如何通过Python SDK获取Azure Git仓库提交中的具体变更行？

我正在测试AI驱动的PR审查工具，需要获取Azure Git仓库PR中的变更行内容交给GPT-4审查。目前使用Azure DevOps SDK只能读取变更文件的完整内容，无法获取仅变更的行，请问有什么解决办法？

方法1：调用Azure DevOps REST API的Diff端点 #

Azure DevOps提供了直接返回PR差异的REST API，结果包含每个文件的变更块（hunk），其中明确标记了变更行的位置和内容，这是最直接的官方解决方案。

代码示例 #

import os
import requests
from dotenv import load_dotenv

load_dotenv()

PAT = os.getenv('PAT')
ORG_URL = os.getenv('ORG_URL')
PROJECT_NAME = os.getenv('PROJECT_NAME')
REPO_ID = os.getenv('REPO_NAME')
PR_ID = 2

# 构建Diff API请求地址
diff_url = f"{ORG_URL}/{PROJECT_NAME}/_apis/git/repositories/{REPO_ID}/pullRequests/{PR_ID}/diffs?api-version=7.1-preview.1"

# 设置认证请求头
headers = {
    'Authorization': f'Basic {PAT}'
}

# 发送请求并解析结果
response = requests.get(diff_url, headers=headers)
response.raise_for_status()
diff_data = response.json()

# 遍历并输出变更行
for file_diff in diff_data['changes']:
    print(f"\n文件路径: {file_diff['item']['path']}")
    if 'hunks' in file_diff:
        for hunk in file_diff['hunks']:
            print(f"变更范围: 原文件行 {hunk['oldStart']}-{hunk['oldStart'] + hunk['oldLength'] -1} → 新文件行 {hunk['newStart']}-{hunk['newStart'] + hunk['newLength'] -1}")
            print("变更内容:")
            for line in hunk['lines']:
                # 根据行类型添加前缀标识
                prefix = '+' if line['lineType'] == 'add' else '-' if line['lineType'] == 'remove' else ' '
                print(f"{prefix}{line['content']}")

方法2：使用Git命令行工具 #

如果运行环境允许执行Git命令，可以通过克隆仓库、对比PR的源分支与目标分支来获取变更行，适合需要本地处理差异的场景。

代码示例 #

import os
import subprocess
from dotenv import load_dotenv

load_dotenv()

REPO_URL = f"{os.getenv('ORG_URL')}/{os.getenv('PROJECT_NAME')}/_git/{os.getenv('REPO_NAME')}"
TARGET_BRANCH = "main"  # PR的目标分支
SOURCE_BRANCH = "feature-branch"  # PR的源分支

# 克隆临时仓库（若不存在）
temp_repo_path = './temp_pr_repo'
if not os.path.exists(temp_repo_path):
    subprocess.run(['git', 'clone', REPO_URL, temp_repo_path], check=True)

os.chdir(temp_repo_path)

# 拉取最新代码
subprocess.run(['git', 'pull'], check=True)

# 获取分支间的差异内容
diff_result = subprocess.run(
    ['git', 'diff', f"{TARGET_BRANCH}...{SOURCE_BRANCH}"],
    capture_output=True,
    text=True,
    check=True
)

# 输出变更行
print(diff_result.stdout)

方法3：结合GitPython本地计算差异 #

通过Azure DevOps SDK获取PR的源/目标Commit ID，再用GitPython库拉取对应版本的文件内容，最后本地计算行差异，适合需要自定义差异解析逻辑的场景。

代码示例 #

import os
import difflib
from git import Repo
from dotenv import load_dotenv
from azure.devops.connection import Connection
from msrest.authentication import BasicAuthentication

load_dotenv()

PAT = os.getenv('PAT')
ORG_URL = os.getenv('ORG_URL')
PROJECT_NAME = os.getenv('PROJECT_NAME')
REPO_ID = os.getenv('REPO_NAME')
PR_ID = 2

# 初始化Azure DevOps连接
credentials = BasicAuthentication('', PAT)
connection = Connection(base_url=ORG_URL, creds=credentials)
git_client = connection.clients.get_git_client()

# 获取PR的源/目标分支Commit信息
pull_request = git_client.get_pull_request(REPO_ID, PR_ID, project=PROJECT_NAME)
target_commit_id = pull_request.target_ref_name.split('/')[-1]
source_commit_id = pull_request.source_ref_name.split('/')[-1]

# 克隆或打开本地仓库
repo_path = './temp_pr_repo'
if not os.path.exists(repo_path):
    repo = Repo.clone_from(f"{ORG_URL}/{PROJECT_NAME}/_git/{REPO_ID}", repo_path)
else:
    repo = Repo(repo_path)

# 获取两个Commit的文件树
target_tree = repo.commit(target_commit_id).tree
source_tree = repo.commit(source_commit_id).tree

# 计算并输出文件差异
diff_index = target_tree.diff(source_tree)
for diff in diff_index:
    print(f"\n文件: {diff.a_path}")
    if diff.change_type in ['M', 'A']:
        # 读取文件的新旧版本内容
        old_content = diff.a_blob.data_stream.read().decode('utf-8').splitlines() if diff.a_blob else []
        new_content = diff.b_blob.data_stream.read().decode('utf-8').splitlines()
        
        # 生成行级差异
        diff_lines = difflib.unified_diff(old_content, new_content, fromfile=diff.a_path, tofile=diff.b_path)
        print('\n'.join(diff_lines))

内容的提问来源于stack exchange，提问作者Federico