def parse_withs_withouts(s1, s2):
"""Convert inputs into unique words and check validity
"""
global withs, withouts
s1, s2 = set(s1.split()), set(s2.split())
if len(s1 & s2) > 0:
SFIToolkit.errprintln(
'The same string cannot be passed to both options with() and without().'
)
SFIToolkit.exit(7103)
withs, withouts = list(s1), list(s2)
def postall(parser, nostore, key_b, key_V, key_y, key_cnames, key_N, key_dofr):
if len(ests) == 0:
SFIToolkit.errprintln('Specify using or run usehdf first.')
SFIToolkit.exit(7103)
for g in ests.keys():
post(g,
parser,
nostore,
key_b,
key_V,
key_y,
key_cnames,
key_N,
key_dofr,
check=False)
def post(gname,
parser,
nostore,
key_b,
key_V,
key_y,
key_cnames,
key_N,
key_dofr,
check=True):
if check:
if len(ests) == 0:
SFIToolkit.errprintln('Specify using or run usehdf first.')
SFIToolkit.exit(7103)
# Select est based on gname
if isinstance(gname, int):
if not 0 <= gname < len(ests):
SFIToolkit.errprintln(
'Value passed to option i() is out of range.')
SFIToolkit.exit(7103)
est = list(ests.values())[gname]
gname = list(ests.keys())[gname]
elif gname in ests:
est = ests[gname]
else:
SFIToolkit.errprintln('Specified group name is not in memory.')
SFIToolkit.exit(7103)
# Check whether gname contains any string in withs
if withs != []:
has_with = False
for s in withs:
if s in gname:
has_with = True
break
if not has_with:
return
# Check whether gname contains any string in withouts
if withouts != []:
has_without = False
for s in withouts:
if s in gname:
has_without = True
break
if has_without:
return
est = ests[gname]
cnames = None
if key_cnames in est:
cnames = list(est[key_cnames])
if callable(parser):
cnames = parser(cnames)
elif isinstance(parser, int) and 0 <= parser < len(custom_parsers):
cnames = custom_parsers[parser](cnames)
elif parser is not None:
SFIToolkit.errprintln('Invalid specification of parser.')
SFIToolkit.exit(7103)
# Generate options for ereturn post
y = ''
if key_y in est:
y = ' dep(' + sn(decode(get_scalar(est[key_y]))) + ')'
obs = ''
if key_N in est:
obs = ' obs(' + str(int(get_scalar(est[key_N]))) + ')'
dof = ''
if key_dofr in est:
dof = ' dof(' + str(int(get_scalar(est[key_dofr]))) + ')'
if key_b in est:
bname = ' ' + coef(est[key_b], cnames)
if key_V in est:
vname = ' ' + vcov(est[key_V], cnames)
st('ereturn post' + bname + vname + ',' + y + obs + dof)
else:
st('ereturn post' + bname + ',' + y + obs + dof)
# Classify and format the remaining objects and then post in e-class
for k, v in est.items():
if not k in (key_b, key_V, key_cnames, key_y, key_N, key_dofr):
generic(k, v)
# est store only works if e(cmd) exists
if 'cmd' not in est:
st('ereturn local cmd posthdf')
if nostore == "":
st('est store ' + gname)
def load(path, rootname, root, groups, append):
global ests
if append == '':
ests = OrderedDict()
load.parents = []
p = Path(path)
gs = ['/'] if root == 'root' else groups.split()
empty_groups = [] # Record empty groups
typeunknown = [] # Record datasets with unknown dtype
# Store all data in ests
with h5py.File(p, 'r') as h:
# Collect all unique parents by default
if len(gs) == 0:
h.visititems(get_parents)
gs = [g if g == '/' else g[1:] for g in set(load.parents)]
if root == 'noroot':
gs = [g for g in gs if g != '/']
if len(gs) == 0:
SFIToolkit.errprintln('No group is found.')
SFIToolkit.exit(7103)
for g in gs:
# Stata name does not allow '/'
e = rootname if g == '/' else g
# Ensure group name is not too long for est store
if len(e) > 27:
SFIToolkit.errprintln(
'Warning: The following group name is too long for Stata:')
SFIToolkit.errprintln(' ' + e)
SFIToolkit.errprintln(
' Try truncating the name to 27 characters.')
e = e[:27]
# Replace any invalid character with '_'
e = sn(e)
if g in h:
d = {}
for k, v in h[g].items():
if isinstance(v, h5py.Dataset):
try:
d[k] = v[()]
except TypeError:
typeunknown.append(g + '/' + k)
if len(d):
if e in ests:
print("Warning: Existing data of group '" + e +
"' are overwritten.")
ests[e] = d
else:
empty_groups.append(g)
else:
SFIToolkit.errprintln(
'A non-existing group name is passed to option groups().')
SFIToolkit.exit(7103)
if len(empty_groups):
print('Groups do not contain any dataset are omitted:')
print(' ' + ' '.join(empty_groups))
if len(typeunknown):
SFIToolkit.errprintln(
'Warning: The following datasets have datatypes not recognized by NumPy and are omitted:'
)
for tu in typeunknown:
SFIToolkit.errprintln(' ' + tu)
if len(ests) == 0:
SFIToolkit.errprintln('No data found.')
SFIToolkit.exit(7103)
elif len(ests) > 300:
SFIToolkit.displayln(
'{text:Warning: The number of groups in memory has passed 300, which is the maximum number of estimation results allowed by Stata.}'
)
SFIToolkit.displayln('{text: Results have to be posted in batches.}')
n_est = str(len(ests))
est_keys = ' '.join(ests.keys())
f = str(p)
st('return scalar n_hdfgroup=' + n_est)
st('return local hdfgroup_names ' + est_keys)
st('return local hdf_file ' + f)