be/src/udf/python/python_env.h
Line | Count | Source |
1 | | // Licensed to the Apache Software Foundation (ASF) under one |
2 | | // or more contributor license agreements. See the NOTICE file |
3 | | // distributed with this work for additional information |
4 | | // regarding copyright ownership. The ASF licenses this file |
5 | | // to you under the Apache License, Version 2.0 (the |
6 | | // "License"); you may not use this file except in compliance |
7 | | // with the License. You may obtain a copy of the License at |
8 | | // |
9 | | // http://www.apache.org/licenses/LICENSE-2.0 |
10 | | // |
11 | | // Unless required by applicable law or agreed to in writing, |
12 | | // software distributed under the License is distributed on an |
13 | | // "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY |
14 | | // KIND, either express or implied. See the License for the |
15 | | // specific language governing permissions and limitations |
16 | | // under the License. |
17 | | |
18 | | #pragma once |
19 | | |
20 | | #include <gen_cpp/BackendService_types.h> |
21 | | |
22 | | #include <filesystem> |
23 | | #include <utility> |
24 | | |
25 | | #include "common/exception.h" |
26 | | #include "common/status.h" |
27 | | |
28 | | namespace doris { |
29 | | |
30 | | namespace fs = std::filesystem; |
31 | | |
32 | | enum class PythonEnvType { CONDA, VENV }; |
33 | | |
34 | | struct PythonVersion { |
35 | | std::string full_version; // e.g. "3.9.16" |
36 | | std::string base_path; // e.g. "/root/anaconda3/envs/python3.9" |
37 | | std::string executable_path; // e.g. "{base_path}/bin/python3" |
38 | | |
39 | 3.23k | PythonVersion() = default; |
40 | | |
41 | | explicit PythonVersion(std::string full_version, std::string base_path, |
42 | | std::string executable_path) |
43 | 61 | : full_version(std::move(full_version)), |
44 | 61 | base_path(std::move(base_path)), |
45 | 61 | executable_path(std::move(executable_path)) {} |
46 | | |
47 | 12.7k | bool operator==(const PythonVersion& other) const { |
48 | 12.7k | return full_version == other.full_version && base_path == other.base_path && |
49 | 12.7k | executable_path == other.executable_path; |
50 | 12.7k | } |
51 | | |
52 | | const std::string& get_base_path() const { return base_path; } |
53 | | |
54 | 122 | const std::string& get_executable_path() const { return executable_path; } |
55 | | |
56 | 36 | bool is_valid() const { |
57 | 36 | return !full_version.empty() && !base_path.empty() && !executable_path.empty() && |
58 | 36 | fs::exists(base_path) && fs::exists(executable_path); |
59 | 36 | } |
60 | | |
61 | 1.91k | std::string to_string() const { |
62 | 1.91k | return fmt::format("[full_version: {}, base_path: {}, executable_path: {}]", full_version, |
63 | 1.91k | base_path, executable_path); |
64 | 1.91k | } |
65 | | }; |
66 | | |
67 | | struct PythonEnvironment { |
68 | | std::string env_name; // e.g. "base" or "myenv" |
69 | | PythonVersion python_version; |
70 | | |
71 | | PythonEnvironment(const std::string& name, const PythonVersion& python_version); |
72 | | |
73 | | std::string to_string() const; |
74 | | |
75 | | bool is_valid() const; |
76 | | |
77 | | static Status scan_from_conda_root_path(const fs::path& conda_root_path, |
78 | | std::vector<PythonEnvironment>* environments); |
79 | | |
80 | | static Status scan_from_venv_root_path(const fs::path& venv_root_path, |
81 | | const std::vector<std::string>& interpreter_paths, |
82 | | std::vector<PythonEnvironment>* environments); |
83 | | }; |
84 | | |
85 | | class PythonEnvScanner { |
86 | | public: |
87 | 20 | PythonEnvScanner(const fs::path& env_root_path) : _env_root_path(env_root_path) {} |
88 | | |
89 | 16 | virtual ~PythonEnvScanner() = default; |
90 | | |
91 | | virtual Status scan() = 0; |
92 | | |
93 | | Status get_versions(std::vector<PythonVersion>* versions) const; |
94 | | |
95 | | Status get_version(const std::string& runtime_version, PythonVersion* version) const; |
96 | | |
97 | 2 | const std::vector<PythonEnvironment>& get_envs() const { return _envs; } |
98 | | |
99 | | std::string root_path() const { return _env_root_path.string(); } |
100 | | |
101 | | virtual PythonEnvType env_type() const = 0; |
102 | | |
103 | | virtual std::string to_string() const = 0; |
104 | | |
105 | | protected: |
106 | | fs::path _env_root_path; |
107 | | std::vector<PythonEnvironment> _envs; |
108 | | }; |
109 | | |
110 | | class CondaEnvScanner : public PythonEnvScanner { |
111 | | public: |
112 | 10 | CondaEnvScanner(const fs::path& python_root_path) : PythonEnvScanner(python_root_path) {} |
113 | | |
114 | | ~CondaEnvScanner() override = default; |
115 | | |
116 | | Status scan() override; |
117 | | |
118 | | std::string to_string() const override; |
119 | | |
120 | 3 | PythonEnvType env_type() const override { return PythonEnvType::CONDA; } |
121 | | }; |
122 | | |
123 | | class VenvEnvScanner : public PythonEnvScanner { |
124 | | public: |
125 | | VenvEnvScanner(const fs::path& python_root_path, |
126 | | const std::vector<std::string>& interpreter_paths) |
127 | 10 | : PythonEnvScanner(python_root_path), _interpreter_paths(interpreter_paths) {} |
128 | | |
129 | 7 | ~VenvEnvScanner() override = default; |
130 | | |
131 | | Status scan() override; |
132 | | |
133 | | std::string to_string() const override; |
134 | | |
135 | 1 | PythonEnvType env_type() const override { return PythonEnvType::VENV; } |
136 | | |
137 | | private: |
138 | | std::vector<std::string> _interpreter_paths; |
139 | | }; |
140 | | |
141 | | // Holds a PythonEnvScanner instance and centralizes the initialization check |
142 | | // to avoid accessing the scanner when the Python UDF feature is disabled. |
143 | | // This class is intended for internal use by PythonVersionManager only. |
144 | | class PythonEnvScannerHolder { |
145 | | public: |
146 | | Status init(PythonEnvType env_type, const fs::path& python_root_path, |
147 | | const std::string& python_venv_interpreter_paths); |
148 | | |
149 | 3.22k | const PythonEnvScanner& get() const { |
150 | 3.22k | if (!_env_scanner) { |
151 | 0 | throw Exception(ErrorCode::NOT_INITIALIZED, |
152 | 0 | "Set 'enable_python_udf_support = true' in be.conf to enable PythonUDF " |
153 | 0 | "feature"); |
154 | 0 | } |
155 | 3.22k | return *_env_scanner; |
156 | 3.22k | } |
157 | | |
158 | | private: |
159 | | std::unique_ptr<PythonEnvScanner> _env_scanner; |
160 | | }; |
161 | | |
162 | | class PythonVersionManager { |
163 | | public: |
164 | 3.23k | static PythonVersionManager& instance() { |
165 | 3.23k | static PythonVersionManager instance; |
166 | 3.23k | return instance; |
167 | 3.23k | } |
168 | | |
169 | | Status init(PythonEnvType env_type, const fs::path& python_root_path, |
170 | 12 | const std::string& python_venv_interpreter_paths) { |
171 | 12 | return _holder.init(env_type, python_root_path, python_venv_interpreter_paths); |
172 | 12 | } |
173 | | |
174 | 3.21k | Status get_version(const std::string& runtime_version, PythonVersion* version) const { |
175 | 3.21k | return _holder.get().get_version(runtime_version, version); |
176 | 3.21k | } |
177 | | |
178 | 2 | const std::vector<PythonEnvironment>& get_envs() const { return _holder.get().get_envs(); } |
179 | | |
180 | 2 | PythonEnvType env_type() const { return _holder.get().env_type(); } |
181 | | |
182 | 7 | std::string to_string() const { return _holder.get().to_string(); } |
183 | | |
184 | | std::vector<TPythonEnvInfo> env_infos_to_thrift() const; |
185 | | |
186 | | std::vector<TPythonPackageInfo> package_infos_to_thrift( |
187 | | const std::vector<std::pair<std::string, std::string>>& packages) const; |
188 | | |
189 | | private: |
190 | | PythonEnvScannerHolder _holder; |
191 | | }; |
192 | | |
193 | | // List installed pip packages for a given Python version. |
194 | | // Returns pairs of (package_name, version). |
195 | | Status list_installed_packages(const PythonVersion& version, |
196 | | std::vector<std::pair<std::string, std::string>>* packages); |
197 | | |
198 | | } // namespace doris |
199 | | |
200 | | namespace std { |
201 | | template <> |
202 | | struct hash<doris::PythonVersion> { |
203 | 12.7k | size_t operator()(const doris::PythonVersion& v) const noexcept { |
204 | 12.7k | return hash<string> {}(v.full_version); |
205 | 12.7k | } |
206 | | }; |
207 | | } // namespace std |