// Copyright (c) 2019, QuantStack and Mamba Contributors // // Distributed under the terms of the BSD 3-Clause License. // // The full license is in the file LICENSE, distributed with this software. #ifndef MAMBA_CORE_SUBDIRDATA_HPP #define MAMBA_CORE_SUBDIRDATA_HPP #include #include #include #include #include #include "mamba/core/error_handling.hpp" #include "mamba/core/package_cache.hpp" #include "mamba/core/subdir_parameters.hpp" #include "mamba/download/downloader.hpp" #include "mamba/download/parameters.hpp" #include "mamba/fs/filesystem.hpp" #include "mamba/specs/channel.hpp" #include "mamba/specs/conda_url.hpp" #include "mamba/specs/platform.hpp" namespace mamba { namespace specs { class Channel; } /** * Handling of a subdirectory metadata. * * These metadata are used and stored to check if a subdirectory index is up to date, * where it comes from, and what protocols are supported to fetch it. */ class SubdirMetadata { public: struct HttpMetadata { std::string url; std::string etag; std::string last_modified; std::string cache_control; }; using expected_subdir_metadata = tl::expected; /** Read the metadata from a lightweight file containing only these metadata. */ static auto read_state_file( // const fs::u8path& state_file, const fs::u8path& repodata_file ) -> expected_subdir_metadata; /** Read the metadata from the ``repodata.json`` header. */ static auto read_from_repodata_json(const fs::u8path& json) -> expected_subdir_metadata; /** Read from any of state file or ``repodata.json`` depending on extension. */ static auto read(const fs::u8path& file) -> expected_subdir_metadata; [[nodiscard]] auto is_valid_metadata(const fs::u8path& file) const -> bool; [[nodiscard]] auto url() const -> const std::string&; [[nodiscard]] auto etag() const -> const std::string&; [[nodiscard]] auto last_modified() const -> const std::string&; [[nodiscard]] auto cache_control() const -> const std::string&; /** Check if zst is available and freshly checked. */ [[nodiscard]] auto has_up_to_date_zst() const -> bool; void set_http_metadata(HttpMetadata data); void set_zst(bool value); void store_file_metadata(const fs::u8path& file); /** Write the metadata to a lightweight file. */ void write_state_file(const fs::u8path& file); friend void to_json(nlohmann::json& j, const SubdirMetadata& data); friend void from_json(const nlohmann::json& j, SubdirMetadata& data); private: #ifdef _WIN32 using time_type = std::chrono::system_clock::time_point; #else using time_type = fs::file_time_type; #endif struct CheckedAt { bool value; std::time_t last_checked; bool has_expired() const; }; HttpMetadata m_http; std::optional m_has_zst; time_type m_stored_mtime; std::size_t m_stored_file_size; friend void to_json(nlohmann::json& j, const CheckedAt& ca); friend void from_json(const nlohmann::json& j, CheckedAt& ca); }; /** * Channel sub-directory (i.e. a platform) packages index. * * Handles downloading of the index from the server and cache generation. * This only handles traditional ``repodata.json`` full indexes. * This abstraction does not load the index in memory, with is done by the @ref Database. * * Upon creation, the caches are checked for a valid and up to date index. * This can be inspected with @ref valid_cache_found. * The created subdirs are typically used with @ref SubdirIndexLoader::download_required_indexes * which will download the missing, invalid, or outdated indexes as needed. */ class SubdirIndexLoader { public: /** * Download the missing, invalid, or outdated indexes as needed in parallel. * * It first creates check requests to update some metadata, then download the indexes. * The result can be inspected with the input subdirs methods, such as * @ref valid_cache_found, @ref valid_json_cache_path etc. */ template [[nodiscard]] static auto download_required_indexes( SubdirIter1 subdirs_first, SubdirIter2 subdirs_last, const SubdirDownloadParams& subdir_params, const specs::AuthenticationDataBase& auth_info, const download::mirror_map& mirrors, const download::Options& download_options, const download::RemoteFetchParams& remote_fetch_params, download::Monitor* check_monitor = nullptr, download::Monitor* download_monitor = nullptr ) -> expected_t; template [[nodiscard]] static auto download_required_indexes( Subdirs& subdirs, const SubdirDownloadParams& subdir_params, const specs::AuthenticationDataBase& auth_info, const download::mirror_map& mirrors, const download::Options& download_options, const download::RemoteFetchParams& remote_fetch_params, download::Monitor* check_monitor = nullptr, download::Monitor* download_monitor = nullptr ) -> expected_t; /** Check existing caches for a valid index validity and freshness. */ static auto create( const SubdirParams& params, specs::Channel channel, specs::DynamicPlatform platform, MultiPackageCache& caches, std::string repodata_filename = "repodata.json" ) -> expected_t; [[nodiscard]] auto is_noarch() const -> bool; [[nodiscard]] auto is_local() const -> bool; [[nodiscard]] auto channel() const -> const specs::Channel&; [[nodiscard]] auto name() const -> std::string; [[nodiscard]] auto channel_id() const -> const std::string&; [[nodiscard]] auto platform() const -> const specs::DynamicPlatform&; [[nodiscard]] auto metadata() const -> const SubdirMetadata&; [[nodiscard]] auto repodata_url() const -> specs::CondaURL; [[nodiscard]] auto caching_is_forbidden() const -> bool; [[nodiscard]] auto valid_cache_found() const -> bool; [[nodiscard]] auto valid_libsolv_cache_path() const -> expected_t; [[nodiscard]] auto writable_libsolv_cache_path() const -> fs::u8path; [[nodiscard]] auto valid_json_cache_path() const -> expected_t; void clear_valid_cache_files(); private: // This paths are pointing to what is found when iterating over the cache directories. // The expired found is the first one, which could be improved by keeping the freshest one. // This could improve caching in some HTTP 304 cases. // A possible improvement would be to keep all path, metadatas, and writable status in a // single vector and sort them by recency. // This would also give a public option for `clear`-ing all writable caches, not just the // valid one. SubdirMetadata m_metadata; specs::Channel m_channel; fs::u8path m_valid_cache_path; std::optional m_expired_cache_path; fs::u8path m_writable_pkgs_dir; specs::DynamicPlatform m_platform; std::string m_repodata_filename; std::string m_json_filename; std::string m_solv_filename; bool m_valid_cache_found = false; bool m_json_cache_valid = false; bool m_solv_cache_valid = false; SubdirIndexLoader( const SubdirParams& params, specs::Channel channel, std::string platform, MultiPackageCache& caches, std::string repodata_fn = "repodata.json" ); [[nodiscard]] auto repodata_url_path() const -> std::string; [[nodiscard]] auto valid_json_cache_path_unchecked() const -> fs::u8path; [[nodiscard]] auto valid_state_file_path_unchecked() const -> fs::u8path; [[nodiscard]] auto valid_libsolv_cache_path_unchecked() const -> fs::u8path; /********************************************************* * Implementation details of SubdirIndexLoader::create * *********************************************************/ void load(const MultiPackageCache& caches, const SubdirParams& params); void load_cache(const MultiPackageCache& caches, const SubdirParams& params); /**************************************************************************** * Implementation details of SubdirIndexLoader::download_required_indexes * ****************************************************************************/ auto use_existing_cache() -> expected_t; auto finalize_transfer(SubdirMetadata::HttpMetadata http_data, const fs::u8path& artifact) -> expected_t; void refresh_last_write_time(const fs::u8path& json_file, const fs::u8path& solv_file); template static auto build_all_check_requests(First subdirs_first, End subdirs_last, const SubdirDownloadParams& params) -> download::MultiRequest; auto build_check_requests(const SubdirDownloadParams& params) -> download::MultiRequest; template static auto build_all_index_requests(First subdirs_first, End subdirs_last, const SubdirDownloadParams& params) -> download::MultiRequest; auto build_index_request(const SubdirDownloadParams& params) -> std::optional; [[nodiscard]] static auto download_requests( download::MultiRequest index_requests, const specs::AuthenticationDataBase& auth_info, const download::mirror_map& mirrors, const download::Options& download_options, const download::RemoteFetchParams& remote_fetch_params, download::Monitor* download_monitor ) -> expected_t; }; /** * Compute an id from a URL. * * This is intended to keep unique, filesystem-safe, cache entries in the cache directory. * @see cache_filename_from_url */ [[nodiscard]] auto cache_name_from_url(std::string url) -> std::string; /** * Compute a filename from a URL. * * This is intended to keep unique, filesystem-safe, cache entries in the cache directory. * Contrary to conda original function, this one expects a full url (that is * channel url + / + repodata_fn). * It is not the responsibility of this function to decide whether it should concatenate base * url and repodata depending on repodata value and old behavior support. */ [[nodiscard]] auto cache_filename_from_url(std::string url) -> std::string; /** * Create cache directory with correct permissions * * @return The path to the directory created */ auto create_cache_dir(const fs::u8path& cache_path) -> std::string; /***************************************** * Implementation of SubdirIndexLoader * *****************************************/ template auto SubdirIndexLoader::download_required_indexes( SubdirIter1 subdirs_first, SubdirIter2 subdirs_last, const SubdirDownloadParams& subdir_params, const specs::AuthenticationDataBase& auth_info, const download::mirror_map& mirrors, const download::Options& download_options, const download::RemoteFetchParams& remote_fetch_params, download::Monitor* check_monitor, download::Monitor* download_monitor ) -> expected_t { auto result = download_requests( build_all_check_requests(subdirs_first, subdirs_last, subdir_params), auth_info, mirrors, download_options, remote_fetch_params, check_monitor ); // Allow to continue if failed checks, unless asked to stop. constexpr auto is_interrupted = [](const auto& e) { return e.error_code() == mamba_error_code::user_interrupted; }; if (!result.has_value() && result.map_error(is_interrupted).error()) { return result; } return download_requests( build_all_index_requests(subdirs_first, subdirs_last, subdir_params), auth_info, mirrors, download_options, remote_fetch_params, download_monitor ); } template auto SubdirIndexLoader::download_required_indexes( Subdirs& subdirs, const SubdirDownloadParams& subdir_params, const specs::AuthenticationDataBase& auth_info, const download::mirror_map& mirrors, const download::Options& download_options, const download::RemoteFetchParams& remote_fetch_params, download::Monitor* check_monitor, download::Monitor* download_monitor ) -> expected_t { return download_required_indexes( subdirs.begin(), subdirs.end(), subdir_params, auth_info, mirrors, download_options, remote_fetch_params, check_monitor, download_monitor ); } template auto SubdirIndexLoader::build_all_check_requests( First subdirs_first, End subdirs_last, const SubdirDownloadParams& params ) -> download::MultiRequest { download::MultiRequest requests; for (; subdirs_first != subdirs_last; ++subdirs_first) { // TODO(C++23): We make a special handling of iterators of pointers due to the // difficulty and necessity to create a range of references from Python objects. SubdirIndexLoader* p_subdir = nullptr; if constexpr (std::is_pointer_v>) { p_subdir = *subdirs_first; } else { p_subdir = &(*subdirs_first); } if (p_subdir != nullptr && !p_subdir->valid_cache_found()) { auto check_list = p_subdir->build_check_requests(params); std::move(check_list.begin(), check_list.end(), std::back_inserter(requests)); } } return requests; } template auto SubdirIndexLoader::build_all_index_requests( First subdirs_first, End subdirs_last, const SubdirDownloadParams& params ) -> download::MultiRequest { download::MultiRequest requests; for (; subdirs_first != subdirs_last; ++subdirs_first) { // TODO(C++23): We make a special handling of iterators of pointers due to the // difficulty and necessity to create a range of references from Python objects. SubdirIndexLoader* p_subdir = nullptr; if constexpr (std::is_pointer_v>) { p_subdir = *subdirs_first; } else { p_subdir = &(*subdirs_first); } if (!p_subdir->valid_cache_found()) { if (auto request = p_subdir->build_index_request(params)) { requests.push_back(*std::move(request)); } } } return requests; } } #endif