def gatherv_str_arr_impl(data):
rank = hpat.distributed_api.get_rank()
n_loc = len(data)
n_all_chars = num_total_chars(data)
# allocate send lens arrays
send_arr_lens = np.empty(n_loc, np.uint32) # XXX offset type is uint32
send_data_ptr = get_data_ptr(data)
for i in range(n_loc):
_str = data[i]
send_arr_lens[i] = len(_str)
recv_counts = gather_scalar(np.int32(n_loc))
recv_counts_char = gather_scalar(np.int32(n_all_chars))
n_total = recv_counts.sum()
n_total_char = recv_counts_char.sum()
# displacements
all_data = StringArray(['']) # dummy arrays on non-root PEs
displs = np.empty(0, np.int32)
displs_char = np.empty(0, np.int32)
if rank == MPI_ROOT:
all_data = pre_alloc_string_array(n_total, n_total_char)
displs = hpat.hiframes.join.calc_disp(recv_counts)
displs_char = hpat.hiframes.join.calc_disp(recv_counts_char)
offset_ptr = get_offset_ptr(all_data)
data_ptr = get_data_ptr(all_data)
c_gatherv(
send_arr_lens.ctypes,
np.int32(n_loc),
offset_ptr,
recv_counts.ctypes,
displs.ctypes,
int32_typ_enum)
c_gatherv(
send_data_ptr,
np.int32(n_all_chars),
data_ptr,
recv_counts_char.ctypes,
displs_char.ctypes,
char_typ_enum)
convert_len_arr_to_offset(offset_ptr, n_total)
return all_data
def test_impl():
A = StringArray(['ABC', 'BB', 'ADEF'])
df = pd.DataFrame({'A': A})
B = df.A.str.contains('BB', regex=False)
return B.sum()
def fix_df_array_impl(column):
return StringArray(column)
def test_impl():
A = StringArray(['ABC', 'BB', 'CDEF'])
B = A == 'ABC'
C = A[B]
return len(C) == 1 and C[0] == 'ABC'
def test_impl():
A = StringArray(['ABC', 'BB', 'CDEF'])
B = A == 'ABC'
return B.sum()
def test_impl():
return StringArray(['ABC', 'BB', 'CDEF'])
def fix_df_array_impl(column): # pragma: no cover
return StringArray(column)