def get_data(path, listfile, seq_length, word_to_id):
all_data, def_positions, identifier_usage = read_data(
path, listfile, word_to_id)
identifier_types = astwalker.possible_types()
num_masks = len(identifier_types)
def_positions = [[[t[1] for t in fp if t[0] == k]
for k in identifier_types] for fp in def_positions]
# def_positions = [[t[1] for t in file_positions] for file_positions in def_positions]
num_masks = len(identifier_types)
#num_masks = 1
data = []
# Prevent indent and dedent tokens from being flagged as variables, which can occur because dedent
# in particular takes up no columns
non_vars = [word_to_id[indent_token], word_to_id[dedent_token]]
for j in range(len(all_data)):
filedata = list(all_data[j])
file_positions = def_positions[j]
identifier_positions = identifier_usage[j]
num_sequences = math.ceil(len(filedata) / seq_length)
input_data = np.zeros([num_sequences, seq_length]).astype("int64")
targets = np.zeros([num_sequences, seq_length]).astype("int64")
masks_data = np.zeros([num_sequences, num_masks,
seq_length]).astype("bool")
id_usage_data = np.zeros([num_sequences, seq_length]).astype("bool")
actual_lengths = []
for i in range(num_sequences):
x = [t[0] for t in filedata[i * seq_length:(i + 1) * seq_length]]
y = [
t[0]
for t in filedata[i * seq_length + 1:(i + 1) * seq_length + 1]
]
masks = [[
t[1] in fp and t[0] not in non_vars for fp in file_positions
] for t in filedata[i * seq_length:(i + 1) * seq_length]]
# masks = [t[1] in file_positions and t[0] not in non_vars
# for t in filedata[i * seq_length:(i + 1) * seq_length]]
ids = [
t[1] in identifier_positions
for t in filedata[i * seq_length + 1:(i + 1) * seq_length + 1]
]
actual_length_x = len(x)
actual_length_y = len(y)
input_data[i, :actual_length_x] = x
targets[i, :actual_length_y] = y
masks_data[i, :, :actual_length_x] = np.transpose(masks)
id_usage_data[i, :actual_length_y] = ids
actual_lengths.append(actual_length_y)
container = Container(input_data, targets, actual_lengths, masks_data,
id_usage_data)
data.append(container)
return data
def get_data(path, listfile, seq_length, word_to_id):
all_data, def_positions, identifier_usage = read_data(path, listfile, word_to_id)
identifier_types = astwalker.possible_types()
num_masks = len(identifier_types)
def_positions = [[[t[1] for t in fp if t[0] == k] for k in identifier_types] for fp in def_positions]
# def_positions = [[t[1] for t in file_positions] for file_positions in def_positions]
num_masks = len(identifier_types)
#num_masks = 1
data = []
# Prevent indent and dedent tokens from being flagged as variables, which can occur because dedent
# in particular takes up no columns
non_vars = [word_to_id[indent_token], word_to_id[dedent_token]]
for j in range(len(all_data)):
filedata = list(all_data[j])
file_positions = def_positions[j]
identifier_positions = identifier_usage[j]
num_sequences = math.ceil(len(filedata) / seq_length)
input_data = np.zeros([num_sequences, seq_length]).astype("int64")
targets = np.zeros([num_sequences, seq_length]).astype("int64")
masks_data = np.zeros([num_sequences, num_masks, seq_length]).astype("bool")
id_usage_data = np.zeros([num_sequences, seq_length]).astype("bool")
actual_lengths = []
for i in range(num_sequences):
x = [t[0] for t in filedata[i * seq_length:(i + 1) * seq_length]]
y = [t[0] for t in filedata[i * seq_length + 1:(i + 1) * seq_length + 1]]
masks = [[t[1] in fp and t[0] not in non_vars for fp in file_positions]
for t in filedata[i * seq_length:(i + 1) * seq_length]]
# masks = [t[1] in file_positions and t[0] not in non_vars
# for t in filedata[i * seq_length:(i + 1) * seq_length]]
ids = [t[1] in identifier_positions for t in filedata[i * seq_length + 1:(i + 1) * seq_length + 1]]
actual_length_x = len(x)
actual_length_y = len(y)
input_data[i, :actual_length_x] = x
targets[i, :actual_length_y] = y
masks_data[i, :, :actual_length_x] = np.transpose(masks)
id_usage_data[i, :actual_length_y] = ids
actual_lengths.append(actual_length_y)
container = Container(input_data, targets, actual_lengths, masks_data, id_usage_data)
data.append(container)
return data
def map_token(map, token):
mask = 0
if token.startswith("(*) "):
mask = 1
token = token.replace("(*) ", "")
if token in map:
return map[token], mask
# Not in map, is it an identifier?
if "|" in token:
spl = token.split("|")
if spl[1] in map:
return map[spl[1]]
elif spl[0] in map:
return map[spl[0]]
elif any([token.startswith(prefix) for prefix in astwalker.possible_types()]):
return pyreader.oov_id
raise KeyError(token)
def map_token(map, token):
mask = 0
if token.startswith("(*) "):
mask = 1
token = token.replace("(*) ", "")
if token in map:
return map[token], mask
# Not in map, is it an identifier?
if "|" in token:
spl = token.split("|")
if spl[1] in map:
return (map[spl[1]],)
elif spl[0] in map:
return (map[spl[0]],)
elif any([token.startswith(prefix) for prefix in astwalker.possible_types()]):
return (pyreader.oov_id,)
raise KeyError(token)
def is_identifier(token_id):
token = self.inv_map[token_id]
if any(token.startswith(p) for p in astwalker.possible_types()):
return 1
return 0
def is_identifier(token_id):
token = self.inv_map[token_id]
if any(token.startswith(p) for p in astwalker.possible_types()):
return 1
return 0
def adjust_flags():
if FLAGS.attention:
FLAGS.attention = FLAGS.attention.split("+")
if "identifiers" in FLAGS.attention:
FLAGS.attention.extend(["identifiers"] * (len(astwalker.possible_types()) - 1))
def adjust_flags():
if FLAGS.attention:
FLAGS.attention = FLAGS.attention.split("+")
if "identifiers" in FLAGS.attention:
FLAGS.attention.extend(["identifiers"] *
(len(astwalker.possible_types()) - 1))
