beeta/chat-service/src/services/CensorService.cpp

#include "CensorService.h"
#include <drogon/drogon.h>
#include <drogon/HttpClient.h>
#include <sstream>
#include <algorithm>
#include <cctype>

namespace services {

CensorService& CensorService::getInstance() {
    static CensorService instance;
    return instance;
}

void CensorService::initialize() {
    initialized_ = true;
    LOG_INFO << "CensorService initialized";
}

void CensorService::scheduleFetch() {
    LOG_INFO << "Scheduling censored words fetch in 2 seconds...";
    drogon::app().getLoop()->runAfter(2.0, [this]() {
        LOG_INFO << "Pre-fetching censored words from backend...";
        fetchCensoredWordsAsync();
    });
    // No periodic refresh - cache invalidation is triggered by backend
}

void CensorService::invalidateCache() {
    LOG_INFO << "Cache invalidation requested, fetching censored words from backend...";
    fetchCensoredWordsAsync();
}

void CensorService::fetchCensoredWordsFromBackend() {
    auto config = drogon::app().getCustomConfig();
    auto backendConfig = config.get("backend_api", Json::Value::null);
    std::string host;
    int port;

    if (backendConfig.isNull() || !backendConfig.isMember("host")) {
        host = "drogon-backend";
        port = 8080;
    } else {
        host = backendConfig.get("host", "drogon-backend").asString();
        port = backendConfig.get("port", 8080).asInt();
    }

    auto client = drogon::HttpClient::newHttpClient("http://" + host + ":" + std::to_string(port));
    auto req = drogon::HttpRequest::newHttpRequest();
    req->setMethod(drogon::Get);
    req->setPath("/api/internal/censored-words");

    std::pair<drogon::ReqResult, drogon::HttpResponsePtr> result = client->sendRequest(req, 5.0);

    if (result.first != drogon::ReqResult::Ok) {
        LOG_ERROR << "Failed to fetch censored words from backend: request failed";
        return;
    }

    auto resp = result.second;
    if (resp->getStatusCode() != drogon::k200OK) {
        LOG_ERROR << "Failed to fetch censored words from backend: HTTP " << resp->getStatusCode();
        return;
    }

    try {
        auto json = resp->getJsonObject();
        if (!json || !(*json)["success"].asBool()) {
            LOG_ERROR << "Failed to fetch censored words: invalid response";
            return;
        }

        std::string wordsStr = (*json)["censored_words"].asString();

        // Build new data in temporary variables
        std::vector<std::string> newWords;
        std::optional<std::regex> newPattern;

        if (!wordsStr.empty()) {
            std::stringstream ss(wordsStr);
            std::string word;
            while (std::getline(ss, word, ',') && newWords.size() < MAX_WORD_COUNT) {
                size_t start = word.find_first_not_of(" \t\r\n");
                size_t end = word.find_last_not_of(" \t\r\n");
                if (start != std::string::npos && end != std::string::npos) {
                    word = word.substr(start, end - start + 1);
                    // Skip empty words and words exceeding max length (ReDoS prevention)
                    if (!word.empty() && word.length() <= MAX_WORD_LENGTH) {
                        newWords.push_back(word);
                    } else if (word.length() > MAX_WORD_LENGTH) {
                        LOG_WARN << "Skipping censored word exceeding " << MAX_WORD_LENGTH << " chars";
                    }
                }
            }
            newPattern = buildCombinedPattern(newWords);
        }

        // Atomic swap under lock
        {
            std::unique_lock<std::shared_mutex> lock(mutex_);
            censoredWords_ = std::move(newWords);
            combinedPattern_ = std::move(newPattern);
        }

        LOG_DEBUG << "Fetched " << censoredWords_.size() << " censored words from backend";
    } catch (const std::exception& e) {
        LOG_ERROR << "Error parsing censored words response: " << e.what();
    }
}

void CensorService::fetchCensoredWordsAsync() {
    auto config = drogon::app().getCustomConfig();
    auto backendConfig = config.get("backend_api", Json::Value::null);
    std::string host;
    int port;

    if (backendConfig.isNull() || !backendConfig.isMember("host")) {
        host = "drogon-backend";
        port = 8080;
    } else {
        host = backendConfig.get("host", "drogon-backend").asString();
        port = backendConfig.get("port", 8080).asInt();
    }

    std::string url = "http://" + host + ":" + std::to_string(port);
    auto client = drogon::HttpClient::newHttpClient(url, drogon::app().getLoop());
    auto req = drogon::HttpRequest::newHttpRequest();
    req->setMethod(drogon::Get);
    req->setPath("/api/internal/censored-words");

    client->sendRequest(req, [this, client](drogon::ReqResult result, const drogon::HttpResponsePtr& resp) {
        if (result != drogon::ReqResult::Ok) {
            LOG_ERROR << "Async fetch censored words failed";
            return;
        }

        if (resp->getStatusCode() != drogon::k200OK) {
            LOG_ERROR << "Async fetch censored words failed: HTTP " << resp->getStatusCode();
            return;
        }

        try {
            auto json = resp->getJsonObject();
            if (!json || !(*json)["success"].asBool()) {
                LOG_ERROR << "Async fetch censored words: invalid response";
                return;
            }

            std::string wordsStr = (*json)["censored_words"].asString();

            // Build new data in temporary variables
            std::vector<std::string> newWords;
            std::optional<std::regex> newPattern;

            if (!wordsStr.empty()) {
                std::stringstream ss(wordsStr);
                std::string word;
                while (std::getline(ss, word, ',') && newWords.size() < MAX_WORD_COUNT) {
                    size_t start = word.find_first_not_of(" \t\r\n");
                    size_t end = word.find_last_not_of(" \t\r\n");
                    if (start != std::string::npos && end != std::string::npos) {
                        word = word.substr(start, end - start + 1);
                        // Skip empty words and words exceeding max length (ReDoS prevention)
                        if (!word.empty() && word.length() <= MAX_WORD_LENGTH) {
                            newWords.push_back(word);
                        } else if (word.length() > MAX_WORD_LENGTH) {
                            LOG_WARN << "Skipping censored word exceeding " << MAX_WORD_LENGTH << " chars";
                        }
                    }
                }
                newPattern = buildCombinedPattern(newWords);
            }

            // Atomic swap under lock
            {
                std::unique_lock<std::shared_mutex> lock(mutex_);
                censoredWords_ = std::move(newWords);
                combinedPattern_ = std::move(newPattern);
            }

            LOG_INFO << "Successfully fetched " << censoredWords_.size() << " censored words from backend";
        } catch (const std::exception& e) {
            LOG_ERROR << "Error parsing async censored words response: " << e.what();
        }
    }, 10.0);
}

std::optional<std::regex> CensorService::buildCombinedPattern(const std::vector<std::string>& words) {
    if (words.empty()) {
        return std::nullopt;
    }

    try {
        // Build combined pattern: \b(word1|word2|word3)\b
        std::string pattern = "\\b(";
        bool first = true;

        for (const auto& word : words) {
            if (!first) {
                pattern += "|";
            }
            first = false;

            // Escape special regex characters
            for (char c : word) {
                if (c == '.' || c == '^' || c == '$' || c == '*' || c == '+' ||
                    c == '?' || c == '(' || c == ')' || c == '[' || c == ']' ||
                    c == '{' || c == '}' || c == '|' || c == '\\') {
                    pattern += '\\';
                }
                pattern += c;
            }
        }

        pattern += ")\\b";

        return std::regex(pattern, std::regex_constants::icase);
    } catch (const std::regex_error& e) {
        LOG_ERROR << "Failed to build combined censored pattern: " << e.what();
        return std::nullopt;
    }
}

std::string CensorService::censor(const std::string& text) {
    if (text.empty()) {
        return text;
    }

    std::shared_lock<std::shared_mutex> lock(mutex_);

    if (!combinedPattern_) {
        return text;
    }

    std::string result;
    try {
        // Replace censored words with asterisks
        std::sregex_iterator begin(text.begin(), text.end(), *combinedPattern_);
        std::sregex_iterator end;

        size_t lastPos = 0;
        for (std::sregex_iterator it = begin; it != end; ++it) {
            const std::smatch& match = *it;
            // Append text before match
            result += text.substr(lastPos, match.position() - lastPos);
            // Replace match with fixed asterisks
            result += "****";
            lastPos = match.position() + match.length();
        }
        // Append remaining text
        result += text.substr(lastPos);
    } catch (const std::regex_error& e) {
        LOG_ERROR << "Regex replace error: " << e.what();
        return text;
    }

    return result;
}

bool CensorService::containsCensoredWords(const std::string& text) {
    if (text.empty()) {
        return false;
    }

    std::shared_lock<std::shared_mutex> lock(mutex_);

    if (!combinedPattern_) {
        return false;
    }

    try {
        return std::regex_search(text, *combinedPattern_);
    } catch (const std::regex_error& e) {
        LOG_ERROR << "Regex search error: " << e.what();
        return false;
    }
}

} // namespace services