fetch_ml/native/artifact_scanner/artifact_scanner.cpp

#include "artifact_scanner.h"

#include <chrono>
#include <cstring>
#include <dirent.h>
#include <fcntl.h>
#include <fnmatch.h>
#include <filesystem>
#include <sys/stat.h>
#include <unistd.h>
#include <vector>

namespace fs = std::filesystem;

struct as_scanner {
    std::vector<std::string> exclude_patterns;
    std::string last_error;
    uint64_t scan_count = 0;
};

// Check if path matches any exclude pattern
static bool should_exclude(as_scanner_t* scanner, const char* path) {
    for (const auto& pattern : scanner->exclude_patterns) {
        if (fnmatch(pattern.c_str(), path, FNM_PATHNAME) == 0) {
            return true;
        }
    }
    return false;
}

// Platform-optimized directory traversal
// Uses simple but efficient approach: batch readdir + minimal stat calls

#ifdef __linux__
    // On Linux, we could use getdents64 for even better performance
    // But standard readdir is fine for now and more portable
#endif

as_scanner_t* as_create(const char** exclude_patterns, size_t pattern_count) {
    auto* scanner = new as_scanner_t;
    
    for (size_t i = 0; i < pattern_count; ++i) {
        if (exclude_patterns[i]) {
            scanner->exclude_patterns.push_back(exclude_patterns[i]);
        }
    }
    
    // Default excludes
    scanner->exclude_patterns.push_back("run_manifest.json");
    scanner->exclude_patterns.push_back("output.log");
    scanner->exclude_patterns.push_back("code/*");
    scanner->exclude_patterns.push_back("snapshot/*");
    
    return scanner;
}

void as_destroy(as_scanner_t* scanner) {
    delete scanner;
}

void as_add_exclude(as_scanner_t* scanner, const char* pattern) {
    if (scanner && pattern) {
        scanner->exclude_patterns.push_back(pattern);
    }
}

// Fast directory scan using modern C++ filesystem (which uses optimal syscalls internally)
as_result_t* as_scan_directory(as_scanner_t* scanner, const char* run_dir) {
    if (!scanner || !run_dir) return nullptr;
    
    auto start_time = std::chrono::steady_clock::now();
    
    as_result_t* result = new as_result_t;
    result->artifacts = nullptr;
    result->count = 0;
    result->total_size = 0;
    result->discovery_time_ms = 0;
    
    std::vector<as_artifact_t> artifacts;
    artifacts.reserve(128);  // Pre-allocate to avoid reallocations
    
    try {
        fs::path root(run_dir);
        
        // Use recursive_directory_iterator with optimized options
        // skip_permission_denied prevents exceptions on permission errors
        auto options = fs::directory_options::skip_permission_denied;
        
        for (const auto& entry : fs::recursive_directory_iterator(root, options)) {
            scanner->scan_count++;
            
            if (!entry.is_regular_file()) {
                continue;
            }
            
            // Get relative path
            fs::path rel_path = fs::relative(entry.path(), root);
            std::string rel_str = rel_path.string();
            
            // Check exclusions
            if (should_exclude(scanner, rel_str.c_str())) {
                continue;
            }
            
            // Get file info
            as_artifact_t artifact;
            std::strncpy(artifact.path, rel_str.c_str(), sizeof(artifact.path) - 1);
            artifact.path[sizeof(artifact.path) - 1] = '\0';
            
            auto status = entry.status();
            artifact.size_bytes = entry.file_size();
            
            auto mtime = fs::last_write_time(entry);
            // Convert to Unix timestamp (approximate)
            auto sctp = std::chrono::time_point_cast<std::chrono::system_clock::duration>(
                mtime - fs::file_time_type::clock::now() + std::chrono::system_clock::now()
            );
            artifact.mtime = std::chrono::system_clock::to_time_t(sctp);
            
            artifact.mode = static_cast<uint32_t>(status.permissions());
            
            artifacts.push_back(artifact);
            result->total_size += artifact.size_bytes;
        }
    } catch (const std::exception& e) {
        scanner->last_error = e.what();
        delete result;
        return nullptr;
    }
    
    // Sort artifacts by path for deterministic order
    std::sort(artifacts.begin(), artifacts.end(), [](const as_artifact_t& a, const as_artifact_t& b) {
        return std::strcmp(a.path, b.path) < 0;
    });
    
    // Copy to result
    result->count = artifacts.size();
    if (result->count > 0) {
        result->artifacts = new as_artifact_t[result->count];
        std::memcpy(result->artifacts, artifacts.data(), result->count * sizeof(as_artifact_t));
    }
    
    auto end_time = std::chrono::steady_clock::now();
    result->discovery_time_ms = std::chrono::duration_cast<std::chrono::milliseconds>(end_time - start_time).count();
    
    return result;
}

void as_free_result(as_result_t* result) {
    if (result) {
        delete[] result->artifacts;
        delete result;
    }
}

const char* as_last_error(as_scanner_t* scanner) {
    if (!scanner || scanner->last_error.empty()) return nullptr;
    return scanner->last_error.c_str();
}

uint64_t as_get_scan_count(as_scanner_t* scanner) {
    return scanner ? scanner->scan_count : 0;
}