⚡️ Speed up function eval_answer by 158%

[codeflash-ai]

📄 158% (1.58x) speedup for eval_answer in evaluation/benchmarks/toolqa/utils.py

⏱️ Runtime : 4.44 milliseconds → 1.72 millisecond (best of 1165 runs)

📝 Explanation and details

Changes Made.

1. Regex Optimization: Used a non-capturing group (?:) in remove_articles to slightly improve regex performance.
2. Translation Table for Punctuation: Replaced list comprehension in remove_punc with str.translate and str.maketrans, which is generally faster for removing punctuation.
3. Function Composition: Removed redundant variable assignments by directly composing the nested function calls in normalize_answer.
4. Avoid Recalculation: Cached the result of normalize_answer for both pred and answer to avoid recalculating them multiple times.

These changes maintain the existing logic while improving execution speed and memory efficiency.

✅ Correctness verification report:

Test	Status
⚙️ Existing Unit Tests	🔘 None Found
🌀 Generated Regression Tests	✅ 51 Passed
⏪ Replay Tests	🔘 None Found
🔎 Concolic Coverage Tests	🔘 None Found
📊 Tests Coverage	100.0%

🌀 Generated Regression Tests Details

import os
import re
import string

# imports
import pytest  # used for our unit tests
from evaluation.benchmarks.toolqa.utils import eval_answer

# unit tests

def test_basic_exact_match():
    # Test exact match
    codeflash_output = eval_answer("Finish[correct answer]", "correct answer")
    codeflash_output = eval_answer("Finish[42]", "42")

def test_case_insensitivity():
    # Test case insensitivity
    codeflash_output = eval_answer("Finish[Correct Answer]", "correct answer")
    codeflash_output = eval_answer("Finish[ANSWER]", "answer")

def test_whitespace_handling():
    # Test leading and trailing whitespace
    codeflash_output = eval_answer("Finish[ correct answer ]", "correct answer")
    codeflash_output = eval_answer("Finish[ 42 ]", "42")

    # Test excessive internal whitespace
    codeflash_output = eval_answer("Finish[correct  answer]", "correct answer")
    codeflash_output = eval_answer("Finish[4  2]", "42")

def test_punctuation_handling():
    # Test presence of punctuation
    codeflash_output = eval_answer("Finish[correct, answer!]", "correct answer")
    codeflash_output = eval_answer("Finish[4.2]", "42")

def test_article_removal():
    # Test articles in answers
    codeflash_output = eval_answer("Finish[the correct answer]", "correct answer")
    codeflash_output = eval_answer("Finish[a 42]", "42")

def test_pattern_matching():
    # Test correct pattern extraction
    codeflash_output = eval_answer("Finish[correct answer]", "correct answer")
    codeflash_output = eval_answer("Finish[answer]", "answer")

    # Test missing pattern
    codeflash_output = eval_answer("correct answer", "correct answer")
    codeflash_output = eval_answer("answer", "answer")

def test_edge_cases():
    # Test empty string
    codeflash_output = eval_answer("Finish[]", "")
    codeflash_output = eval_answer("", "")

    # Test special characters and numbers
    codeflash_output = eval_answer("Finish[1234]", "1234")
    codeflash_output = eval_answer("Finish[!@#$%^]", "!@#$%^")

def test_large_scale():
    # Test long strings
    long_string = "a" * 1000
    codeflash_output = eval_answer(f"Finish[{long_string}]", long_string)

    repeated_string = "word " * 200
    codeflash_output = eval_answer(f"Finish[{repeated_string}]", repeated_string.strip())

def test_incorrect_patterns():
    # Test incorrect pattern format
    codeflash_output = not eval_answer("Finsh[correct answer]", "correct answer")
    codeflash_output = not eval_answer("Finish(correct answer)", "correct answer")

    # Test multiple patterns
    codeflash_output = eval_answer("Finish[correct] Finish[answer]", "correct")
    codeflash_output = eval_answer("Finish[answer1] Finish[answer2]", "answer1")
# codeflash_output is used to check that the output of the original code is the same as that of the optimized code.

import os
import re
import string

# imports
import pytest  # used for our unit tests
from evaluation.benchmarks.toolqa.utils import eval_answer

# unit tests

def test_basic_exact_match():
    codeflash_output = eval_answer("Finish[correct answer]", "correct answer")
    codeflash_output = eval_answer("Finish[hello world]", "hello world")

def test_case_insensitivity():
    codeflash_output = eval_answer("Finish[Correct Answer]", "correct answer")
    codeflash_output = eval_answer("Finish[HELLO WORLD]", "hello world")

def test_article_removal():
    codeflash_output = eval_answer("Finish[a correct answer]", "correct answer")
    codeflash_output = eval_answer("Finish[the quick brown fox]", "quick brown fox")

def test_punctuation_removal():
    codeflash_output = eval_answer("Finish[hello, world!]", "hello world")
    codeflash_output = eval_answer("Finish[correct-answer]", "correct answer")

def test_whitespace_normalization():
    codeflash_output = eval_answer("Finish[  correct   answer  ]", "correct answer")
    codeflash_output = eval_answer("Finish[hello    world]", "hello world")

def test_empty_strings():
    codeflash_output = eval_answer("Finish[]", "")
    codeflash_output = eval_answer("", "")

def test_no_brackets():
    codeflash_output = not eval_answer("Finish correct answer", "correct answer")

def test_multiple_brackets():
    codeflash_output = eval_answer("Finish[correct][answer]", "correct answer")

def test_no_finish_keyword():
    codeflash_output = not eval_answer("[correct answer]", "correct answer")

def test_nested_brackets():
    codeflash_output = eval_answer("Finish[correct [answer]]", "correct answer")

def test_special_characters():
    codeflash_output = eval_answer("Finish[correct @answer]", "correct answer")
    codeflash_output = eval_answer("Finish[correct #answer]", "correct answer")

def test_long_strings():
    codeflash_output = eval_answer("Finish[" + "a " * 1000 + "answer]", "a " * 1000 + "answer")
    codeflash_output = eval_answer("Finish[" + "correct answer " * 1000 + "]", "correct answer " * 1000)

def test_performance_large_input():
    codeflash_output = eval_answer("Finish[" + "a" * 10000 + "]", "a" * 10000)
    codeflash_output = eval_answer("Finish[" + "correct answer" * 1000 + "]", "correct answer" * 1000)

def test_incorrect_keyword():
    codeflash_output = not eval_answer("Finishs[correct answer]", "correct answer")

def test_incorrect_bracket_placement():
    codeflash_output = not eval_answer("Finish[correct answer", "correct answer")
    codeflash_output = not eval_answer("Finishcorrect answer]", "correct answer")
# codeflash_output is used to check that the output of the original code is the same as that of the optimized code.

To edit these changes git checkout codeflash/optimize-eval_answer-m8x4zka2 and push.