def release_live_import(live_import):
"""
Cleanup and exit live import module.
"""
if live_import is not None:
live_import.ReleaseLiveImport()
FreeLibrary(live_import._handle)
def test_1703286_A(self):
from _ctypes import LoadLibrary, FreeLibrary
# On winXP 64-bit, advapi32 loads at an address that does
# NOT fit into a 32-bit integer. FreeLibrary must be able
# to accept this address.
# These are tests for http://www.python.org/sf/1703286
handle = LoadLibrary("advapi32")
FreeLibrary(handle)
def load_shared_lib(shared_lib_path: str) -> ctypes.CDLL:
"""
Load a shared library using ctypes freeing the resource at end.
"""
plugin_dll = ctypes.cdll.LoadLibrary(shared_lib_path)
try:
yield plugin_dll
finally:
if sys.platform == 'win32':
from _ctypes import FreeLibrary
FreeLibrary(plugin_dll._handle)
else:
from _ctypes import dlclose
dlclose(plugin_dll._handle)
def UnloadCamDLL(self):
del self.dll
self.dll = None
FreeLibrary(self.libHandle)
def test_1703286_A(self):
from _ctypes import LoadLibrary, FreeLibrary
handle = LoadLibrary('advapi32')
FreeLibrary(handle)
def __del__(self):
FreeLibrary(self.dll._handle)
def deleteLib(daLib):
from _ctypes import FreeLibrary
hlib = daLib._handle
del daLib
FreeLibrary(hlib)
def rungap50(country, data, adjfact, vintage_name, changey, path, tipo,
logfile):
import ctypes as ct
import numpy as np
import pandas as pd
from _ctypes import FreeLibrary
import f90nml
projpath = path + '/'
path = path + '$' # add to find end of dtring
dllversion = "GAP50DLL20190503.dll"
with open(logfile, 'a') as f:
f.write('\n---computing ' + tipo.upper() + ' for ' + country.upper() +
' via GAP50.DLL v.' + dllversion)
lib = ct.CDLL(projpath + "lib/" + dllversion)
gap50 = getattr(lib, "pytogap")
# Read Namelist
nml = f90nml.read(projpath + 'priors/' + tipo.upper() + '_DLL_' +
country.upper() + '_' +
vintage_name.replace('final', '') + '.nml')
nmlstr = str(nml)
nt = len(nml['prior']['lab']) # number of parameters
ny = len(nml['ssm']['endogenous']) # number of endogenous series
inter = nml['GAP'][
'Inter'] # INTER: 1=(Back,PC),2=(Back,Cycle),3=(Forw,PC),4=(Forw,Cycle)
hor = max(0, nml['GAP']['Anchor'][1]) # horizon for anchored estimates
starty = nml['GAP']['Startyear']
stri = str(nml['ssm']['exogenous'])
pos = list(find_all(stri, ','))
if len(stri) > 0:
nz = len(pos) + 1 # number of exogenous series
else:
nz = 0
name_vars = list()
if tipo == 'nawru':
name_vars.append('LUR')
nf = hor
else:
name_vars.append('SR')
nf = 10
if tipo == 'nawru':
if inter < 3:
name_vars.append('DWINF')
else:
name_vars.append('DRULC')
# TODO : DEBUG
# else:
# name_vars.append('CU')
elif ny == 2:
name_vars.append('CU')
if nz == 1:
name_vars.append(stri.strip())
if nz > 1:
name_vars.append(stri[0:pos[0]].strip())
for i in range(0, len(pos) - 1):
name_vars.append(stri[pos[i] + 1:pos[i + 1]].strip())
name_vars.append(stri[pos[len(pos) - 1] + 1:].strip())
nmax = changey - starty + 1
# Select variables
(inds, indn, xn) = selectdata(data, changey, name_vars[0:ny], nmax)
ismax1 = np.amax(inds) # maximum missing position
ismax2 = 0
if nz > 1: #number of exog series
(indse, indne, xne) = selectdata(data, changey, name_vars[ny + 1:],
nmax)
ismax2 = np.amax(indse) # maximum missing position
ismax = max(ismax1, ismax2)
nobs = nmax - ismax # number of in-sample observations
yk = np.arange(0, ny + nz, dtype=float) * 0 # setup the nobs+nf - long
for i in range(0, nobs + max(nf, hor) -
1): # concatenate columns until it is nrow x ncol
yk = np.c_[yk, np.arange(0, ny + nz, dtype=float) * 0]
yk[0:ny, 0:nobs] = xn[0:ny, ismax:nobs + ismax]
yk[0:ny, nobs:] = -99999.0
if nz > 0:
yk[ny, :] = 1.0 # set the constant series
if nz > 1:
yk[ny + 1:, 0:nobs] = xne[0:, ismax:nobs + ismax]
for j in range(nobs, nobs + max(nf, hor)):
yk[ny + 1:, j] = 0
# yk[ny+1:,j] = xne[0:,nobs+ismax-1]
# Set other dll Input
b_string0 = nmlstr.encode('utf-8')
b_string1 = path.encode('utf-8')
arr = (ct.c_char * 5000 * 2)()
arr[0].value = b_string0
arr[1].value = b_string1
nstring = ct.pointer(ct.c_int(2))
nobsp = ct.pointer(ct.c_int(nobs)) # setup the pointer
nyp = ct.pointer(ct.c_int(ny))
nzp = ct.pointer(ct.c_int(nz))
nfp = ct.pointer(ct.c_int(nf))
horp = ct.pointer(ct.c_int(hor))
ntp = ct.pointer(ct.c_int(nt))
# Set dll Output
# Both ML and Bayes
unobs = np.arange(0, 26, dtype=float) * 0 # setup the nobs+nf - long
for i in range(
1, nobs +
max(nf, hor)): # concatenate columns until it is nrow x ncol
unobs = np.c_[unobs, np.arange(0, 26, dtype=float) * 0]
# Bayes
ac = np.arange(0, 4 * ny, dtype=float) * 0 # setup the nobs+nf - long
for i in range(1, 5): # concatenate columns until it is nrow x ncol
ac = np.c_[ac, np.arange(0, 4 * ny, dtype=float) * 0]
param = np.arange(0, nt, dtype=float) * 0 # setup the nobs+nf - long
for i in range(1, 409): # concatenate columns until it is nrow x ncol
param = np.c_[param, np.arange(0, nt, dtype=float) * 0]
marginal = np.arange(0, nobs + nf,
dtype=float) * 0 # setup the nobs+nf - long
for i in range(1, 1200): # concatenate columns until it is nrow x ncol
marginal = np.c_[marginal, np.arange(0, nobs + nf, dtype=float) * 0]
margl = np.arange(0, 13, dtype=float) * 0 # setup the nobs+nf - long
for i in range(1, 2): # concatenate columns until it is nrow x ncol
margl = np.c_[margl, np.arange(0, 13, dtype=float) * 0]
with open(logfile, 'a') as f:
f.write('\n-----Variable list : ' + str(name_vars))
gap50(nstring, arr, nobsp, nyp, nzp, nfp, horp, ntp,
np.ctypeslib.as_ctypes(yk), np.ctypeslib.as_ctypes(unobs),
np.ctypeslib.as_ctypes(ac), np.ctypeslib.as_ctypes(param),
np.ctypeslib.as_ctypes(marginal), np.ctypeslib.as_ctypes(margl))
handle = lib._handle
FreeLibrary(handle)
# --------------------------------------------------------
# UNOBSERVABLES (nob+nf,26) Maximum Likelihood estimates
# 1 series + forecasts, 2 rmse, 3 akrmse,
# 4 smoothed trend + forecast, 5 rmse, 6 akrmse
# 7 smoothed cycle + forecast, 8 rmse, 9 akrmse
# 10 filtered trend, 11 filtered cycle, 12 rmse
# 13 1st series innovations
# 14 PC series + forecast
# 15 rmse
# 16 akrmse
# 17 PC innovations
# 18 PC idiosync smoothed - empty!
# 19 PC idiosync filtered - empty!
# 20 PC Smoothed component - mu + beta'*c(t|T)
# 21 PC filtered component - mu + beta'*c(t|t)"
# 22 PC fitted values - mu + beta'*c(t|T)+gam'*z
# 23 Anchored trend
# 24 Anchored cycle
# ---------------------------------------------------------
if tipo == 'nawru':
if nml['GAP']['Anchor'][1] < 0:
nawru_series = pd.Series(unobs[3, 0:nmax - ismax + nf] +
float(adjfact[country.lower()]),
index=range(changey + 1 - nmax + ismax,
changey + nf +
1)).rename('NAWRU')
else:
nawru_series = pd.Series(unobs[22, 0:nmax - ismax + nf] +
float(adjfact[country.lower()]),
index=range(changey + 1 - nmax + ismax,
changey + nf +
1)).rename('NAWRU')
with open(logfile, 'a') as f:
f.write('->PASSED')
return nawru_series
else:
# ------------------------------------------------------------------------
# UNOBSERVABLES, Bayesian estimates
# by cols: ser1, p2.5, p5, p95, p97.5, trend, p2.5, p5, p95, p97.5,
# cycle, p2.5, p5, p95, p97.5, ser2, p2.5, p5, p95, p97.5, inn
# slope, p2.5, p5, p95, p97.5
# ------------------------------------------------------------------------
with open(logfile, 'a') as f:
f.write('->PASSED')
return pd.Series(unobs[5, 0:nmax + nf],
index=range(changey + 1 - nmax + ismax,
changey + ismax + nf + 1)).rename('SRKF')