一次利用大模型完成Jacoco code coverage报告合并的尝试

prompt

最近学习了pseudo写prompt，一直没有机会尝试，刚好最近在讨论Jacoco覆盖率合并的事情，因此就想用大模型试试，prompt如下：

You major at merge cover report by jacoco.Both reports are for different versions of the same project.
你的主要任务就是合并由Jacoco生成的代码覆盖报告，两个报告都是针对同一个项目不同版本的代码。两个文件就是两个输入的report1、report2的对应文件
下面是一些如何处理的为代码：

def same_line_fc(report1,report2):
    # report2 is nearly cover report
    # comparte two coverage reports.You find line number of the same line  that the the html class is nc in report2 and fc in report1.
    # the line number of report2 append to a list 
    # the same line is the same java code,not the line number.
    #the line number append to a list


def merge_coverage_reports(report1, report2):
    # Merge two coverage reports
    same_list_number = same_line_fc(report1, report2)
    # change report2's html the line in the same_list_number html class to fc.And return the new report2


print(merge_coverage_reports(report1, report2))

大模型反馈

我把两个html上传到了大模型，在通过上面的prmopt想得到最终的结果，其实要是这样每次合并都需大模型成本也很高。但是没想到的是大模型直接给我了一个处理代码，而不是结果。

from bs4 import BeautifulSoup

def same_line_fc(report1_content, report2_content):
    soup1 = BeautifulSoup(report1_content, 'html.parser')
    soup2 = BeautifulSoup(report2_content, 'html.parser')

    # 获取所有行
    lines1 = soup1.find_all('span', class_='fc')
    lines2 = soup2.find_all('span', class_='nc')

    same_line_numbers = []

    # 创建一个字典来保存每行代码和对应的行号
    lines_dict1 = {line.get_text(): line['id'] for line in lines1}
    lines_dict2 = {line.get_text(): line['id'] for line in lines2}

    # 找到在report1中为fc，而在report2中为nc的行
    for code_line, line_id in lines_dict1.items():
        if code_line in lines_dict2:
            same_line_numbers.append(lines_dict2[code_line].replace('L', ''))

    return same_line_numbers

def merge_coverage_reports(report1_content, report2_content):
    same_list_number = same_line_fc(report1_content, report2_content)

    soup2 = BeautifulSoup(report2_content, 'html.parser')

    # 修改report2中相应行的类
    for line_number in same_list_number:
        line_span = soup2.find('span', id=f'L{line_number}')
        if line_span:
            line_span['class'] = 'fc'

    return str(soup2)

# 读取HTML文件内容
with open('./to/report1.html', 'r', encoding='utf-8') as file1:
    report1_content = file1.read()

with open('./to/report2.html', 'r', encoding='utf-8') as file2:
    report2_content = file2.read()

# 合并报告
merged_report_content = merge_coverage_reports(report1_content, report2_content)

# 输出合并后的报告
output_path = './to/merged_report.html'
with open(output_path, 'w', encoding='utf-8') as file_out:
    file_out.write(merged_report_content)

print(f"Merged report saved to {output_path}")

我也没修改直接案列就用了，结果也是出奇的好。

合并后的截图如下：

特别注意

其实如上的合并在Jacoco的解决原理上并不科学，一些合并前是红色，合并后是绿色的代码也并不一定是合理的以及正确的，这有可能只是一种视觉上的合并而不是真正意义上的Code Coverage的合并，但是大模型的这次利用给我开启了新的思路。