path: root/src/Cache.cpp

#include "../include/Cache.hpp"
#include "../include/env.hpp"
#include "../include/ResourceCache.hpp"
#include "../include/FileUtil.hpp"
#include "../include/Gif.hpp"
#include "../include/Chatbar.hpp"
#include "../include/WebPagePreview.hpp"
#include <boost/filesystem/convenience.hpp>
#include <unordered_map>
#include <process.hpp>
#include <odhtdb/Hash.hpp>
#include <sibs/SafeSerializer.hpp>
#include <sibs/SafeDeserializer.hpp>
#include <libpreview.h>

#if OS_FAMILY == OS_FAMILY_POSIX
#include <pwd.h>
#else
#include <string>
#endif

using namespace std;
using namespace TinyProcessLib;

namespace dchat
{
    static unordered_map<string, ContentByUrlResult> contentUrlCache;
    const i64 CONTENT_NOT_VISIBLE_AGE_MS = 20000; // Delete content from cache after a specified amount of time if the content is not visible on the screen
    
    static boost::filesystem::path getHomeDir()
    {
    #if OS_FAMILY == OS_FAMILY_POSIX
        const char *homeDir = getenv("HOME");
        if(!homeDir)
        {
            passwd *pw = getpwuid(getuid());
            homeDir = pw->pw_dir;
        }
        return boost::filesystem::path(homeDir);
    #elif OS_FAMILY == OS_FAMILY_WINDOWS
        BOOL ret;
        HANDLE hToken;
        std::wstring homeDir;
        DWORD homeDirLen = MAX_PATH;
        homeDir.resize(homeDirLen);

        if (!OpenProcessToken(GetCurrentProcess(), TOKEN_READ, &hToken))
            return Result<FileString>::Err("Failed to open process token");

        if (!GetUserProfileDirectory(hToken, &homeDir[0], &homeDirLen))
        {
            CloseHandle(hToken);
            return Result<FileString>::Err("Failed to get home directory");
        }

        CloseHandle(hToken);
        homeDir.resize(wcslen(homeDir.c_str()));
        return boost::filesystem::path(homeDir);
    #endif
    }
    
    boost::filesystem::path Cache::getDchatDir()
    {
        boost::filesystem::path dchatHomeDir = getHomeDir() / ".local" / "share" / "dchat";
        boost::filesystem::create_directories(dchatHomeDir);
        return dchatHomeDir;
    }
    
    boost::filesystem::path Cache::getImagesDir()
    {
        boost::filesystem::path imagesDir = getDchatDir() / "images";
        boost::filesystem::create_directories(imagesDir);
        return imagesDir;
    }
    
    void Cache::loadBindsFromFile()
    {
        StringView fileContent;
        try
        {
            fileContent = getFileContent(getDchatDir() / "binds");
            sibs::SafeDeserializer deserializer((const u8*)fileContent.data, fileContent.size);
            
            while(!deserializer.empty())
            {
                u8 keySize = deserializer.extract<u8>();
                string key;
                key.resize(keySize);
                deserializer.extract((u8*)&key[0], keySize);
                
                u8 valueSize = deserializer.extract<u8>();
                string value;
                value.resize(valueSize);
                deserializer.extract((u8*)&value[0], valueSize);
                
                Chatbar::addBind(key, value, false);
            }
        }
        catch(FileException &e)
        {
            fprintf(stderr, "Failed to read binds from file, reason: %s\n", e.what());
        }
        
        delete fileContent.data;
    }
    
    void Cache::replaceBindsInFile(const unordered_map<string, string> &binds)
    {
        sibs::SafeSerializer serializer;
        for(auto &it : binds)
        {
            serializer.add((u8)it.first.size());
            serializer.add((const u8*)it.first.data(), it.first.size());
            
            serializer.add((u8)it.second.size());
            serializer.add((const u8*)it.second.data(), it.second.size());
        }
        fileReplace(getDchatDir() / "binds", StringView((const char*)serializer.getBuffer().data(), serializer.getBuffer().size()));
    }
    
    static ContentByUrlResult loadImageFromFile(const boost::filesystem::path &filepath)
    {
        StringView fileContent;
        try
        {
            fileContent = getFileContent(filepath);
            
            sf::String webPageTitle;
            sf::String webPageDescription;
            bool foundHtmlContent = false;
            preview_state state;
            preview_init(&state);
            size_t offset = 0;
            do
            {
                // TODO: Get file content before doing this, the file might be in utf-16 encoding. That can happen for example if file contains html.
                // Content type can be retrieved from HTTP response header when downloading content
                offset += preview_step(&state, fileContent.data + offset, fileContent.size - offset);
                if(state.step_result == PREVIEW_FOUND_IMAGE)
                {
                    if(Gif::isDataGif(fileContent))
                    {
                        Gif *gif = new Gif(move(fileContent));
                        return { gif, ContentByUrlResult::Type::CACHED };
                    }
                    else
                    {
                        sf::Texture *texture = new sf::Texture();
                        if(texture->loadFromMemory(fileContent.data, fileContent.size))
                        {
                            delete[] fileContent.data;
                            fileContent.data = nullptr;
                            texture->setSmooth(true);
                            texture->generateMipmap();
                            return { texture, ContentByUrlResult::Type::CACHED };
                        }
                        delete texture;
                    }
                    break;
                }
                else if(state.step_result == PREVIEW_FOUND_TITLE)
                {
                    foundHtmlContent = true;
                    webPageTitle = sf::String::fromUtf8(state.meta_content, state.meta_content + state.meta_content_length);
                }
                else if(state.step_result == PREVIEW_FOUND_DESCRIPTION)
                {
                    foundHtmlContent = true;
                    webPageDescription = sf::String::fromUtf8(state.meta_content, state.meta_content + state.meta_content_length);
                }
            } while(offset < fileContent.size);
            
            delete[] fileContent.data;
            fileContent.data = nullptr;
            
            if(foundHtmlContent)
            {
                // TODO: Use move semantics for webPageTitle and webPageDescription when SFML supports it
                WebPagePreview *webPagePreview = new WebPagePreview(webPageTitle, webPageDescription);
                return { webPagePreview, ContentByUrlResult::Type::CACHED };
            }
        }
        catch(std::exception &e)
        {
            fprintf(stderr, "Failed to load image %s, reason: %s\n", filepath.string().c_str(), e.what());
        }
        return { (sf::Texture*)nullptr, ContentByUrlResult::Type::FAILED_DOWNLOAD };
    }
    
    Cache::Cache() : 
        alive(true)
    {
        downloadWaitThread = thread([this]
        {
            while(alive)
            {
                for(vector<ImageDownloadInfo>::iterator it = imageDownloadProcesses.begin(); it != imageDownloadProcesses.end();)
                {
                    int exitStatus;
                    if(it->process->try_get_exit_status(exitStatus))
                    {
                        boost::filesystem::path filepath = getImagesDir();
                        odhtdb::Hash urlHash(it->url.data(), it->url.size());
                        filepath /= urlHash.toString();
                            
                        ContentByUrlResult contentByUrlResult;
                        bool failed = exitStatus != 0;
                        if(failed)
                        {
                            contentByUrlResult = { (sf::Texture*)nullptr, ContentByUrlResult::Type::FAILED_DOWNLOAD };
                        }
                        else
                        {
                            contentByUrlResult = loadImageFromFile(filepath);
                            contentByUrlResult.lastAccessed = chrono::duration_cast<chrono::milliseconds>(chrono::steady_clock::now().time_since_epoch()).count();
                            if(contentByUrlResult.type == ContentByUrlResult::Type::CACHED)
                            {
                                printf("Download content from url: %s\n", it->url.c_str());
                            }
                        }
                        
                        imageDownloadMutex.lock();
                        contentUrlCache[it->url] = contentByUrlResult;
                        boost::filesystem::path downloadingFilepath = filepath;
                        downloadingFilepath += ".downloading";
                        // Intentionally ignore failure, program should not crash if we fail to remove these files...
                        boost::system::error_code err;
                        boost::filesystem::remove(downloadingFilepath, err);
                        imageDownloadMutex.unlock();
                        it = imageDownloadProcesses.erase(it);
                    }
                    else
                        ++it;
                }
                
                while(alive && imageDownloadProcesses.empty() && imageDownloadProcessesQueue.empty())
                    this_thread::sleep_for(chrono::milliseconds(20));
                
                if(!imageDownloadProcessesQueue.empty())
                {
                    imageDownloadMutex.lock();
                    for(auto imageDownloadInfo : imageDownloadProcessesQueue)
                    {
                        imageDownloadProcesses.push_back(imageDownloadInfo);
                    }
                    imageDownloadProcessesQueue.clear();
                    imageDownloadMutex.unlock();
                }
                
                this_thread::sleep_for(chrono::milliseconds(20));
            }
        });
        
        checkContentAccessTimeThread = thread([this]
        {
            while(alive)
            {
                this_thread::sleep_for(chrono::milliseconds(500));
                lock_guard<mutex> lock(imageDownloadMutex);
                
                i64 now = chrono::duration_cast<chrono::milliseconds>(chrono::steady_clock::now().time_since_epoch()).count();
                for(unordered_map<string, ContentByUrlResult>::iterator it = contentUrlCache.begin(); it != contentUrlCache.end();)
                {
                    if(it->second.type == ContentByUrlResult::Type::CACHED && now - it->second.lastAccessed > CONTENT_NOT_VISIBLE_AGE_MS)
                    {
                        switch(it->second.cachedType)
                        {
                            case ContentByUrlResult::CachedType::TEXTURE:
                            {
                                delete it->second.texture;
                                break;
                            }
                            case ContentByUrlResult::CachedType::GIF:
                            {
                                delete it->second.gif;
                                break;
                            }
                            case ContentByUrlResult::CachedType::WEB_PAGE_PREVIEW:
                            {
                                delete it->second.webPagePreview;
                                break;
                            }
                            default:
                                break;
                        }
                        it = contentUrlCache.erase(it);
                    }
                    else
                        ++it;
                }
            }
        });
    }
    
    Cache::~Cache()
    {
        alive = false;
        downloadWaitThread.join();
        checkContentAccessTimeThread.join();
    }
    
    static void replaceFileIgnoreError(const boost::filesystem::path &path)
    {
        try
        {
            fileReplace(path, StringView());
        }
        catch(FileException &e)
        {
            fprintf(stderr, "Failed to replace file: %s, reason: %s\n", path.string().c_str(), e.what());
        }
    }
    
    static string stringReplaceChar(const string &str, const string &from, const string &to)
    {
        string result = str;
        size_t pos = 0;
        while((pos = result.find(from, pos)) != string::npos)
        {
            result.replace(pos, from.size(), to);
            pos += to.size();
        }
        return result;
    }
    
    const ContentByUrlResult Cache::getContentByUrl(const string &url, int downloadLimitBytes)
    {
        lock_guard<mutex> lock(imageDownloadMutex);
        auto it = contentUrlCache.find(url);
        if(it != contentUrlCache.end())
        {
            it->second.lastAccessed = chrono::duration_cast<chrono::milliseconds>(chrono::steady_clock::now().time_since_epoch()).count();
            return it->second;
        }
        
        // TODO: Verify hashed url is not too long for filepath on windows
        boost::filesystem::path filepath = getImagesDir();
        odhtdb::Hash urlHash(url.data(), url.size());
        filepath /= urlHash.toString();
        
        boost::filesystem::path downloadingFilepath = filepath;
        downloadingFilepath += ".downloading";
        if(boost::filesystem::exists(downloadingFilepath))
        {
            // Intentionally ignore failure, program should not crash if we fail to remove these files...
            boost::system::error_code err;
            boost::filesystem::remove(filepath, err);
            boost::filesystem::remove(downloadingFilepath, err);
        }
        
        // TODO: Do not load content in this thread. Return LOADING status and load it in another thread, because with a lot of images, chat can freeze
        ContentByUrlResult contentByUrlResult = loadImageFromFile(filepath);
        if(contentByUrlResult.type == ContentByUrlResult::Type::CACHED)
        {
            contentByUrlResult.lastAccessed = chrono::duration_cast<chrono::milliseconds>(chrono::steady_clock::now().time_since_epoch()).count();
            contentUrlCache[url] = contentByUrlResult;
            printf("Loaded content from file cache: %s\n", url.c_str());
            return contentByUrlResult;
        }
        else if(contentByUrlResult.type == ContentByUrlResult::Type::FAILED_DOWNLOAD && boost::filesystem::exists(filepath))
        {
            contentUrlCache[url] = contentByUrlResult;
            return contentByUrlResult;
        }
        
        replaceFileIgnoreError(downloadingFilepath);
        ContentByUrlResult result((sf::Texture*)nullptr, ContentByUrlResult::Type::DOWNLOADING);
        contentUrlCache[url] = result;
        
        string downloadLimitBytesStr = to_string(downloadLimitBytes);
        
        string escapedUrl = stringReplaceChar(url, "'", "");
        escapedUrl = stringReplaceChar(escapedUrl, "\\", "");
        Process::string_type cmd = "curl -L --silent -o '";
        cmd += filepath.native();
        cmd += "' --max-filesize " + downloadLimitBytesStr + " --range 0-" + downloadLimitBytesStr + " --url '" + escapedUrl + "'";
        // TODO: Use this instead of curl on windows: certutil.exe -urlcache -split -f "https://url/to/file" path/and/name/to/save/as/file
        Process *process = new Process(cmd, "", nullptr, nullptr, false);
        ImageDownloadInfo imageDownloadInfo { process, url };
        imageDownloadProcessesQueue.emplace_back(imageDownloadInfo);
        return result;
    }
}