def find_hist_model(equations):
''' takes a unrolled model and create a model which can be run for historic periode \n
The model calculates residuals for equations with a <res= > clause and \n
and the identities are also calculeted'''
hist = []
for f in find_frml(equations):
hist.append(find_res(f))
a, fr, n, udtryk = split_frml(f.upper())
if kw_frml_name(n, 'IDENT') or kw_frml_name(n, 'I'):
# identites are just replicated in order to calculate backdata
hist.append(f)
return (' '.join(hist))
def exounroll(in_equations):
''' takes a model and makes a new model by enhancing frml's with <exo=,j=,jr=>
in their frml name.
:exo: the value can be fixed in to a value valuename_x by setting valuename_d=1
:jled: a additiv adjustment element is added to the frml
:jrled: a multiplicativ adjustment element is added to the frml
'''
nymodel = []
equations = in_equations[:] # we want do change the e
for comment, command, value in find_statements(equations.upper()):
# print('>>',comment,'<',command,'>',value)
if comment:
nymodel.append(comment)
elif command == 'FRML':
a, fr, n, udtryk = split_frml((command + ' ' + value).upper())
#print(fr,n, kw_frml_name(n.upper(),'EXO'))
# we want a relatov adjustment
if kw_frml_name(n, 'JRLED') or kw_frml_name(n, 'JR'):
lhs, rhs = udtryk.split('=', 1)
udtryk = lhs + \
'= (' + rhs[:-1] + ')*(1+' + lhs.strip() + '_JR )' + '$'
if kw_frml_name(n, 'JLED') or kw_frml_name(n, 'J'):
lhs, rhs = udtryk.split('=', 1)
# we want a absolute adjustment
udtryk = lhs + '=' + rhs[:-1] + '+' + lhs.strip() + '_J' + '$'
if kw_frml_name(n, 'EXO'):
lhs, rhs = udtryk.split('=', 1)
endogen = lhs.strip()
dummy = endogen + '_D'
exogen = endogen + '_X'
udtryk = lhs + \
'=(' + rhs[:-1] + ')*(1-' + dummy + ')+' + exogen + '*' + dummy + '$'
nymodel.append(command + ' ' + n + ' ' + udtryk)
else:
nymodel.append(command + ' ' + value)
equations = '\n'.join(nymodel)
return equations
def __init__(self,
mmodel,
df=None,
endovar=None,
onlyendocur=False,
timeit=False,
silent=True,
forcenum=False,
per='',
ljit=0,
nchunk=None,
endoandexo=False):
self.df = df if type(df) == pd.DataFrame else mmodel.lastdf
self.endovar = sorted(mmodel.endogene if endovar == None else endovar)
self.endoandexo = endoandexo
self.mmodel = mmodel
self.onlyendocur = onlyendocur
self.silent = silent
self.maxdif = 9999999999999
self.forcenum = forcenum
self.timeit = timeit
self.per = per
self.ljit = ljit
self.nchunk = nchunk
print(f'Prepare model til calculate derivatives for Newton solver')
self.declared_endo_list0 = [
pt.kw_frml_name(self.mmodel.allvar[v]['frmlname'], 'ENDO', v)
for v in self.endovar
]
self.declared_endo_list = [
v[:-6] if v.endswith('___RES') else v
for v in self.declared_endo_list0
] # real endogeneous variables
self.declared_endo_set = set(self.declared_endo_list)
assert len(self.declared_endo_list) == len(self.declared_endo_set)
self.placdic = {v: i for i, v in enumerate(self.endovar)}
if self.endoandexo:
self.exovar = [
v for v in sorted(mmodel.exogene)
if not v in self.declared_endo_set
]
self.exoplacdic = {v: i for i, v in enumerate(self.exovar)}
else:
self.exoplacdic = {}
# breakpoint()
self.diffendocur = self.modeldiff()
self.diff_model = self.get_diffmodel()
def un_normalize(frml):
'''This function makes sure that all formulas are unnormalized.
if the formula is already decorated with <endo=name> this is keept
else the lhs_varriable is used in <endo=>
'''
frml_name, frml_index, frml_rest = findindex(frml.upper())
this_endo = pt.kw_frml_name(frml_name.upper(), 'ENDO')
lhs, rhs = frml_rest.split('=')
if this_endo:
lhs_var = this_endo.strip()
frml_name_out = frml_name
else:
lhs_var = lhs.strip()
frml_name_out = f'<endo={lhs_var}>' if frml_name == '<>' else f'{frml_name[:-1]},endo={lhs_var}>'
print(this_endo)
new_rest = f'{lhs_var}___res = ( {rhs.strip()} ) - ( {lhs.strip()} )'
return f'{frml_name_out} {frml_index if len(frml_index) else ""} {new_rest}'
def find_res(f):
''' Finds the expression which calculates the residual in a formel. FRML <res=a,endo=b> x=a*b+c $ '''
from sympy import solve, sympify
a, fr, n, udtryk = split_frml(f.upper())
udres = ''
tres = kw_frml_name(n, 'RES')
if tres:
lhs, rhs = udtryk.split('=', 1)
res = lhs.strip() + '_J' if tres == 'J' else lhs.strip() + \
'_JR' if tres == 'JR' else tres
# we take the the $ out
kat = sympify('Eq(' + lhs + ',' + rhs[0:-1] + ')')
res_frml = sympify('res_frml')
res_frml = solve(kat, res)
udres = 'FRML ' + 'RES' + ' ' + \
res.ljust(25) + ' = ' + str(res_frml)[1:-1] + ' $'
return udres
def __init__(self,
mmodel,
df=None,
endovar=None,
onlyendocur=False,
timeit=False,
silent=True,
forcenum=False,
per='',
ljit=0,
nchunk=None,
endoandexo=False):
"""
Args:
mmodel (TYPE): Model to analyze.
df (TYPE, optional): Dataframe. if None mmodel.lastdf will be used
endovar (TYPE, optional): if set defines which endogeneous to include . Defaults to None.
onlyendocur (TYPE, optional): Only calculate for the curren endogeneous variables. Defaults to False.
timeit (TYPE, optional): writeout time informations . Defaults to False.
silent (TYPE, optional): Defaults to True.
forcenum (TYPE, optional): Force differentiation to be numeric else try sumbolic (slower) Defaults to False.
per (TYPE, optional): Period for which to calculte the jacobi . Defaults to ''.
ljit (TYPE, optional): Trigger just in time compilation of the differential coiefficient. Defaults to 0.
nchunk (TYPE, optional): Chunks for which the model is written - relevant if ljit == True. Defaults to None.
endoandexo (TYPE, optional): Calculate for both endogeneous and exogeneous . Defaults to False.
Returns:
None.
"""
self.df = df if type(df) == pd.DataFrame else mmodel.lastdf
self.endovar = sorted(mmodel.endogene if endovar == None else endovar)
self.endoandexo = endoandexo
self.mmodel = mmodel
self.onlyendocur = onlyendocur
self.silent = silent
self.maxdif = 9999999999999
self.forcenum = forcenum
self.timeit = timeit
self.per = per
self.ljit = ljit
self.nchunk = nchunk
if not self.silent:
print(f'Prepare model for calculate derivatives for Newton solver')
self.declared_endo_list0 = [
pt.kw_frml_name(self.mmodel.allvar[v]['frmlname'], 'ENDO', v)
for v in self.endovar
]
self.declared_endo_list = [
v[:-6] if v.endswith('___RES') else v
for v in self.declared_endo_list0
] # real endogeneous variables
self.declared_endo_set = set(self.declared_endo_list)
assert len(self.declared_endo_list) == len(self.declared_endo_set)
self.placdic = {v: i for i, v in enumerate(self.endovar)}
self.newmodel = mmodel.__class__
if self.endoandexo:
self.exovar = [
v for v in sorted(mmodel.exogene)
if not v in self.declared_endo_set
]
self.exoplacdic = {v: i for i, v in enumerate(self.exovar)}
else:
self.exoplacdic = {}
# breakpoint()
self.diffendocur = self.modeldiff()
self.diff_model = self.get_diffmodel()
'where': ['land', 'sea']
}, 'we have brought you your {weed} from {where} ')
sub_frml({
'weed': ['scrubbery', 'herring'],
'where': ['land', 'sea']
}, 'we have brought you your {weed} from {where} ')
a = {'bankdic': {'bank': ['Danske', 'Nordea'], 'danske': ['yes', 'no']}}
sub_frml(a['bankdic'], 'Dette er {bank}')
sub_frml(a['bankdic'], 'Dette er {bank}', sep=' and ')
sub_frml(a['bankdic'], 'Dette er {bank}', plus='+', sep='')
sub_frml(a['bankdic'], 'Dette er {bank}', xvar='danske', lig='yes')
sub_frml(a['bankdic'], 'Dette er {bank}', xvar='danske', lig='no')
sub_frml(a['bankdic'], 'Dette er {bank}')
list_extract(
'list bankdic = bank : Danske , Nordea / danske : yes , no $')
kw_frml_name('<res=abe>', 'res')
kw_frml_name('<res=abe,animal>', 'animal')
find_res('FRML <res=x> ib =x+y+v $')
find_res('FRML <res=J> ib =x+y+v + ib_j $')
find_res('FRML <res=JR> ib =(x+y+v + ib_j)*(1+ ib_JR) $)')
find_hist_model('FRML <res=x> ib =x+y+v $ frml <ident> y=c+i+x-m $')
(exounroll('frml <j> ib=x+y $ frml <jr> ib=x+y $ frml <j,jr'))
find_arg('log', 'x=log(a+b)+77')
find_arg('log', 'x=log(a+log(b))+77')
#%%
x = (dounloop('''list bankdic = bank : Danske , Nordea /
ko : yes , no $
do bankdic ko = no $
frml x {bank}_income = {bank}_a +{bank}_b $
enddo $
frml x ialt=sum(bankdic,{bank}_income $'''))