介绍
后台是用springboot技术,其他同事做的算法是python,现在的需求是springboot调用python,python又需要调用其他的数据文件,比如dat文件,这个文件是app通过蓝牙获取智能戒指数据以后,保存到后台,后台需要将数据写入到dat文件里,供python调用,本身难点并不大,主要本地环境死windows,服务器环境是ubuntu,并且是通过docker部署的,这样文件的路径就会产生问题,需要适配
代码部分
先处理一下文件路径问题,路径如下
如何能保证不同环境都能正确读到文件呢
答案是:使用临时文件复制
在 Spring Boot 调用 Python 脚本的场景中使用临时文件的方式来实现跨平台和 Docker 兼容,主要基于以下几个关键原因:
1. 解决路径一致性问题
问题:不同操作系统路径格式不同(Windows: C:\path,Linux: /path),且 Docker 容器内路径与宿主机不同
方案:临时目录提供统一的绝对路径基准
Path tempDir = Files.createTempDirectory("prefix"); // 自动适应各平台路径格式
2. 解决文件权限问题
问题:直接操作项目资源文件可能因权限不足失败(尤其是 Docker 默认以非 root 用户运行)
方案:临时目录确保可写权限
// 显式设置权限(Linux/Unix需要)
targetPath.toFile().setExecutable(true);
3. 解决资源打包问题
问题:Spring Boot 打包后,resources 下的文件存在于 JAR 内,无法直接通过文件系统访问
方案:运行时复制到临时目录解压
ClassPathResource resource = new ClassPathResource("python/script.py");
Files.copy(resource.getInputStream(), tempPath); // 从JAR内解压到文件系统
4. 解决 Docker 环境隔离
问题:Docker 容器有独立文件系统,无法直接访问宿主机的项目资源
方案:构建镜像时复制资源,运行时使用临时目录
COPY src/main/resources/python /app/python # 构建时固化路径
5. 多线程安全写入
问题:多线程并发写入同一文件会导致冲突
方案:每个线程使用独立临时文件
Path threadSpecificFile = tempDir.resolve("thread_" + Thread.currentThread().getId() + ".dat");
6. 资源清理保障
问题:运行后残留文件可能积累
方案:标准化清理流程
finally {
deleteDirectory(tempDir.toFile()); // 确保删除临时文件
}
7. 调试与日志追踪
问题:直接操作原始文件难以追踪运行时状态
方案:临时文件提供独立运行环境
System.out.println("临时目录: " + tempDir); // 明确显示运行时文件位置
代码部分的思路大致是,先将文件复制到临时路径,然后往临时路径的文件里写内容,这个临时路径是可变的,所以不能写死,需要将路径当入参传给python文件
代码
@Service
public class PythonService {
public String executePythonScript(String waveData) {
Path tempDir = null;
try {
System.out.println("1:"+ DateUtils.dateTimeNow());
// 1. 创建临时目录(使用NIO API确保跨平台兼容)
tempDir = Files.createTempDirectory("python_workspace");
System.out.println("2:"+ DateUtils.dateTimeNow());
// 2. 复制资源
copyPythonResourcesToTemp(tempDir);
System.out.println("3:"+ DateUtils.dateTimeNow());
//写入数据
Path tempFile = tempDir.resolve("python/raw_data/bp_106_63.dat");
try (BufferedWriter writer = new BufferedWriter(new FileWriter(tempFile.toFile()))) {
writer.write("");//先清空数据
writer.write(waveData);
}
// 3. 构建命令(使用绝对路径)
String pythonScriptPath = tempDir.resolve("python/realTime_predict.py").toString();
String[] command = {
"python3",
pythonScriptPath
};
String pythonPath = "C:\\xxxx\\Python\\Python311\\python.exe"; // 替换为你的实际路径
boolean isWindows = System.getProperty("os.name").toLowerCase().contains("win");
if (isWindows) {
command = new String[]{pythonPath, pythonScriptPath,tempFile.toString()};
} else {
command = new String[]{"python3",pythonScriptPath,tempFile.toString()};
}
System.out.println("4:"+ DateUtils.dateTimeNow());
// 4. 执行命令
ProcessBuilder pb = new ProcessBuilder(command);
pb.directory(tempDir.toFile());
pb.redirectErrorStream(true);
System.out.println("5:"+ DateUtils.dateTimeNow());
Process process = pb.start();
String output = new BufferedReader(
new InputStreamReader(process.getInputStream(), StandardCharsets.UTF_8))
.lines().collect(Collectors.joining("\n"));
System.out.println("6:"+ DateUtils.dateTimeNow());
int exitCode = process.waitFor();
System.out.println("output:"+ output);
if (exitCode != 0) {
return "计算出错";
}
System.out.println("7:"+ DateUtils.dateTimeNow());
System.out.println("output:"+ output);
String[] split = output.split("\n");
return split[split.length-1];
} catch (Exception e) {
e.printStackTrace();
//throw new RuntimeException("执行Python脚本出错: " + e.getMessage(), e);
} finally {
// 生产环境建议保留日志,开发时可清理
if (tempDir != null) {
deleteDirectory(tempDir.toFile());
}
}
return "";
}
private void copyPythonResourcesToTemp(Path tempDir) throws IOException {
// 创建python子目录
Files.createDirectories(tempDir.resolve("python"));
// 创建必要的子目录结构
Files.createDirectories(tempDir.resolve("python/raw_data"));
// 使用Spring的ResourceUtils复制所有文件
copyResourceToTemp("python/realTime_predict.py", tempDir.resolve("python/realTime_predict.py"));
copyResourceToTemp("python/best_model.pth", tempDir.resolve("python/best_model.pth"));
copyResourceToTemp("python/model.py", tempDir.resolve("python/model.py"));
copyResourceToTemp("python/readData.py", tempDir.resolve("python/readData.py"));
// 确保先创建raw_data目录再复制数据文件
copyResourceToTemp("python/raw_data/bp_106_63.dat", tempDir.resolve("python/raw_data/bp_106_63.dat"));
// 复制requirements.txt(如果存在)
copyResourceIfExists("python/requirements.txt", tempDir.resolve("python/requirements.txt"));
// 可选:复制requirements.txt
ClassPathResource requirements = new ClassPathResource("python/requirements.txt");
if (requirements.exists()) {
Files.copy(requirements.getInputStream(),
tempDir.resolve("python/requirements.txt"),
StandardCopyOption.REPLACE_EXISTING);
}
}
// 新增方法:安全复制资源(仅当资源存在时)
private void copyResourceIfExists(String resourcePath, Path targetPath) throws IOException {
ClassPathResource resource = new ClassPathResource(resourcePath);
if (resource.exists()) {
try (InputStream in = resource.getInputStream()) {
Files.copy(in, targetPath, StandardCopyOption.REPLACE_EXISTING);
}
}
}
// 专用资源复制方法(处理JAR内资源)
private void copyResourceToTemp(String resourcePath, Path targetPath) throws IOException {
ClassPathResource resource = new ClassPathResource(resourcePath);
try (InputStream in = resource.getInputStream()) {
Files.copy(in, targetPath, StandardCopyOption.REPLACE_EXISTING);
}
// 设置可执行权限(Linux/Unix需要)
targetPath.toFile().setExecutable(true);
}
// 辅助方法:递归删除目录
private void deleteDirectory(File directory) {
if (directory.exists()) {
File[] files = directory.listFiles();
if (files != null) {
for (File file : files) {
if (file.isDirectory()) {
deleteDirectory(file);
} else {
file.delete();
}
}
}
directory.delete();
}
}
}
python部分
if __name__ == "__main__":
# 测试预测
if len(sys.argv) < 2:
print("请提供数据文件路径作为参数")
sys.exit(1)
input_file = sys.argv[1] # 获取Java传递的文件路径参数
try:
sbp, dbp = predict_bp(input_file)
print(f"{sbp}/{dbp} mmHg")
except Exception as e:
print(f"错误: {str(e)}", file=sys.stderr)
sys.exit(1)
input_file就是
command = new String[]{“python3”,pythonScriptPath,tempFile.toString()};
里的tempFile.toString()
注意点
java调用python输出中文是乱码,在python里设置
from pathlib import Path
# 强制设置标准输出编码为UTF-8
sys.stdout = io.TextIOWrapper(sys.stdout.buffer, encoding='utf-8', errors='replace')
python找不到python/raw_data里的文件
def get_data_file_path(self, filename):
"""获取数据文件的绝对路径"""
# 1. 尝试在同目录下的raw_data文件夹查找
script_dir = os.path.dirname(os.path.abspath(__file__))
data_path = os.path.join(script_dir, "raw_data", filename)
# 2. 如果在开发环境找不到,尝试在上级目录的raw_data查找
if not os.path.exists(data_path):
parent_dir = os.path.dirname(script_dir)
data_path = os.path.join(parent_dir, "raw_data", filename)
# 3. 如果还是找不到,尝试在Docker环境路径查找
if not os.path.exists(data_path):
data_path = os.path.join("/app/python/raw_data", filename)
if not os.path.exists(data_path):
raise FileNotFoundError(f"Data file not found at: {data_path}")
return data_path
dockerFile注意点
dockerfile这个打包,困扰了我一天,主要遇到了库拉取不下来,下载离线库,又遇到了缺少其他依赖的问题
1.docker build报错
# 第一阶段:构建Python环境(使用官方镜像+国内pip源)
FROM python:3.9-slim AS python-builder
这个就遇到了
ERROR: failed to solve: python:3.9-slim: failed to resolve source metadata for docker.io/library/python:3.9-slim: failed commit on ref “unknown-sha256:b1fd1b5f83b18a7a7377874e3791c8104d5cf26c52677291a31d8805a9a3e5b0”: “unknown-sha256:b1fd1b5f83b18a7a7377874e3791c8104d5cf26c52677291a31d8805a9a3e5b0” failed size validation: 7630 != 7317: failed precondition
还有另一个库openjdk:11-jre-slim,也报错
ERROR: failed to solve: adoptopenjdk:11-jre-hotspot: failed to resolve source metadata for docker.io/library/adoptopenjdk:11-jre-hotspot: failed commit on ref “unknown-sha256:09a07bc840c63d79cfcc70a8960e0cead643b14cfdf6bdbca14a22bd6a9d3991”: “unknown-sha256:09a07bc840c63d79cfcc70a8960e0cead643b14cfdf6bdbca14a22bd6a9d3991” failed size validation: 7634 != 7377: failed precondition
解决的办法是,先docker pull一下这些库,因为比较大,docker build的时候可能会中断
docker pull python:3.9
docker pull openjdk:11-jdk-slim
2.离线包的问题
构建镜像(禁用网络访问)
docker build --no-cache --pull=false --network=none -t ring_1.0.0 .
因为上边的问题,我本来想着都用离线包,把requirements.txt里的依赖库全下载到本地,然后copy到docker里,结果
6.811 ERROR: Could not find a version that satisfies the requirement nvidia-cuda-cupti-cu1212.1.105; platform_system == “Linux” and platform_machine == “x86_64” (from torch) (from versions: none)
6.812 ERROR: No matching distribution found for nvidia-cuda-cupti-cu1212.1.105; platform_system == “Linux” and platform_machine == “x86_64”
如果一个一个去找,非常麻烦,还不一定匹配,所以还是需要用在线打包
3.在线打包的注意点
一定一定注意,名称是否正确
FROM python:3.9-slim as python-builder
我本来是这个,结果还是一直出错,后来发现是名称错了,可以用
docker images查看一下镜像名称
结果我本地的python镜像名是3.9不是3.9-slim,同样的,FROM openjdk:11-jre-slim 这个也要确认名称
FROM python:3.9 as python-builder
Dockerfile文件内容
# 第一阶段:构建Python环境
FROM python:3.9 as python-builder
WORKDIR /app
COPY ruoyi-admin/src/main/resources/python/requirements.txt .
RUN pip install --user -r requirements.txt && \
mkdir -p /app/python/raw_data
# 第二阶段:构建Java应用
FROM openjdk:11-jdk-slim
# 安装基础Python环境
RUN apt-get update && \
apt-get install -y --no-install-recommends \
python3 \
python3-pip \
&& rm -rf /var/lib/apt/lists/*
# 从python阶段复制依赖
COPY --from=python-builder /root/.local /root/.local
ENV PATH=/root/.local/bin:$PATH
# 设置工作目录
WORKDIR /app
# 复制应用文件
COPY ruoyi-admin/target/ring.jar /ring.jar
COPY ruoyi-admin/src/main/resources/python /app/python
# 设置权限
RUN chmod -R 755 /app/python && \
find /app/python -name "*.py" -exec chmod +x {} \; && \
chmod -R 777 /app/python/raw_data # 确保数据目录可写
# 环境变量
ENV PYTHON_SCRIPT_PATH=/app/python
ENV PYTHONUNBUFFERED=1
EXPOSE 8080
ENTRYPOINT ["java", "-jar", "/ring.jar"]