generate-balanced-sample.ts

import fs from 'fs';

// Configuration
const SAMPLE_SIZE = 50;
const INPUT_FILE = 'bugs.json';
const OUTPUT_FILE = 'comments-easy-bugs.json';
const DIFFICULTY_LEVEL = 'easy'; // Only include easy bugs

interface Bug {
    repo: string;
    LLMReviewPassed: boolean;
    reReviewedAfterJan27th: boolean;
    difficultyLevel?: string; // Add difficulty level property
    [key: string]: any; // for other properties we don't care about in this script
}

// Read and parse the bugs file
const bugs = JSON.parse(fs.readFileSync(INPUT_FILE, 'utf-8')) as Bug[];

// Filter for bugs that passed LLM review, were re-reviewed after Jan 27th, and have "easy" difficulty
const eligibleBugs = bugs.filter(bug => 
    bug.LLMReviewPassed === true && 
    bug.reReviewedAfterJan27th === true &&
    bug.difficultyLevel === DIFFICULTY_LEVEL
);

console.log(`Found ${eligibleBugs.length} eligible ${DIFFICULTY_LEVEL} bugs out of ${bugs.length} total bugs`);

// Group bugs by repository
const bugsByRepo = eligibleBugs.reduce((acc: { [key: string]: any[] }, bug: any) => {
    if (!acc[bug.repo]) {
        acc[bug.repo] = [];
    }
    acc[bug.repo].push(bug);
    return acc;
}, {});

// Get list of repos with their bug counts
const repoStats = Object.entries(bugsByRepo)
    .map(([repo, bugs]) => ({
        repo,
        count: bugs.length
    }))
    .sort((a, b) => b.count - a.count); // Sort by count descending

console.log('\n=== Available Repositories with Easy Bugs ===');
repoStats.forEach(({repo, count}) => {
    console.log(`${repo}: ${count} eligible ${DIFFICULTY_LEVEL} bugs`);
});

// Select bugs trying to maximize repo diversity
let selectedBugs: any[] = [];
let currentRepoIndex = 0;

while (selectedBugs.length < SAMPLE_SIZE && currentRepoIndex < repoStats.length) {
    const repo = repoStats[currentRepoIndex].repo;
    const repoBugs = bugsByRepo[repo];
    
    // Take one bug from current repo if available
    if (repoBugs.length > 0) {
        selectedBugs.push(repoBugs.shift()); // Take and remove the first bug
    }
    
    // Move to next repo, or back to start if we've gone through all repos
    currentRepoIndex = (currentRepoIndex + 1) % repoStats.length;
    
    // Break if we've gone through all repos and can't find more bugs
    if (currentRepoIndex === 0 && selectedBugs.length < SAMPLE_SIZE && 
        repoStats.every(({repo}) => bugsByRepo[repo].length === 0)) {
        break;
    }
}

// Write to output file
fs.writeFileSync(OUTPUT_FILE, JSON.stringify(selectedBugs, null, 2));

// Print statistics
console.log('\n=== Sample Generation Statistics ===');
console.log(`Total ${DIFFICULTY_LEVEL} bugs selected: ${selectedBugs.length}`);
console.log('\nDistribution by repository:');
const finalDistribution = selectedBugs.reduce((acc: { [key: string]: number }, bug: any) => {
    acc[bug.repo] = (acc[bug.repo] || 0) + 1;
    return acc;
}, {});

Object.entries(finalDistribution)
    .sort((a, b) => b[1] - a[1])
    .forEach(([repo, count]) => {
        console.log(`- ${repo}: ${count} bugs`);
    });