def computeClimatology(experiment, filetype, domain, periods=None, offset=0, griddef=None, varlist=None,
ldebug=False, loverwrite=False, lparallel=False, pidstr='', logger=None):
''' worker function to compute climatologies for given file parameters. '''
# input type checks
if not isinstance(experiment,Exp): raise TypeError
if not isinstance(filetype,basestring): raise TypeError
if not isinstance(domain,(np.integer,int)): raise TypeError
if periods is not None and not (isinstance(periods,(tuple,list)) and isInt(periods)): raise TypeError
if not isinstance(offset,(np.integer,int)): raise TypeError
if not isinstance(loverwrite,(bool,np.bool)): raise TypeError
if griddef is not None and not isinstance(griddef,GridDefinition): raise TypeError
#if pidstr == '[proc01]': raise TypeError # to test error handling
# load source
dataset_name = experiment.name
fileclass = fileclasses[filetype] # used for target file name
tsfile = fileclass.tsfile.format(domain,'')
expfolder = experiment.avgfolder
filepath = '{:s}/{:s}'.format(expfolder, tsfile)
logger.info('\n\n{0:s} *** Processing Experiment {1:<15s} *** '.format(pidstr,"'{:s}'".format(dataset_name)) +
'\n{0:s} *** {1:^37s} *** \n'.format(pidstr,"'{:s}'".format(tsfile)))
# check file and read begin/enddates
if not os.path.exists(filepath):
#raise IOError, "Source file '{:s}' does not exist!".format(filepath)
# print message and skip
skipmsg = "\n{:s} >>> File '{:s}' in dataset '{:s}' is missing --- skipping!".format(pidstr,tsfile,dataset_name)
skipmsg += "\n{:s} >>> ('{:s}')\n".format(pidstr,filepath)
logger.warning(skipmsg)
# N.B.: this can cause a lot of error messages, when not all files are present
else: # if monthly source file exists
import netCDF4 as nc
ncfile = nc.Dataset(filepath,mode='r')
begintuple = ncfile.begin_date.split('-')
endtuple = ncfile.end_date.split('-')
ncfile.close()
# N.B.: at this point we don't want to initialize a full GDAL-enabled dataset, since we don't even
# know if we need it, and it creates a lot of overhead
# determine age of source file
if not loverwrite: sourceage = datetime.fromtimestamp(os.path.getmtime(filepath))
# figure out start date
filebegin = int(begintuple[0]) # first element is the year
fileend = int(endtuple[0]) # first element is the year
begindate = offset + filebegin
if not ( filebegin <= begindate <= fileend ): raise DateError
# handle cases where the first month in the record is not January
firstmonth = int(begintuple[1]) # second element is the month
shift = firstmonth-1 # will be zero for January (01)
## loop over periods
if periods is None: periods = [begindate-fileend]
# periods.sort(reverse=True) # reverse, so that largest chunk is done first
source = None # will later be assigned to the source dataset
for period in periods:
# figure out period
enddate = begindate + period
if filebegin > enddate: raise DateError, 'End date earlier than begin date.'
if enddate-1 > fileend: # if filebegin is 1979 and the simulation is 10 years, fileend will be 1988, not 1989!
# if end date is not available, skip period
endmsg = "\n{:s} --- Invalid Period for '{:s}': End Date {:4d} not in File! --- \n".format(pidstr,dataset_name,enddate)
endmsg += "{:s} --- ('{:s}')\n".format(pidstr,filepath)
logger.info(endmsg)
else: ## perform averaging for selected period
# determine if sink file already exists, and what to do about it
periodstr = '{0:4d}-{1:4d}'.format(begindate,enddate)
gridstr = '' if griddef is None or griddef.name is 'WRF' else '_'+griddef.name
filename = fileclass.climfile.format(domain,gridstr,'_'+periodstr)
if ldebug: filename = 'test_' + filename
if lparallel: tmppfx = 'tmp_wrfavg_{:s}_'.format(pidstr[1:-1])
else: tmppfx = 'tmp_wrfavg_'.format(pidstr[1:-1])
tmpfilename = tmppfx + filename
assert os.path.exists(expfolder)
filepath = expfolder+filename
tmpfilepath = expfolder+tmpfilename
lskip = False # else just go ahead
if os.path.exists(filepath):
if not loverwrite:
age = datetime.fromtimestamp(os.path.getmtime(filepath))
# if sink file is newer than source file, skip (do not recompute)
if age > sourceage and os.path.getsize(filepath) > 1e6: lskip = True
# N.B.: NetCDF files smaller than 1MB are usually incomplete header fragments from a previous crash
#print sourceage, age
if not lskip: os.remove(filepath)
# depending on last modification time of file or overwrite setting, start computation, or skip
if lskip:
# print message
skipmsg = "\n{:s} >>> Skipping: file '{:s}' in dataset '{:s}' already exists and is newer than source file.".format(pidstr,filename,dataset_name)
skipmsg += "\n{:s} >>> ('{:s}')\n".format(pidstr,filepath)
logger.info(skipmsg)
else:
## begin actual computation
beginmsg = "\n{:s} <<< Computing '{:s}' (d{:02d}) Climatology from {:s}".format(
pidstr,dataset_name,domain,periodstr)
if griddef is None: beginmsg += " >>> \n"
else: beginmsg += " ('{:s}' grid) >>> \n".format(griddef.name)
logger.info(beginmsg)
## actually load datasets
if source is None:
source = loadWRF_TS(experiment=experiment, filetypes=[filetype], domains=domain) # comes out as a tuple...
if not lparallel and ldebug: logger.info('\n'+str(source)+'\n')
# prepare sink
if os.path.exists(tmpfilepath): os.remove(tmpfilepath) # remove old temp files
sink = DatasetNetCDF(name='WRF Climatology', folder=expfolder, filelist=[tmpfilename], atts=source.atts.copy(), mode='w')
sink.atts.period = periodstr
# initialize processing
if griddef is None: lregrid = False
else: lregrid = True
CPU = CentralProcessingUnit(source, sink, varlist=varlist, tmp=lregrid, feedback=ldebug) # no need for lat/lon
# start processing climatology
if shift != 0:
logger.info('{0:s} (shifting climatology by {1:d} month, to start with January) \n'.format(pidstr,shift))
CPU.Climatology(period=period, offset=offset, shift=shift, flush=False)
# N.B.: immediate flushing should not be necessary for climatologies, since they are much smaller!
# reproject and resample (regrid) dataset
if lregrid:
CPU.Regrid(griddef=griddef, flush=True)
logger.info('%s --- '+str(griddef.geotansform)+' --- \n'%(pidstr))
# sync temporary storage with output dataset (sink)
CPU.sync(flush=True)
# add Geopotential Height Variance
if 'GHT_Var' in sink and 'Z_var' not in sink:
data_array = ( sink['GHT_Var'].data_array - sink['Z'].data_array**2 )**0.5
atts = dict(name='Z_var',units='m',long_name='Square Root of Geopotential Height Variance')
sink += Variable(axes=sink['Z'].axes, data=data_array, atts=atts)
# add (relative) Vorticity Variance
if 'Vorticity_Var' in sink and 'zeta_var' not in sink:
data_array = ( sink['Vorticity_Var'].data_array - sink['zeta'].data_array**2 )**0.5
atts = dict(name='zeta_var',units='1/s',long_name='Square Root of Relative Vorticity Variance')
sink += Variable(axes=sink['zeta'].axes, data=data_array, atts=atts)
# add names and length of months
sink.axisAnnotation('name_of_month', name_of_month, 'time',
atts=dict(name='name_of_month', units='', long_name='Name of the Month'))
if not sink.hasVariable('length_of_month'):
sink += Variable(name='length_of_month', units='days', axes=(sink.time,), data=days_per_month,
atts=dict(name='length_of_month',units='days',long_name='Length of Month'))
# close... and write results to file
sink.sync()
sink.close()
writemsg = "\n{:s} >>> Writing to file '{:s}' in dataset {:s}".format(pidstr,filename,dataset_name)
writemsg += "\n{:s} >>> ('{:s}')\n".format(pidstr,filepath)
logger.info(writemsg)
# rename file to proper name
if os.path.exists(filepath): os.remove(filepath) # remove old file
os.rename(tmpfilepath,filepath) # this will overwrite the old file
# print dataset
if not lparallel and ldebug:
logger.info('\n'+str(sink)+'\n')
# clean up (not sure if this is necessary, but there seems to be a memory leak...
del sink, CPU; gc.collect() # get rid of these guys immediately
# clean up and return
if source is not None: source.unload(); del source
# N.B.: source is only loaded once for all periods
# N.B.: garbage is collected in multi-processing wrapper as well
# return
return 0 # so far, there is no measure of success, hence, if there is no crash...
def performExtraction(dataset, mode, stnfct, dataargs, loverwrite=False, varlist=None, lwrite=True, lreturn=False,
ldebug=False, lparallel=False, pidstr='', logger=None):
''' worker function to extract point data from gridded dataset '''
# input checking
if not isinstance(dataset,basestring): raise TypeError
if not isinstance(dataargs,dict): raise TypeError # all dataset arguments are kwargs
if not callable(stnfct): raise TypeError # function to load station dataset
if lparallel:
if not lwrite: raise IOError, 'In parallel mode we can only write to disk (i.e. lwrite = True).'
if lreturn: raise IOError, 'Can not return datasets in parallel mode (i.e. lreturn = False).'
# logging
if logger is None: # make new logger
logger = logging.getLogger() # new logger
logger.addHandler(logging.StreamHandler())
else:
if isinstance(logger,basestring):
logger = logging.getLogger(name=logger) # connect to existing one
elif not isinstance(logger,logging.Logger):
raise TypeError, 'Expected logger ID/handle in logger KW; got {}'.format(str(logger))
lclim = False; lts = False
if mode == 'climatology': lclim = True
elif mode == 'time-series': lts = True
else: raise NotImplementedError, "Unrecognized Mode: '{:s}'".format(mode)
## extract meta data from arguments
module, dataargs, loadfct, filepath, datamsgstr = getMetaData(dataset, mode, dataargs)
dataset_name = dataargs.dataset_name; periodstr = dataargs.periodstr; avgfolder = dataargs.avgfolder
# load template dataset
stndata = stnfct() # load station dataset from function
if not isinstance(stndata, Dataset): raise TypeError
# N.B.: the loading function is necessary, because DataseNetCDF instances do not pickle well
# determine age of source file
if not loverwrite: sourceage = datetime.fromtimestamp(os.path.getmtime(filepath))
# get filename for target dataset and do some checks
filename = getTargetFile(stndata.name, dataset, mode, module, dataargs, lwrite)
if ldebug: filename = 'test_' + filename
if not os.path.exists(avgfolder): raise IOError, "Dataset folder '{:s}' does not exist!".format(avgfolder)
lskip = False # else just go ahead
if lwrite:
if lreturn:
tmpfilename = filename # no temporary file if dataset is passed on (can't rename the file while it is open!)
else:
if lparallel: tmppfx = 'tmp_exstns_{:s}_'.format(pidstr[1:-1])
else: tmppfx = 'tmp_exstns_'.format(pidstr[1:-1])
tmpfilename = tmppfx + filename
filepath = avgfolder + filename
tmpfilepath = avgfolder + tmpfilename
if os.path.exists(filepath):
if not loverwrite:
age = datetime.fromtimestamp(os.path.getmtime(filepath))
# if source file is newer than sink file or if sink file is a stub, recompute, otherwise skip
if age > sourceage and os.path.getsize(filepath) > 1e5: lskip = True
# N.B.: NetCDF files smaller than 100kB are usually incomplete header fragments from a previous crashed
if not lskip: os.remove(filepath) # recompute
# depending on last modification time of file or overwrite setting, start computation, or skip
if lskip:
# print message
skipmsg = "\n{:s} >>> Skipping: file '{:s}' in dataset '{:s}' already exists and is newer than source file.".format(pidstr,filename,dataset_name)
skipmsg += "\n{:s} >>> ('{:s}')\n".format(pidstr,filepath)
logger.info(skipmsg)
else:
## actually load datasets
source = loadfct() # load source
# check period
if 'period' in source.atts and dataargs.periodstr != source.atts.period: # a NetCDF attribute
raise DateError, "Specifed period is inconsistent with netcdf records: '{:s}' != '{:s}'".format(periodstr,source.atts.period)
# print message
if lclim: opmsgstr = "Extracting '{:s}'-type Point Data from Climatology ({:s})".format(stndata.name, periodstr)
elif lts: opmsgstr = "Extracting '{:s}'-type Point Data from Time-series".format(stndata.name)
else: raise NotImplementedError, "Unrecognized Mode: '{:s}'".format(mode)
# print feedback to logger
logger.info('\n{0:s} *** {1:^65s} *** \n{0:s} *** {2:^65s} *** \n'.format(pidstr,datamsgstr,opmsgstr))
if not lparallel and ldebug: logger.info('\n'+str(source)+'\n')
## create new sink/target file
# set attributes
atts=source.atts.copy()
atts['period'] = dataargs.periodstr if dataargs.periodstr else 'time-series'
atts['name'] = dataset_name; atts['station'] = stndata.name
atts['title'] = '{:s} (Stations) from {:s} {:s}'.format(stndata.title,dataset_name,mode.title())
# make new dataset
if lwrite: # write to NetCDF file
if os.path.exists(tmpfilepath): os.remove(tmpfilepath) # remove old temp files
sink = DatasetNetCDF(folder=avgfolder, filelist=[tmpfilename], atts=atts, mode='w')
else: sink = Dataset(atts=atts) # ony create dataset in memory
# initialize processing
CPU = CentralProcessingUnit(source, sink, varlist=varlist, tmp=False, feedback=ldebug)
# extract data at station locations
CPU.Extract(template=stndata, flush=True)
# get results
CPU.sync(flush=True)
# print dataset
if not lparallel and ldebug:
logger.info('\n'+str(sink)+'\n')
# write results to file
if lwrite:
sink.sync()
writemsg = "\n{:s} >>> Writing to file '{:s}' in dataset {:s}".format(pidstr,filename,dataset_name)
writemsg += "\n{:s} >>> ('{:s}')\n".format(pidstr,filepath)
logger.info(writemsg)
# rename file to proper name
if not lreturn:
sink.unload(); sink.close(); del sink # destroy all references
if os.path.exists(filepath): os.remove(filepath) # remove old file
os.rename(tmpfilepath,filepath)
# N.B.: there is no temporary file if the dataset is returned, because an open file can't be renamed
# clean up and return
source.unload(); del source#, CPU
if lreturn:
return sink # return dataset for further use (netcdf file still open!)
else:
return 0 # "exit code"
else: source = loadGPCC_TS(varlist=None,resolution=res)
source = source(time=timeSlice(period))
#source.load()
print(source)
print('\n')
# prepare sink
gridstr = res if grid == 'GPCC' else grid
filename = getFileName(grid=gridstr, period=period, name='GPCC', filepattern=avgfile)
if os.path.exists(avgfolder+filename): os.remove(avgfolder+filename)
atts =dict(period=periodstr, name='GPCC', title='GPCC Climatology')
sink = DatasetNetCDF(name='GPCC Climatology', folder=avgfolder, filelist=[filename], atts=source.atts, mode='w')
# sink = addGDALtoDataset(sink, griddef=source.griddef)
# initialize processing
CPU = CentralProcessingUnit(source, sink, tmp=True)
if period is not None:
# determine averaging interval
offset = source.time.getIndex(period[0]-1979)/12 # origin of monthly time-series is at January 1979
# start processing climatology
CPU.Climatology(period=period[1]-period[0], offset=offset, flush=False)
# CPU.sync(flush=True)
# get NARR coordinates
if grid is not 'GPCC':
griddef = loadPickledGridDef(grid=grid, res=None, folder=grid_folder)
#new_grid = import_module(grid[0:4]).__dict__[grid+'_grid']
# if grid == 'NARR':
# from datasets.NARR import NARR_grid
# reproject and resample (regrid) dataset
print('\n')
source = loadNARR_TS()
print(source)
print('\n')
# prepare sink
gridstr = '' if grid is 'NARR' else '_'+grid
filename = avgfile.format(gridstr,'_'+periodstr)
if os.path.exists(avgfolder+filename): os.remove(avgfolder+filename)
sink = DatasetNetCDF(name='NARR Climatology', folder=avgfolder, filelist=[filename], atts=source.atts, mode='w')
sink.atts.period = periodstr
# determine averaging interval
offset = source.time.getIndex(period[0]-1979)/12 # origin of monthly time-series is at January 1979
# initialize processing
# CPU = CentralProcessingUnit(source, sink, varlist=['precip', 'T2'], tmp=True) # no need for lat/lon
CPU = CentralProcessingUnit(source, sink, varlist=None, tmp=True) # no need for lat/lon
# start processing climatology
CPU.Climatology(period=period[1]-period[0], offset=offset, flush=False)
# sync temporary storage with output
CPU.sync(flush=True)
# # make new masks
# sink.mask(sink.landmask, maskSelf=False, varlist=['snow','snowh','zs'], invert=True, merge=False)
# add names and length of months
sink.axisAnnotation('name_of_month', name_of_month, 'time',
atts=dict(name='name_of_month', units='', long_name='Name of the Month'))
#print ' === month === '
# sink += VarNC(sink.dataset, name='length_of_month', units='days', axes=(sink.time,), data=days_per_month,
def performRegridding(dataset,
mode,
griddef,
dataargs,
loverwrite=False,
varlist=None,
lwrite=True,
lreturn=False,
ldebug=False,
lparallel=False,
pidstr='',
logger=None):
''' worker function to perform regridding for a given dataset and target grid '''
# input checking
if not isinstance(dataset, basestring): raise TypeError
if not isinstance(dataargs, dict):
raise TypeError # all dataset arguments are kwargs
if not isinstance(griddef, GridDefinition): raise TypeError
if lparallel:
if not lwrite:
raise IOError, 'Can only write to disk in parallel mode (i.e. lwrite = True).'
if lreturn:
raise IOError, 'Can not return datasets in parallel mode (i.e. lreturn = False).'
# logging
if logger is None: # make new logger
logger = logging.getLogger() # new logger
logger.addHandler(logging.StreamHandler())
else:
if isinstance(logger, basestring):
logger = logging.getLogger(name=logger) # connect to existing one
elif not isinstance(logger, logging.Logger):
raise TypeError, 'Expected logger ID/handle in logger KW; got {}'.format(
str(logger))
## extract meta data from arguments
dataargs, loadfct, srcage, datamsgstr = getMetaData(
dataset, mode, dataargs)
dataset_name = dataargs.dataset_name
periodstr = dataargs.periodstr
avgfolder = dataargs.avgfolder
# get filename for target dataset and do some checks
filename = getTargetFile(
dataset=dataset,
mode=mode,
dataargs=dataargs,
lwrite=lwrite,
grid=griddef.name.lower(),
)
# prepare target dataset
if ldebug: filename = 'test_' + filename
if not os.path.exists(avgfolder):
raise IOError, "Dataset folder '{:s}' does not exist!".format(
avgfolder)
lskip = False # else just go ahead
if lwrite:
if lreturn:
tmpfilename = filename # no temporary file if dataset is passed on (can't rename the file while it is open!)
else:
if lparallel: tmppfx = 'tmp_regrid_{:s}_'.format(pidstr[1:-1])
else: tmppfx = 'tmp_regrid_'.format(pidstr[1:-1])
tmpfilename = tmppfx + filename
filepath = avgfolder + filename
tmpfilepath = avgfolder + tmpfilename
if os.path.exists(filepath):
if not loverwrite:
age = datetime.fromtimestamp(os.path.getmtime(filepath))
# if source file is newer than sink file or if sink file is a stub, recompute, otherwise skip
if age > srcage and os.path.getsize(filepath) > 1e6:
lskip = True
if hasattr(griddef,
'filepath') and griddef.filepath is not None:
gridage = datetime.fromtimestamp(
os.path.getmtime(griddef.filepath))
if age < gridage: lskip = False
# N.B.: NetCDF files smaller than 1MB are usually incomplete header fragments from a previous crashed
# depending on last modification time of file or overwrite setting, start computation, or skip
if lskip:
# print message
skipmsg = "\n{:s} >>> Skipping: file '{:s}' in dataset '{:s}' already exists and is newer than source file.".format(
pidstr, filename, dataset_name)
skipmsg += "\n{:s} >>> ('{:s}')\n".format(pidstr, filepath)
logger.info(skipmsg)
else:
## actually load datasets
source = loadfct() # load source
# check period
if 'period' in source.atts and dataargs.periodstr != source.atts.period: # a NetCDF attribute
raise DateError, "Specifed period is inconsistent with netcdf records: '{:s}' != '{:s}'".format(
periodstr, source.atts.period)
# print message
if mode == 'climatology':
opmsgstr = 'Regridding Climatology ({:s}) to {:s} Grid'.format(
periodstr, griddef.name)
elif mode == 'time-series':
opmsgstr = 'Regridding Time-series to {:s} Grid'.format(
griddef.name)
else:
raise NotImplementedError, "Unrecognized Mode: '{:s}'".format(mode)
# print feedback to logger
logger.info(
'\n{0:s} *** {1:^65s} *** \n{0:s} *** {2:^65s} *** \n'
.format(pidstr, datamsgstr, opmsgstr))
if not lparallel and ldebug: logger.info('\n' + str(source) + '\n')
## create new sink/target file
# set attributes
atts = source.atts.copy()
atts['period'] = periodstr
atts['name'] = dataset_name
atts['grid'] = griddef.name
if mode == 'climatology':
atts['title'] = '{:s} Climatology on {:s} Grid'.format(
dataset_name, griddef.name)
elif mode == 'time-series':
atts['title'] = '{:s} Time-series on {:s} Grid'.format(
dataset_name, griddef.name)
# make new dataset
if lwrite: # write to NetCDF file
if os.path.exists(tmpfilepath):
os.remove(tmpfilepath) # remove old temp files
sink = DatasetNetCDF(folder=avgfolder,
filelist=[tmpfilename],
atts=atts,
mode='w')
else:
sink = Dataset(atts=atts) # ony create dataset in memory
# initialize processing
CPU = CentralProcessingUnit(source,
sink,
varlist=varlist,
tmp=False,
feedback=ldebug)
# perform regridding (if target grid is different from native grid!)
if griddef.name != dataset:
# reproject and resample (regrid) dataset
CPU.Regrid(griddef=griddef, flush=True)
# get results
CPU.sync(flush=True)
# add geolocators
sink = addGeoLocator(sink,
griddef=griddef,
lgdal=True,
lreplace=True,
lcheck=True)
# N.B.: WRF datasets come with their own geolocator arrays - we need to replace those!
# add length and names of month
if mode == 'climatology' and not sink.hasVariable(
'length_of_month') and sink.hasVariable('time'):
addLengthAndNamesOfMonth(
sink,
noleap=True if dataset.upper() in ('WRF', 'CESM') else False)
# print dataset
if not lparallel and ldebug:
logger.info('\n' + str(sink) + '\n')
# write results to file
if lwrite:
sink.sync()
writemsg = "\n{:s} >>> Writing to file '{:s}' in dataset {:s}".format(
pidstr, filename, dataset_name)
writemsg += "\n{:s} >>> ('{:s}')\n".format(pidstr, filepath)
logger.info(writemsg)
# rename file to proper name
if not lreturn:
sink.unload()
sink.close()
del sink # destroy all references
if os.path.exists(filepath):
os.remove(filepath) # remove old file
os.rename(
tmpfilepath,
filepath) # this would also overwrite the old file...
# N.B.: there is no temporary file if the dataset is returned, because an open file can't be renamed
# clean up and return
source.unload()
del source, CPU
if lreturn:
return sink # return dataset for further use (netcdf file still open!)
else:
return 0 # "exit code"
print('\n')
print(' *** Processing Resolution %s from %s *** '%(res,periodstr))
print('\n')
source = loadCFSR_TS(resolution=res)
print(source)
print('\n')
# prepare sink
filename = avgfile.format('_'+res,'_'+periodstr)
if os.path.exists(avgfolder+filename): os.remove(avgfolder+filename)
sink = DatasetNetCDF(name='CFSR Climatology', folder=avgfolder, filelist=[filename], atts=source.atts, mode='w')
sink.atts.period = periodstr
# determine averaging interval
offset = source.time.getIndex(period[0]-1979)/12 # origin of monthly time-series is at January 1979
# initialize processing
CPU = CentralProcessingUnit(source, sink, tmp=True)
# start processing climatology
CPU.Climatology(period=period[1]-period[0], offset=offset, flush=False)
# shift longitude axis by 180 degrees left (i.e. 0 - 360 -> -180 - 180)
CPU.Shift(lon=-180, flush=False)
# sync temporary storage with output (sink variable; do not flush!)
CPU.sync(flush=False)
# make new masks
if sink.hasVariable('landmask'):
sink.mask(sink.landmask, maskSelf=False, varlist=['snow','snowh','zs'], invert=True, merge=False)
# add names and length of months
def performExtraction(dataset,
mode,
stnfct,
dataargs,
loverwrite=False,
varlist=None,
lwrite=True,
lreturn=False,
ldebug=False,
lparallel=False,
pidstr='',
logger=None):
''' worker function to extract point data from gridded dataset '''
# input checking
if not isinstance(dataset, basestring): raise TypeError
if not isinstance(dataargs, dict):
raise TypeError # all dataset arguments are kwargs
if not callable(stnfct):
raise TypeError # function to load station dataset
if lparallel:
if not lwrite:
raise IOError, 'In parallel mode we can only write to disk (i.e. lwrite = True).'
if lreturn:
raise IOError, 'Can not return datasets in parallel mode (i.e. lreturn = False).'
# logging
if logger is None: # make new logger
logger = logging.getLogger() # new logger
logger.addHandler(logging.StreamHandler())
else:
if isinstance(logger, basestring):
logger = logging.getLogger(name=logger) # connect to existing one
elif not isinstance(logger, logging.Logger):
raise TypeError, 'Expected logger ID/handle in logger KW; got {}'.format(
str(logger))
lclim = False
lts = False
if mode == 'climatology': lclim = True
elif mode == 'time-series': lts = True
else: raise NotImplementedError, "Unrecognized Mode: '{:s}'".format(mode)
## extract meta data from arguments
dataargs, loadfct, srcage, datamsgstr = getMetaData(
dataset, mode, dataargs)
dataset_name = dataargs.dataset_name
periodstr = dataargs.periodstr
avgfolder = dataargs.avgfolder
# load template dataset
stndata = stnfct() # load station dataset from function
if not isinstance(stndata, Dataset): raise TypeError
# N.B.: the loading function is necessary, because DataseNetCDF instances do not pickle well
# get filename for target dataset and do some checks
filename = getTargetFile(dataset=dataset,
mode=mode,
dataargs=dataargs,
lwrite=lwrite,
station=stndata.name)
if ldebug: filename = 'test_' + filename
if not os.path.exists(avgfolder):
raise IOError, "Dataset folder '{:s}' does not exist!".format(
avgfolder)
lskip = False # else just go ahead
if lwrite:
if lreturn:
tmpfilename = filename # no temporary file if dataset is passed on (can't rename the file while it is open!)
else:
if lparallel: tmppfx = 'tmp_exstns_{:s}_'.format(pidstr[1:-1])
else: tmppfx = 'tmp_exstns_'.format(pidstr[1:-1])
tmpfilename = tmppfx + filename
filepath = avgfolder + filename
tmpfilepath = avgfolder + tmpfilename
if os.path.exists(filepath):
if not loverwrite:
age = datetime.fromtimestamp(os.path.getmtime(filepath))
# if source file is newer than sink file or if sink file is a stub, recompute, otherwise skip
if age > srcage and os.path.getsize(filepath) > 1e5:
lskip = True
# N.B.: NetCDF files smaller than 100kB are usually incomplete header fragments from a previous crashed
# depending on last modification time of file or overwrite setting, start computation, or skip
if lskip:
# print message
skipmsg = "\n{:s} >>> Skipping: file '{:s}' in dataset '{:s}' already exists and is newer than source file.".format(
pidstr, filename, dataset_name)
skipmsg += "\n{:s} >>> ('{:s}')\n".format(pidstr, filepath)
logger.info(skipmsg)
else:
## actually load datasets
source = loadfct() # load source
# check period
if 'period' in source.atts and dataargs.periodstr != source.atts.period: # a NetCDF attribute
raise DateError, "Specifed period is inconsistent with netcdf records: '{:s}' != '{:s}'".format(
periodstr, source.atts.period)
# print message
if lclim:
opmsgstr = "Extracting '{:s}'-type Point Data from Climatology ({:s})".format(
stndata.name, periodstr)
elif lts:
opmsgstr = "Extracting '{:s}'-type Point Data from Time-series".format(
stndata.name)
else:
raise NotImplementedError, "Unrecognized Mode: '{:s}'".format(mode)
# print feedback to logger
logger.info(
'\n{0:s} *** {1:^65s} *** \n{0:s} *** {2:^65s} *** \n'
.format(pidstr, datamsgstr, opmsgstr))
if not lparallel and ldebug: logger.info('\n' + str(source) + '\n')
## create new sink/target file
# set attributes
atts = source.atts.copy()
atts[
'period'] = dataargs.periodstr if dataargs.periodstr else 'time-series'
atts['name'] = dataset_name
atts['station'] = stndata.name
atts['title'] = '{:s} (Stations) from {:s} {:s}'.format(
stndata.title, dataset_name, mode.title())
# make new dataset
if lwrite: # write to NetCDF file
if os.path.exists(tmpfilepath):
os.remove(tmpfilepath) # remove old temp files
sink = DatasetNetCDF(folder=avgfolder,
filelist=[tmpfilename],
atts=atts,
mode='w')
else:
sink = Dataset(atts=atts) # ony create dataset in memory
# initialize processing
CPU = CentralProcessingUnit(source,
sink,
varlist=varlist,
tmp=False,
feedback=ldebug)
# extract data at station locations
CPU.Extract(template=stndata, flush=True)
# get results
CPU.sync(flush=True)
# print dataset
if not lparallel and ldebug:
logger.info('\n' + str(sink) + '\n')
# write results to file
if lwrite:
sink.sync()
writemsg = "\n{:s} >>> Writing to file '{:s}' in dataset {:s}".format(
pidstr, filename, dataset_name)
writemsg += "\n{:s} >>> ('{:s}')\n".format(pidstr, filepath)
logger.info(writemsg)
# rename file to proper name
if not lreturn:
sink.unload()
sink.close()
del sink # destroy all references
if os.path.exists(filepath):
os.remove(filepath) # remove old file
os.rename(tmpfilepath, filepath)
# N.B.: there is no temporary file if the dataset is returned, because an open file can't be renamed
# clean up and return
source.unload()
del source #, CPU
if lreturn:
return sink # return dataset for further use (netcdf file still open!)
else:
return 0 # "exit code"
def performRegridding(dataset, mode, griddef, dataargs, loverwrite=False, varlist=None, lwrite=True,
lreturn=False, ldebug=False, lparallel=False, pidstr='', logger=None):
''' worker function to perform regridding for a given dataset and target grid '''
# input checking
if not isinstance(dataset,basestring): raise TypeError
if not isinstance(dataargs,dict): raise TypeError # all dataset arguments are kwargs
if not isinstance(griddef,GridDefinition): raise TypeError
if lparallel:
if not lwrite: raise IOError, 'Can only write to disk in parallel mode (i.e. lwrite = True).'
if lreturn: raise IOError, 'Can not return datasets in parallel mode (i.e. lreturn = False).'
# logging
if logger is None: # make new logger
logger = logging.getLogger() # new logger
logger.addHandler(logging.StreamHandler())
else:
if isinstance(logger,basestring):
logger = logging.getLogger(name=logger) # connect to existing one
elif not isinstance(logger,logging.Logger):
raise TypeError, 'Expected logger ID/handle in logger KW; got {}'.format(str(logger))
## extract meta data from arguments
dataargs, loadfct, srcage, datamsgstr = getMetaData(dataset, mode, dataargs)
dataset_name = dataargs.dataset_name; periodstr = dataargs.periodstr; avgfolder = dataargs.avgfolder
# get filename for target dataset and do some checks
filename = getTargetFile(dataset=dataset, mode=mode, dataargs=dataargs, lwrite=lwrite,
grid=griddef.name.lower(), period=None, filetype=None)
# prepare target dataset
if ldebug: filename = 'test_' + filename
if not os.path.exists(avgfolder): raise IOError, "Dataset folder '{:s}' does not exist!".format(avgfolder)
lskip = False # else just go ahead
if lwrite:
if lreturn: tmpfilename = filename # no temporary file if dataset is passed on (can't rename the file while it is open!)
else:
if lparallel: tmppfx = 'tmp_regrid_{:s}_'.format(pidstr[1:-1])
else: tmppfx = 'tmp_regrid_'.format(pidstr[1:-1])
tmpfilename = tmppfx + filename
filepath = avgfolder + filename
tmpfilepath = avgfolder + tmpfilename
if os.path.exists(filepath):
if not loverwrite:
age = datetime.fromtimestamp(os.path.getmtime(filepath))
# if source file is newer than sink file or if sink file is a stub, recompute, otherwise skip
if age > srcage and os.path.getsize(filepath) > 1e6:
lskip = True
if hasattr(griddef, 'filepath') and griddef.filepath is not None:
gridage = datetime.fromtimestamp(os.path.getmtime(griddef.filepath))
if age < gridage: lskip = False
# N.B.: NetCDF files smaller than 1MB are usually incomplete header fragments from a previous crashed
if not lskip: os.remove(filepath) # recompute
# depending on last modification time of file or overwrite setting, start computation, or skip
if lskip:
# print message
skipmsg = "\n{:s} >>> Skipping: file '{:s}' in dataset '{:s}' already exists and is newer than source file.".format(pidstr,filename,dataset_name)
skipmsg += "\n{:s} >>> ('{:s}')\n".format(pidstr,filepath)
logger.info(skipmsg)
else:
## actually load datasets
source = loadfct() # load source
# check period
if 'period' in source.atts and dataargs.periodstr != source.atts.period: # a NetCDF attribute
raise DateError, "Specifed period is inconsistent with netcdf records: '{:s}' != '{:s}'".format(periodstr,source.atts.period)
# print message
if mode == 'climatology': opmsgstr = 'Regridding Climatology ({:s}) to {:s} Grid'.format(periodstr, griddef.name)
elif mode == 'time-series': opmsgstr = 'Regridding Time-series to {:s} Grid'.format(griddef.name)
else: raise NotImplementedError, "Unrecognized Mode: '{:s}'".format(mode)
# print feedback to logger
logger.info('\n{0:s} *** {1:^65s} *** \n{0:s} *** {2:^65s} *** \n'.format(pidstr,datamsgstr,opmsgstr))
if not lparallel and ldebug: logger.info('\n'+str(source)+'\n')
## create new sink/target file
# set attributes
atts=source.atts.copy()
atts['period'] = periodstr; atts['name'] = dataset_name; atts['grid'] = griddef.name
if mode == 'climatology': atts['title'] = '{:s} Climatology on {:s} Grid'.format(dataset_name, griddef.name)
elif mode == 'time-series': atts['title'] = '{:s} Time-series on {:s} Grid'.format(dataset_name, griddef.name)
# make new dataset
if lwrite: # write to NetCDF file
if os.path.exists(tmpfilepath): os.remove(tmpfilepath) # remove old temp files
sink = DatasetNetCDF(folder=avgfolder, filelist=[tmpfilename], atts=atts, mode='w')
else: sink = Dataset(atts=atts) # ony create dataset in memory
# initialize processing
CPU = CentralProcessingUnit(source, sink, varlist=varlist, tmp=False, feedback=ldebug)
# perform regridding (if target grid is different from native grid!)
if griddef.name != dataset:
# reproject and resample (regrid) dataset
CPU.Regrid(griddef=griddef, flush=True)
# get results
CPU.sync(flush=True)
# add geolocators
sink = addGeoLocator(sink, griddef=griddef, lgdal=True, lreplace=True, lcheck=True)
# N.B.: WRF datasets come with their own geolocator arrays - we need to replace those!
# add length and names of month
if mode == 'climatology' and not sink.hasVariable('length_of_month') and sink.hasVariable('time'):
addLengthAndNamesOfMonth(sink, noleap=True if dataset.upper() in ('WRF','CESM') else False)
# print dataset
if not lparallel and ldebug:
logger.info('\n'+str(sink)+'\n')
# write results to file
if lwrite:
sink.sync()
writemsg = "\n{:s} >>> Writing to file '{:s}' in dataset {:s}".format(pidstr,filename,dataset_name)
writemsg += "\n{:s} >>> ('{:s}')\n".format(pidstr,filepath)
logger.info(writemsg)
# rename file to proper name
if not lreturn:
sink.unload(); sink.close(); del sink # destroy all references
if os.path.exists(filepath): os.remove(filepath) # remove old file
os.rename(tmpfilepath,filepath) # this would also overwrite the old file...
# N.B.: there is no temporary file if the dataset is returned, because an open file can't be renamed
# clean up and return
source.unload(); del source, CPU
if lreturn:
return sink # return dataset for further use (netcdf file still open!)
else:
return 0 # "exit code"
def computeClimatology(experiment,
filetype,
domain,
periods=None,
offset=0,
griddef=None,
varlist=None,
ldebug=False,
loverwrite=False,
lparallel=False,
pidstr='',
logger=None):
''' worker function to compute climatologies for given file parameters. '''
# input type checks
if not isinstance(experiment, Exp): raise TypeError
if not isinstance(filetype, basestring): raise TypeError
if not isinstance(domain, (np.integer, int)): raise TypeError
if periods is not None and not (isinstance(periods, (tuple, list))
and isInt(periods)):
raise TypeError
if not isinstance(offset, (np.integer, int)): raise TypeError
if not isinstance(loverwrite, (bool, np.bool)): raise TypeError
if griddef is not None and not isinstance(griddef, GridDefinition):
raise TypeError
#if pidstr == '[proc01]': raise TypeError # to test error handling
# load source
dataset_name = experiment.name
fileclass = fileclasses[filetype] # used for target file name
tsfile = fileclass.tsfile.format(domain, '')
expfolder = experiment.avgfolder
filepath = '{:s}/{:s}'.format(expfolder, tsfile)
logger.info('\n\n{0:s} *** Processing Experiment {1:<15s} *** '.
format(pidstr, "'{:s}'".format(dataset_name)) +
'\n{0:s} *** {1:^37s} *** \n'.format(
pidstr, "'{:s}'".format(tsfile)))
# check file and read begin/enddates
if not os.path.exists(filepath):
#raise IOError, "Source file '{:s}' does not exist!".format(filepath)
# print message and skip
skipmsg = "\n{:s} >>> File '{:s}' in dataset '{:s}' is missing --- skipping!".format(
pidstr, tsfile, dataset_name)
skipmsg += "\n{:s} >>> ('{:s}')\n".format(pidstr, filepath)
logger.warning(skipmsg)
# N.B.: this can cause a lot of error messages, when not all files are present
else: # if monthly source file exists
import netCDF4 as nc
ncfile = nc.Dataset(filepath, mode='r')
begintuple = ncfile.begin_date.split('-')
endtuple = ncfile.end_date.split('-')
ncfile.close()
# N.B.: at this point we don't want to initialize a full GDAL-enabled dataset, since we don't even
# know if we need it, and it creates a lot of overhead
# determine age of source file
if not loverwrite:
sourceage = datetime.fromtimestamp(os.path.getmtime(filepath))
# figure out start date
filebegin = int(begintuple[0]) # first element is the year
fileend = int(endtuple[0]) # first element is the year
begindate = offset + filebegin
if not (filebegin <= begindate <= fileend): raise DateError
# handle cases where the first month in the record is not January
firstmonth = int(begintuple[1]) # second element is the month
shift = firstmonth - 1 # will be zero for January (01)
## loop over periods
if periods is None: periods = [begindate - fileend]
# periods.sort(reverse=True) # reverse, so that largest chunk is done first
source = None # will later be assigned to the source dataset
for period in periods:
# figure out period
enddate = begindate + period
if filebegin > enddate:
raise DateError, 'End date earlier than begin date.'
if enddate - 1 > fileend: # if filebegin is 1979 and the simulation is 10 years, fileend will be 1988, not 1989!
# if end date is not available, skip period
endmsg = "\n{:s} --- Invalid Period for '{:s}': End Date {:4d} not in File! --- \n".format(
pidstr, dataset_name, enddate)
endmsg += "{:s} --- ('{:s}')\n".format(pidstr, filepath)
logger.info(endmsg)
else: ## perform averaging for selected period
# determine if sink file already exists, and what to do about it
periodstr = '{0:4d}-{1:4d}'.format(begindate, enddate)
gridstr = '' if griddef is None or griddef.name is 'WRF' else '_' + griddef.name
filename = fileclass.climfile.format(domain, gridstr,
'_' + periodstr)
if ldebug: filename = 'test_' + filename
if lparallel: tmppfx = 'tmp_wrfavg_{:s}_'.format(pidstr[1:-1])
else: tmppfx = 'tmp_wrfavg_'.format(pidstr[1:-1])
tmpfilename = tmppfx + filename
assert os.path.exists(expfolder)
filepath = expfolder + filename
tmpfilepath = expfolder + tmpfilename
lskip = False # else just go ahead
if os.path.exists(filepath):
if not loverwrite:
age = datetime.fromtimestamp(
os.path.getmtime(filepath))
# if sink file is newer than source file, skip (do not recompute)
if age > sourceage and os.path.getsize(filepath) > 1e6:
lskip = True
# N.B.: NetCDF files smaller than 1MB are usually incomplete header fragments from a previous crash
#print sourceage, age
if not lskip: os.remove(filepath)
# depending on last modification time of file or overwrite setting, start computation, or skip
if lskip:
# print message
skipmsg = "\n{:s} >>> Skipping: file '{:s}' in dataset '{:s}' already exists and is newer than source file.".format(
pidstr, filename, dataset_name)
skipmsg += "\n{:s} >>> ('{:s}')\n".format(
pidstr, filepath)
logger.info(skipmsg)
else:
if griddef is None: lregrid = False
else: lregrid = True
## begin actual computation
beginmsg = "\n{:s} <<< Computing '{:s}' (d{:02d}) Climatology from {:s}".format(
pidstr, dataset_name, domain, periodstr)
if not lregrid: beginmsg += " >>> \n"
else:
beginmsg += " ('{:s}' grid) >>> \n".format(
griddef.name)
logger.info(beginmsg)
## actually load datasets
if source is None:
source = loadWRF_TS(
experiment=experiment,
filetypes=[filetype],
domains=domain) # comes out as a tuple...
if not lparallel and ldebug:
logger.info('\n' + str(source) + '\n')
# prepare sink
if os.path.exists(tmpfilepath):
os.remove(tmpfilepath) # remove old temp files
sink = DatasetNetCDF(name='WRF Climatology',
folder=expfolder,
filelist=[tmpfilename],
atts=source.atts.copy(),
mode='w')
sink.atts.period = periodstr
# if lregrid: addGDALtoDataset(sink, griddef=griddef)
# initialize processing
CPU = CentralProcessingUnit(
source,
sink,
varlist=varlist,
tmp=lregrid,
feedback=ldebug) # no need for lat/lon
# start processing climatology
if shift != 0:
logger.info(
'{0:s} (shifting climatology by {1:d} month, to start with January) \n'
.format(pidstr, shift))
CPU.Climatology(period=period,
offset=offset,
shift=shift,
flush=False)
# N.B.: immediate flushing should not be necessary for climatologies, since they are much smaller!
# reproject and resample (regrid) dataset
if lregrid:
CPU.Regrid(griddef=griddef, flush=True)
logger.info('{:s} --- {:s} --- \n'.format(
pidstr, griddef.name))
logger.debug('{:s} --- {:s} --- \n'.format(
pidstr, str(griddef)))
# sync temporary storage with output dataset (sink)
CPU.sync(flush=True)
# add Geopotential Height Variance
if 'GHT_Var' in sink and 'Z_var' not in sink:
data_array = (sink['GHT_Var'].data_array -
sink['Z'].data_array**2)**0.5
atts = dict(
name='Z_var',
units='m',
long_name=
'Square Root of Geopotential Height Variance')
sink += Variable(axes=sink['Z'].axes,
data=data_array,
atts=atts)
# add (relative) Vorticity Variance
if 'Vorticity_Var' in sink and 'zeta_var' not in sink:
data_array = (sink['Vorticity_Var'].data_array -
sink['zeta'].data_array**2)**0.5
atts = dict(
name='zeta_var',
units='1/s',
long_name=
'Square Root of Relative Vorticity Variance')
sink += Variable(axes=sink['zeta'].axes,
data=data_array,
atts=atts)
# add names and length of months
sink.axisAnnotation('name_of_month',
name_of_month,
'time',
atts=dict(
name='name_of_month',
units='',
long_name='Name of the Month'))
if not sink.hasVariable('length_of_month'):
sink += Variable(name='length_of_month',
units='days',
axes=(sink.time, ),
data=days_per_month,
atts=dict(
name='length_of_month',
units='days',
long_name='Length of Month'))
# close... and write results to file
sink.sync()
sink.close()
writemsg = "\n{:s} >>> Writing to file '{:s}' in dataset {:s}".format(
pidstr, filename, dataset_name)
writemsg += "\n{:s} >>> ('{:s}')\n".format(
pidstr, filepath)
logger.info(writemsg)
# rename file to proper name
if os.path.exists(filepath):
os.remove(filepath) # remove old file
os.rename(tmpfilepath,
filepath) # this will overwrite the old file
# print dataset
if not lparallel and ldebug:
logger.info('\n' + str(sink) + '\n')
# clean up (not sure if this is necessary, but there seems to be a memory leak...
del sink, CPU
gc.collect() # get rid of these guys immediately
# clean up and return
if source is not None:
source.unload()
del source
# N.B.: source is only loaded once for all periods
# N.B.: garbage is collected in multi-processing wrapper as well
# return
return 0 # so far, there is no measure of success, hence, if there is no crash...
filename = avgfile.format('_' + res, '_' + periodstr)
if os.path.exists(avgfolder + filename):
os.remove(avgfolder + filename)
sink = DatasetNetCDF(name='CFSR Climatology',
folder=avgfolder,
filelist=[filename],
atts=source.atts,
mode='w')
sink.atts.period = periodstr
# determine averaging interval
offset = source.time.getIndex(
period[0] -
1979) / 12 # origin of monthly time-series is at January 1979
# initialize processing
CPU = CentralProcessingUnit(source, sink, tmp=True)
# start processing climatology
CPU.Climatology(period=period[1] - period[0],
offset=offset,
flush=False)
# shift longitude axis by 180 degrees left (i.e. 0 - 360 -> -180 - 180)
CPU.Shift(lon=-180, flush=False)
# sync temporary storage with output
CPU.sync(flush=True)
# make new masks
if sink.hasVariable('landmask'):
sink.mask(sink.landmask,
name='GPCC',
filepattern=avgfile)
if os.path.exists(avgfolder + filename):
os.remove(avgfolder + filename)
atts = dict(period=periodstr,
name='GPCC',
title='GPCC Climatology')
sink = DatasetNetCDF(name='GPCC Climatology',
folder=avgfolder,
filelist=[filename],
atts=source.atts,
mode='w')
# sink = addGDALtoDataset(sink, griddef=source.griddef)
# initialize processing
CPU = CentralProcessingUnit(source, sink, tmp=True)
if period is not None:
# determine averaging interval
offset = source.time.getIndex(
period[0] - 1979
) / 12 # origin of monthly time-series is at January 1979
# start processing climatology
CPU.Climatology(period=period[1] - period[0],
offset=offset,
flush=False)
# CPU.sync(flush=True)
# get NARR coordinates
if grid is not 'GPCC':
griddef = loadPickledGridDef(grid=grid,
if os.path.exists(avgfolder + filename):
os.remove(avgfolder + filename)
sink = DatasetNetCDF(name='CRU Climatology',
folder=avgfolder,
filelist=[filename],
atts=source.atts,
mode='w')
sink.atts.period = periodstr
# determine averaging interval
offset = source.time.getIndex(
period[0] -
1979) / 12 # origin of monthly time-series is at January 1979
# initialize processing
# CPU = CentralProcessingUnit(source, sink, varlist=['wetfrq'])
CPU = CentralProcessingUnit(source, sink)
# start processing
print('')
print(' +++ processing +++ ')
CPU.Climatology(period=period[1] - period[0],
offset=offset,
flush=False)
# sync temporary storage with output
CPU.sync(flush=False)
print('\n')
# add landmask
print ' === landmask === '
tmpatts = dict(
name='landmask',
units='',
