p4dataset / construct_dataset.py
Mihai-Valentin DUMITRU
initial commit
065fa03
Raw History Blame Contribute Delete
5.39 kB
#!/usr/bin/env python3
import argparse
import json
import logging
import multiprocessing as mp
import os
import traceback
import chardet
from tqdm import tqdm
import helpers
from helpers import getLogger, setup_logging
def read_file_safely(filepath):
"""Reads a file safely, attempting multiple encodings if necessary.
Args:
filepath (str): Path to the file.
Returns:
str: Decoded file content.
"""
try:
with open(filepath, "rb") as fin:
raw_data = fin.read()
detected = chardet.detect(raw_data)
encoding = detected.get("encoding", "utf-8")
try:
return raw_data.decode(encoding)
except UnicodeDecodeError:
logging.warning("Failed to decode %s with detected encoding %s", filepath, encoding)
return ""
except Exception as e:
logging.warning("Error reading file: %s", filepath)
logging.warning("Error details: %s", "".join(traceback.format_exception(e)))
return None
def get_all_files(directory):
"""Get all files in a directory.
Args:
directory (str): The directory to search.
Returns:
list: A list of file paths.
"""
files = []
for root, _, filenames in os.walk(directory):
for filename in filenames:
files.append(os.path.join(root, filename))
return files
def process_directory(directory):
"""Process files in a directory and generate JSON entries.
Args:
directory (str): The root directory to process.
Returns:
list: A list of JSON strings representing processed files.
"""
id_ = 0
lines = []
logging.info("Starting to process directory: %s", directory)
tasks = get_all_files(directory)
for root, _, files in os.walk(directory):
for filename in files:
filepath = os.path.join(root, filename)
try:
user, repo, path = filepath.split(os.sep, 2)[2].split(os.sep, 2)
repo_name = os.path.join(user, repo)
path = "/" + path # match the format used in the Stack v2
# with open(filepath, "rt") as fin:
# content = fin.read()
content = read_file_safely(filepath)
if content is None:
continue
entry = {
"repo_name": repo_name,
"repo_path": path,
"id": id_,
"content": content
}
lines.append(json.dumps(entry) + "\n")
id_ += 1
except Exception as e:
logging.warning("Failed to process file: %s", filepath)
estr = "".join(traceback.format_exception(e))
logging.warning("Error details: %s", estr)
logging.info("Finished processing directory: %s", directory)
return lines
def process_file(filepath):
setup_logging()
logger = getLogger()
try:
user, repo, path = filepath.split(os.sep, 2)[2].split(os.sep, 2)
repo_name = os.path.join(user, repo)
path = "/" + path # match the format used in the Stack v2
content = read_file_safely(filepath)
if content is None:
logger.warning("Failed to read file: %s", filepath)
return None
entry = {
"repo_name": repo_name,
"repo_path": path,
"content": content
}
return entry
except Exception as e:
logger.warning("Failed to process file: %s", filepath)
estr = "".join(traceback.format_exception(e))
logger.warning("Error details: %s", estr)
def construct_dataset(srcdir, dspath):
"""Construct a dataset from files in a directory.
Args:
srcdir (str): The directory to process.
dspath (str): The path to the output dataset file.
"""
logging.info("Starting dataset construction from directory: %s", srcdir)
tasks = get_all_files(srcdir)
id_ = 0
results = []
with mp.Pool(processes=mp.cpu_count()) as pool:
with tqdm(total=len(tasks), desc="Processing files",
unit="file") as pbar:
for result in pool.imap_unordered(process_file, tasks):
if result is None:
continue
result["id"] = id_
id_ += 1
results.append(result)
pbar.update()
lines = [json.dumps(entry) + "\n" for entry in results]
with open(dspath, "wt") as fout:
fout.writelines(lines)
logging.info("Dataset successfully written to: %s", dspath)
def main():
parser = argparse.ArgumentParser(description="Process files in a directory to construct a dataset.")
parser.add_argument("srcdir", type=str, help="Source directory to process.")
parser.add_argument("dspath", type=str, help="Path to the output dataset file.")
parser.add_argument("--num_processes", type=int, default=mp.cpu_count(),
help="Number of processes to use (default: number of CPU cores)")
parser.add_argument("--logdir", type=str, default=None,
help="Name of the log directory (default: %(default)s)")
args = parser.parse_args()
construct_dataset(args.srcdir, args.dspath)
if __name__ == "__main__":
main()