[0,:] if self.method_y == "comp": yeah = (self.field_y >= dalisty[indy][0]) * (self.field_y <= dalisty[indy][1]) self.index_y = yeah[:,0] self.index_x2d = self.index_x self.index_y2d = self.index_y ## AND NOW THE LITTLE BIT MORE COMPLICATED CASES ## CASE 1 LAT AND LON ARE FIXED elif test == "fixedfixed": idy,idx = np.unravel_index( np.argmin( ( self.field_x - dalistx[indx][0])**2 + (self.field_y - dalisty[indy][0])**2 ), self.field_x.shape ) #TBD: pb with staggered coord if idx not in self.index_x: self.index_x.append(idx) if idy not in self.index_y: self.index_y.append(idy) self.index_x2d.append(idx) self.index_y2d.append(idy) ## CASE 2 LON IS FIXED BUT NOT LAT elif test in ["fixedfree","fixedcomp"]: # find where are requested x values for each y on the free dimension # NB: this does not work for non-bijective cases e.g. polar stereographic for iy in range(self.dim_y): idx = np.argmin( np.abs( self.field_x[iy,:] - dalistx[indx][0] ) ) # if comp is requested we select only indexes which yield values between requested min and max storeval = (self.method_y == "comp") and (self.field_y[iy,idx] > dalisty[indy][0]) and (self.field_y[iy,idx] < dalisty[indy][1]) storeval = storeval or (self.method_y == "free") if storeval: if idx not in self.index_x: self.index_x.append(idx) if iy not in self.index_y and self.method_y == "comp": self.index_y.append(iy) if idx not in self.index_x2d or iy not in self.index_y2d: self.index_x2d.append(idx) self.index_y2d.append(iy) ## CASE 3 LAT IS FIXED BUT NOT LON elif test in ["freefixed","compfixed"]: # find where are requested y values for each x on the free dimension # NB: this does not work for non-bijective cases e.g. polar stereographic for ix in range(self.dim_x): idy = np.argmin( np.abs( self.field_y[:,ix] - dalisty[indy][0] ) ) # if comp is requested we select only indexes which yield values between requested min and max storeval = (self.method_x == "comp") and (self.field_x[idy,ix] > dalistx[indx][0]) and (self.field_x[idy,ix] < dalistx[indx][1]) storeval = storeval or (self.method_x == "free") if storeval: if idy not in self.index_y: self.index_y.append(idy) if ix not in self.index_x and self.method_x == "comp": self.index_x.append(ix) if ix not in self.index_x2d or idy not in self.index_y2d: self.index_x2d.append(ix) self.index_y2d.append(idy) ## check index tab if len(self.index_x) == 0 or len(self.index_y) == 0: print "!! ERROR !! no indices found. check prescribed latitudes or longitudes" ; exit() ## ensure the array is a numpy array for getfield to work self.index_x = np.array(self.index_x) self.index_y = np.array(self.index_y) self.index_x2d = np.array(self.index_x2d) self.index_y2d = np.array(self.index_y2d) ### print extrema printx = self.field_x[np.ix_(self.index_y2d, self.index_x2d)] printy = self.field_y[np.ix_(self.index_y2d, self.index_x2d)] if self.verbose: print "**** OK. x values (min,max).", printx.min(),printx.max() print "**** OK. y values (min,max).", printy.min(),printy.max() # get the field from the NETCDF file and perform averages # ------------------------------- def getfield(self): ## first tell what is to be done if self.dimplot > 2: print "**** !! ERROR !! "+str(self.dimplot)+"D plots not supported!" ; exit() elif self.dimplot == 0 and self.verbose: print "**** OK. 0D value requested." elif self.dimplot == 1 and self.verbose: print "**** OK. 1D plot requested." elif self.verbose: print "**** OK. 2D section requested." # well, now get field from netcdf file # part below is necessary otherwise there is an index error below if self.index_x.size == 1: self.index_x = self.index_x[0] if self.index_y.size == 1: self.index_y = self.index_y[0] if self.index_z.size == 1: self.index_z = self.index_z[0] if self.index_t.size == 1: self.index_t = self.index_t[0] # then retrieve what is requested by user # each self.dim case corresponds to tests in the beginning of getdim. time0 = timelib.time() if self.verbose: print "**** OK. I am getting values from files. Please wait." if self.dim == 1: nt = self.index_t.size ; nz = 1 ; ny = 1 ; nx = 1 self.field = self.f.variables[self.var][self.index_t] elif self.dim == 2: nt = 1 ; nz = 1 ; ny = self.index_y.size ; nx = self.index_x.size self.field = self.f.variables[self.var][self.index_y,self.index_x] elif self.dim == 3: nt = self.index_t.size ; nz = 1 ; ny = self.index_y.size ; nx = self.index_x.size self.field = self.f.variables[self.var][self.index_t,self.index_y,self.index_x] # this is far faster than retrieving each term with a loop elif self.dim == 4: nt = self.index_t.size ; nz = self.index_z.size ; ny = self.index_y.size ; nx = self.index_x.size self.field = self.f.variables[self.var][self.index_t,self.index_z,self.index_y,self.index_x] else: print "!! ERROR !! field would have more than four dimensions ?" ; exit() # dirty hack (AS) ref1_dirty_hack # waiting for more fundamental modifications. case when self.index_y is a bool array. # ... be careful if no point... if type(self.index_x[0]) == np.bool_: nx = np.sum(self.index_x) ## gives the size of the True part! if type(self.index_y[0]) == np.bool_: ny = np.sum(self.index_y) ## gives the size of the True part! # NB: ... always 4D array but possibly with "size 1" dimensions # ... if one dimension is missing because 1D 2D or 3D requests, make it appear again self.field = np.reshape(self.field,(nt,nz,ny,nx)) if self.verbose: print "**** OK. I got %7.1e values. This took me %6.4f seconds" % (nx*ny*nz*nt,timelib.time() - time0) if self.verbose: print "**** OK. I got var "+self.var+" with shape",self.field.shape # reduce coordinates to useful points # ... TBD: this should be ordered in the case of non-regular projections if self.method_x in ["free","comp"] and self.method_y in ["free","comp"]: # we need 2D coordinates (free) or we get broadcast problem (comp) so we use np.ix self.field_x = self.field_x[np.ix_(self.index_y2d, self.index_x2d)] self.field_y = self.field_y[np.ix_(self.index_y2d, self.index_x2d)] else: # we are OK with 1D coordinates self.field_x = self.field_x[self.index_y2d, self.index_x2d] self.field_y = self.field_y[self.index_y2d, self.index_x2d] self.field_z = self.field_z[self.index_z] self.field_t = self.field_t[self.index_t] # extract relevant horizontal points # TBD: is compfree OK with computing on irregular grid? test = self.method_x + self.method_y if test in ["fixedfixed","freefree"]: pass elif test in ["fixedfree","fixedcomp"] or test in ["freefixed","compfixed"]: time0 = timelib.time() # now have to obtain the new indexes which correspond to the extracted self.field # for it to work with unique index, ensure that any index_* is a numpy array if not isinstance(self.index_x, np.ndarray): self.index_x = np.array([self.index_x]) if not isinstance(self.index_y, np.ndarray): self.index_y = np.array([self.index_y]) if not isinstance(self.index_z, np.ndarray): self.index_z = np.array([self.index_z]) if not isinstance(self.index_t, np.ndarray): self.index_t = np.array([self.index_t]) for val in self.index_x: self.index_x2d[np.where(self.index_x2d == val)] = np.where(self.index_x == val)[0] for val in self.index_y: self.index_y2d[np.where(self.index_y2d == val)] = np.where(self.index_y == val)[0] ##### VERY EXPENSIVE ## recast self.field with 2D horizontal arrays because we might have extracted ## more than what is to be actually plot or computed, in particular for comps on 2D lat,lon coordinates #self.field = self.field[np.ix_(self.index_t,self.index_z,self.index_y2d,self.index_x2d)] #(nt,nz,ny,nx) = self.field.shape # prepare the loop on all relevant horizontal points if self.method_x in ["comp","free"]: nnn = self.index_x2d.shape[0] ; what_I_am_supposed_to_do = "keepx" elif self.method_y in ["comp","free"]: nnn = self.index_y2d.shape[0] ; what_I_am_supposed_to_do = "keepy" # LOOP to extract only useful values over horizontal dimensions # only take diagonal terms, do not loop on all self.index_x2d*self.index_y2d # ... this method is fast enough, perhaps there is a faster way though # ... (for sure the method with np.diag is much slower) for iii in range(nnn): ix = self.index_x2d[iii] ; iy = self.index_y2d[iii] for iz in range(self.index_z.size): for it in range(self.index_t.size): if what_I_am_supposed_to_do == "keepx": self.field[it,iz,0,ix] = self.field[it,iz,iy,ix] elif what_I_am_supposed_to_do == "keepy": self.field[it,iz,iy,0] = self.field[it,iz,iy,ix] if self.verbose: print "**** OK. I got to pick the right values for your request. This took me %6.4f seconds" % (timelib.time() - time0) # we only keep the one value that was modified on the dimension which is not free if what_I_am_supposed_to_do == "keepx": self.field = self.field[:,:,0,:] ; ny = 1 ; self.field = np.reshape(self.field,(nt,nz,ny,nx)) elif what_I_am_supposed_to_do == "keepy": self.field = self.field[:,:,:,0] ; nx = 1 ; self.field = np.reshape(self.field,(nt,nz,ny,nx)) # make a mask in case there are non-NaN missing values. (what about NaN missing values?) # ... this is important for computations below (see ppcompute) masked = np.ma.masked_where(np.abs(self.field) > 1e25,self.field) if masked.mask.any() == True: if self.verbose: print "!! WARNING !! Values over +-1e25 are considered missing values." self.field = masked self.field.set_fill_value([np.NaN]) # perform computations # ------------------------------- # available: mean, max, min, meanarea # TB: integrals? for derivatives, define a function self.dx() def computations(self): nt,nz,ny,nx = self.field.shape # treat the case of mean on fields normalized with grid mesh area # ... this is done in the .area() method. # after that self.field contains field*area/totarea if "area" in self.compute: if "comp" in self.method_x+self.method_y: self.area() else: if self.verbose: print "!! WARNING !! No area accounted for (computing on t and/or z axis)." # now ready to compute [TBD: we would like to have e.g. mean over x,y and min over t ??] if self.method_t == "comp": if self.verbose: print "**** OK. Computing over t axis." if "mean" in self.compute: self.field = ppcompute.mean(self.field,axis=0) elif self.compute == "min": self.field = ppcompute.min(self.field,axis=0) elif self.compute == "max": self.field = ppcompute.max(self.field,axis=0) else: print "!! ERROR !! operation not supported." ; exit() nt = 1 ; self.field = np.reshape(self.field,(nt,nz,ny,nx)) if self.method_z == "comp": if self.verbose: print "**** OK. Computing over z axis." if "mean" in self.compute: self.field = ppcompute.mean(self.field,axis=1) elif self.compute == "min": self.field = ppcompute.min(self.field,axis=1) elif self.compute == "max": self.field = ppcompute.max(self.field,axis=1) else: print "!! ERROR !! operation not supported." ; exit() nz = 1 ; self.field = np.reshape(self.field,(nt,nz,ny,nx)) if self.method_y == "comp": if self.verbose: print "**** OK. Computing over y axis." if self.compute == "mean": self.field = ppcompute.mean(self.field,axis=2) elif self.compute == "min": self.field = ppcompute.min(self.field,axis=2) elif self.compute == "max": self.field = ppcompute.max(self.field,axis=2) elif self.compute == "meanarea": self.field = ppcompute.sum(self.field,axis=2) else: print "!! ERROR !! operation not supported." ; exit() ny = 1 ; self.field = np.reshape(self.field,(nt,nz,ny,nx)) if self.field_x.ndim == 2: self.field_x = self.field_x[0,:] # TBD: this is OK for regular grid but not for irregular if self.method_x == "comp": if self.verbose: print "**** OK. Computing over x axis." if self.compute == "mean": self.field = ppcompute.mean(self.field,axis=3) elif self.compute == "min": self.field = ppcompute.min(self.field,axis=3) elif self.compute == "max": self.field = ppcompute.max(self.field,axis=3) elif self.compute == "meanarea": self.field = ppcompute.sum(self.field,axis=3) else: print "!! ERROR !! operation not supported." ; exit() nx = 1 ; self.field = np.reshape(self.field,(nt,nz,ny,nx)) if self.field_y.ndim == 2: self.field_y = self.field_y[:,0] # TBD: this is OK for regular grid but not for irregular # remove all dimensions with size 1 to prepare plot (and check the resulting dimension with dimplot) self.field = np.squeeze(self.field) if self.field.ndim != self.dimplot: print "!! ERROR !! Problem: self.field is different than plot dimensions", self.field.ndim, self.dimplot ; exit() if self.verbose: print "**** OK. Final shape for "+self.var+" after averaging and squeezing",self.field.shape # get areas for computations and ponderate self.field by area/totarea # ------------------------------------------------------------------- def area(self): if self.verbose: print "**** OK. Get area array for computations." # create a request object for area # ... and copy known attributes from self aire = onerequest() aire.copy(self) # get area field name aire.var = "nothing" for c in glob_listarea: if c in aire.f.variables.keys(): aire.var = c # do not try to calculate areas automatically if aire.var == "nothing": print "!! ERROR !! area variable not found... needs to be added in set_ppclass.txt?" exit() # define area request dimensions aire.getdim() # ensure this is a 2D horizontal request and define indexes # ... areas are not supposed to vary with time and height aire.method_x = "free" ; aire.method_y = "free" aire.getindexhori() ; aire.dimplot = 2 aire.method_z = "fixed" ; aire.field_z = np.array([0]) ; aire.index_z = np.array([0]) aire.method_t = "fixed" ; aire.field_t = np.array([0]) ; aire.index_t = np.array([0]) # read the 2D area array in netCDF file aire.getfield() aire.field = np.squeeze(aire.field) # reduce with self horizontal indexes if "fixed" in self.method_x+self.method_y: aire.field = aire.field[self.index_y,self.index_x] # calculate total area # ... 2D comp is easy. 1D comp is a bit less easy but simple array manipulation. if "free" in self.method_x+self.method_y: if self.method_x == "free": totarea = ppcompute.sum(aire.field,axis=0) totarea = np.reshape(totarea,(1,totarea.size)) totarea = np.tile(totarea,(1,self.index_x)) elif self.method_y == "free": totarea = ppcompute.sum(aire.field,axis=1) totarea = np.reshape(totarea,(totarea.size,1)) totarea = np.tile(totarea,(1,self.index_x.size)) elif self.method_x == "comp" and self.method_y == "comp": aire.field = aire.field[np.ix_(self.index_y, self.index_x)] # reduce to requested indexes only totarea = ppcompute.sum(ppcompute.sum(aire.field,axis=1),axis=0) else: if self.verbose: print "!! WARNING !! Not account for areas. Only averaging over z and/or t axis." # normalize by total area print "**** OK. I can now normalize field by areas." aire.field = aire.field / totarea # tile area array over self t and z axis so that area field could be multiplied with self.field aire.field = np.tile(aire.field,(self.index_t.size,self.index_z.size,1,1)) if self.field.shape != aire.field.shape: print "!! ERROR !! Problem in area(). Check array shapes." print "Field vs. aire:",self.field.shape,aire.field.shape ; exit() else: self.field = self.field*aire.field # define coordinates for plot # ------------------------------- def definecoord(self): I_got_abs = False ; I_got_ord = False # here is the thing. time is usually taken as an abscissa so we start with time. if self.method_t == "free": self.absc = self.field_t ; self.absclab = self.name_t I_got_abs = True # then we usually have x as an abscissa. if self.method_x == "free": if I_got_abs: self.ordi = self.field_x ; self.ordilab = self.name_x I_got_ord = True else: self.absc = self.field_x ; self.absclab = self.name_x I_got_abs = True # ... or we have y if self.method_y == "free": if I_got_abs: self.ordi = self.field_y ; self.ordilab = self.name_y I_got_ord = True else: self.absc = self.field_y ; self.absclab = self.name_y I_got_abs = True # ... and we end with z because it is usually not an abscissa (profiles). if self.method_z == "free": if self.field_z[0] > self.field_z[1]: self.invert_axes = True # the axis will be turned upside-down if I_got_abs: self.ordi = self.field_z ; self.ordilab = self.name_z I_got_ord = True else: self.absc = self.field_z ; self.absclab = self.name_z I_got_abs = True self.swap_axes = True # says that altitude is not supposed to remain as an abscissa if I_got_abs and self.verbose: print "**** OK. abscissa:",self.absclab, self.absc.shape if I_got_ord and self.verbose: print "**** OK. ordinate:",self.ordilab, self.ordi.shape