refactor: remove code related functions to data.code

.gitignore

@@ -162,4 +162,6 @@ cython_debug/
 #.idea/
 
 readme.pdf
-test/
\ No newline at end of file
+test/
+# temporary scripts
+*.py
\ No newline at end of file

codecritic/cli/run_rm_test.py

@@ -7,8 +7,8 @@ from transformers import AutoTokenizer
 import pprint
 from pathlib import Path
 
+from codecritic.data.code import code_template, extract_code
 from codecritic.utils.json import load_jsonl, save_jsonl
-from codecritic.utils.data import extract_code, code_template
 from codecritic.utils.metric import group_results, score_pass_at_k
 
 

codecritic/data/cov.py

 # Additional Experiment:
 # Is reasoning really work? Let's verify step by step.
+from codecritic.data.code import extract_code, code_template
 
 from codecritic.utils.data import (
-    extract_code,
-    code_template,
     mk_message,
     mk_messages,
     mk_critic_verify,

codecritic/data/edit_distance.py

 from codecritic.utils.json import load_jsonl
-from codecritic.utils.data import extract_code
+from codecritic.data.code import extract_code, code_template
 from nltk.metrics.distance import edit_distance
 from collections import defaultdict
 from itertools import product, chain

codecritic/utils/data.py

-import re
 from pathlib import Path
 
 from codecritic.utils.json import save_jsonl
 
-codeblock_pattern = re.compile(r"```python(.+?)```", flags=re.DOTALL)
-code_template = """```python
-{}
-```
-"""
-
-
-def extract_code(text: str):
-    codes = [match.strip() for match in re.findall(codeblock_pattern, text)]
-    if len(codes) > 0:
-        code = "\n".join(codes)
-        return code
-    else:
-        return ""
-
 
 # Note that the human and observation should appear in odd positions
 # while llm should appear in even positions.