cpplogging/src/cpplogging/manager/consumer/formatting.cpp

#include <map>
#include <chrono>
#include <cstring>
#include <iomanip>

#include <cpplogging/manager/manager.h>
#include <cpplogging/manager/consumer/consumer.h>

using namespace ::cpplogging;

enum class parser_state
{
    is_text,            //!< current pos is inside normal text
    is_token_begin,     //!< current pos is in token begin (except char '$')
    is_in_token,        //!< current pos is in token (everything after '{')
    is_in_format,       //!< current pos is in format (everything after ':')
    is_in_format_align, //!< current pos is in format align (expect char '-')
    is_in_format_zero,  //!< current pos is in format align (expect char '0')
    is_in_format_dot,   //!< current pos is in format dot (expect char '.')
    is_in_format_type,  //!< current pos is in format type (expect char 's', 'f', x', 'X' or 'd')
};

enum class value_token
    : uint8_t
{
    unknown = 0,    //!< unknown
    level,          //!< log level
    uptime,         //!< system uptime
    systime,        //!< unix timestamp
    runtime,        //!< runtime of the application
    sender,         //!< sender of the log message
    sender_type,    //!< type of sender
    thread,         //!< thread id
    filename,       //!< filename only
    filepath,       //!< whole filepath
    line,           //!< line number
    name,           //!< name of the logger
    message,        //!< log message
    newline,        //!< write new line
};

enum class value_format
    : uint8_t
{
    unknown = 0,
    decimal,
    real,
    hex_lower,
    hex_upper,
    string,
    string_upper,
};

struct format_item
{
    const uint16_t  magic       { 0 };                      //!< magic value to identify the item
    value_token     token       { value_token::unknown };   //!< value to print
    value_format    format      { value_format::unknown };  //!< format to print the value with
    uint8_t         width       { 0 };                      //!< with of the value
    uint8_t         precision   { 0 };                      //!< precision (for float values)
    bool            right_align { false };                  //!< use right alignment
    bool            zero_fill   { false };                  //!< fill with zeros (if not than spaces)

    inline void reset()
        { memset(this, 0, sizeof(*this)); }

    inline std::string as_string() const
        { return std::string(reinterpret_cast<const char*>(this), sizeof(*this)); }

} __attribute__ ((packed));

static_assert(sizeof(format_item) == 8, "format_item has invalid size!");

struct op_less_invariant_string
{
    inline bool operator()(const std::string& lhs, const std::string& rhs) const
    {
        auto c1 = lhs.c_str();
        auto c2 = rhs.c_str();
        auto l1 = lhs.size();
        auto l2 = rhs.size();
        while (l1 > 0 && l2 > 0 && std::tolower(*c1) == std::tolower(*c2))
        {
            ++c1;
            ++c2;
            --l1;
            --l2;
        }
        return l1 > 0 && l2 > 0
            ? std::tolower(*c1) < std::tolower(*c2)
            : l1 < l2;
    }
};

struct invariant_string_map
    : public std::map<std::string, value_token, op_less_invariant_string>
{
    using map::map;

    inline value_token get_token(const std::string& token) const
    {
        auto it = find(token);
        return it == end()
            ? value_token::unknown
            : it->second;
    }
};

static invariant_string_map value_token_map({
    { "level",       value_token::level       },
    { "uptime",      value_token::uptime      },
    { "systime",     value_token::systime     },
    { "runtime",     value_token::runtime     },
    { "sender",      value_token::sender      },
    { "sender_type", value_token::sender_type },
    { "thread",      value_token::thread      },
    { "filename",    value_token::filename    },
    { "filepath",    value_token::filepath    },
    { "line",        value_token::line        },
    { "name",        value_token::name        },
    { "message",     value_token::message     },
    { "newline",     value_token::newline     },
});

consumer::format_type consumer::parse_format(const std::string& format)
{
    consumer::format_type ret;

    parser_state state = parser_state::is_text;
    const char * s = format.c_str();
    const char * e = s + format.size();
    const char * c = s;
    const char * x = s;
    char * tmp;
    unsigned long long int u;
    format_item item;

    while (c <= e)
    {
        switch (*c)
        {
            /* if this is the EOF */
            case '\0':
                if (s < c)
                {
                    if (!ret.empty() && !ret.back().empty() && *ret.back().c_str() != '\0')
                        ret.back() += std::string(s, static_cast<size_t>(c - s - 1));
                    else
                        ret.emplace_back(s, c - s);
                }
                break;
            /* the format string starts with $ */
            case '$':
                switch (state)
                {
                    /* if we are 'in_text', we switch to 'is_token_begin' */
                    case parser_state::is_text:
                        state = parser_state::is_token_begin;
                        break;
                    /* in an other case switch back to 'is_text' */
                    default:
                        state = parser_state::is_text;
                        item.reset();
                        break;
                }
                break;

            /* after '$' we expect '{' */
            case '{':
                switch (state)
                {
                    /* if we have seen '$' we are in 'is_token_begin' and switch to 'is_in_token' */
                    case parser_state::is_token_begin:
                        if (s + 1 < c)
                        {
                            if (!ret.empty() && !ret.back().empty() && *ret.back().c_str() != '\0')
                                ret.back() += std::string(s, static_cast<size_t>(c - s - 1));
                            else
                                ret.emplace_back(s, c - s - 1);
                        }
                        s = c - 1;
                        x = c + 1;
                        state = parser_state::is_in_token;
                        break;
                    /* in an other case switch back to 'is_text' */
                    default:
                        state = parser_state::is_text;
                        item.reset();
                        break;
                }
                break;

            /* inside the token we have ':' to seperate the format */
            case ':':
                switch (state)
                {
                    /* if we are in 'is_in_token' we extract the token and switch to 'is_in_format' */
                    case parser_state::is_in_token:
                        state = parser_state::is_in_format;
                        item.token = value_token_map.get_token(std::string(x, static_cast<size_t>(c - x)));
                        if (item.token == value_token::unknown)
                        {
                            state = parser_state::is_text;
                            item.reset();
                        }
                        x = c + 1;
                        break;
                    /* in an other case switch back to 'is_text' */
                    default:
                        state = parser_state::is_text;
                        item.reset();
                        break;
                }
                break;

            /* inside the format parsing we can have '-' to indicate right align */
            case '-':
                switch (state)
                {
                    /* if we are in 'is_in_format' we switch to 'is_in_format_align' */
                    case parser_state::is_in_format:
                        state = parser_state::is_in_format_align;
                        item.right_align = true;
                        x = c + 1;
                        break;
                    /* in an other case switch back to 'is_text' */
                    default:
                        state = parser_state::is_text;
                        item.reset();
                        break;
                }
                break;

            /* inside the format parsing we can have '0' to indicate zero fill */
            case '0':
                switch (state)
                {
                    /* if we are in state 'is_in_format' or 'is_in_format_align' we switch to 'is_in_format_zero' */
                    case parser_state::is_in_format:
                    case parser_state::is_in_format_align:
                        state = parser_state::is_in_format_zero;
                        item.zero_fill = true;
                        x = c + 1;
                        break;
                    /* if we are in state 'is_in_format_zero' we ignore zeros, because they belong to the length */
                    case parser_state::is_in_format_zero:
                        break;
                    /* in an other case switch back to 'is_text' */
                    default:
                        state = parser_state::is_text;
                        item.reset();
                        break;
                }
                break;

            /* inside the format parsing we can have '.' to indicate the precision */
            case '.':
                switch (state)
                {
                    /* if we are in state 'is_in_format', 'is_in_format_align' or 'is_in_format_zero' we switch to 'is_in_format_dot' */
                    case parser_state::is_in_format:
                    case parser_state::is_in_format_align:
                    case parser_state::is_in_format_zero:
                        u = std::strtoull(x, &tmp, 10);
                        if (x == tmp || x > c)
                        {
                            /* if parsing the with failed, we switch back to 'is_text' */
                            state = parser_state::is_text;
                            item.reset();
                        }
                        else
                        {
                            item.width = static_cast<uint8_t>(u);
                            x = c + 1;
                            state = parser_state::is_in_format_dot;
                        }
                        break;
                    /* in an other case switch back to 'is_text' */
                    default:
                        state = parser_state::is_text;
                        item.reset();
                        break;
                }
                break;

            /* if 'n' follows '$' we write a newline */
            case 'n':
                switch (state)
                {
                    case parser_state::is_token_begin:
                        item.token = value_token::newline;
                        ret.emplace_back(item.as_string());
                        item.reset();
                        state = parser_state::is_text;
                        s = c + 1;
                        break;
                    default:
                        break;
                }
                break;

            /* inside the format parsing we can have 's', 'f', 'd', 'x' or 'X' to indicate the format */
            case 's':
            case 'S':
            case 'f':
            case 'd':
            case 'x':
            case 'X':
                switch (state)
                {
                    /* if we are in state 'is_in_token' everything is fine */
                    case parser_state::is_in_token:
                        break;
                    /* if we are in state 'is_in_format', 'is_in_format_align' or 'is_in_format_zero' we switch to 'is_in_format_type' */
                    case parser_state::is_in_format:
                    case parser_state::is_in_format_align:
                    case parser_state::is_in_format_zero:
                    case parser_state::is_in_format_dot:
                        u = std::strtoull(x, &tmp, 10);
                        if (x != c && (x == tmp || x > c))
                        {
                            /* if parsing the with failed, we switch back to 'is_text' */
                            state = parser_state::is_text;
                            item.reset();
                        }
                        else
                        {
                            if (state == parser_state::is_in_format_dot)
                                item.precision = static_cast<uint8_t>(u);
                            else
                                item.width = static_cast<uint8_t>(u);

                            /* set the type and switch to 'is_in_format_type' */
                            x = c + 1;
                            state = parser_state::is_in_format_type;
                            switch(*c)
                            {
                                case 's': item.format = value_format::string;       break;
                                case 'S': item.format = value_format::string_upper; break;
                                case 'f': item.format = value_format::real;         break;
                                case 'd': item.format = value_format::decimal;      break;
                                case 'x': item.format = value_format::hex_lower;    break;
                                case 'X': item.format = value_format::hex_upper;    break;
                                default: break;
                            }
                        }
                        break;
                    /* in an other case switch back to 'is_text' */
                    default:
                        state = parser_state::is_text;
                        item.reset();
                        break;
                }
                break;

            /* inside the token/format parsing '}' indicates the end */
            case '}':
                switch (state)
                {
                    /* if we are in 'is_in_token' we extract the token, add the item and switch to 'is_text' */
                    case parser_state::is_in_token:
                        state = parser_state::is_text;
                        item.token = value_token_map.get_token(std::string(x, static_cast<size_t>(c - x)));
                        if (item.token == value_token::unknown)
                        {
                            state = parser_state::is_text;
                        }
                        else
                        {
                            ret.emplace_back(item.as_string());
                        }
                        item.reset();
                        s = c + 1;
                        break;
                    /* if we are in 'is_in_format_type' we extract the token, add the item and switch to 'is_text' */
                    case parser_state::is_in_format_type:
                        state = parser_state::is_text;
                        ret.emplace_back(item.as_string());
                        item.reset();
                        s = c + 1;
                        break;
                    /* if we are in 'is_in_format', is_in_format_zero' or 'is_in_format_align'
                       we extract the with, add the item and switch to 'is_text'*/
                    case parser_state::is_in_format:
                    case parser_state::is_in_format_zero:
                    case parser_state::is_in_format_align:
                    case parser_state::is_in_format_dot:
                        u = std::strtoull(x, &tmp, 10);
                        s = c + 1;
                        if (x != tmp && x <= c)
                        {
                            if (state == parser_state::is_in_format_dot)
                                item.precision = static_cast<uint8_t>(u);
                            else
                                item.width = static_cast<uint8_t>(u);
                            ret.emplace_back(item.as_string());
                        }
                        item.reset();
                        state = parser_state::is_text;
                        break;
                    /* in an other case switch back to 'is_text' */
                    default:
                        state = parser_state::is_text;
                        item.reset();
                        break;
                }
                break;

            /* any non parsing relevant char */
            default:
                switch (state)
                {
                    /* in case 'is_in_format', 'is_in_format_align' and 'is_in_format_zero'
                       we need to check if the current char is a digit, if not we switch to 'is_text' */
                    case parser_state::is_in_format:
                    case parser_state::is_in_format_align:
                    case parser_state::is_in_format_zero:
                        if (    *c != '.'
                            && (*c < '0' || *c > '9'))
                        {
                            state = parser_state::is_text;
                            item.reset();
                        }
                        break;
                    /* any other case is fine */
                    default:
                        break;
                }
                break;
        }
        ++c;
    }

    return ret;
}

static const auto start_time = std::chrono::steady_clock::now();

void consumer::write_formatted(const log_entry& e, const format_type& f, std::ostream& os)
{
    using duration_f = std::chrono::duration<double,                 std::ratio<1>>;
    using duration_u = std::chrono::duration<unsigned long long int, std::ratio<1>>;

    static const std::string LogLevelUpperDebug("DEBUG", 5);
    static const std::string LogLevelUpperInfo ("INFO",  4);
    static const std::string LogLevelUpperWarn ("WARN",  4);
    static const std::string LogLevelUpperError("ERROR", 5);

    static const std::string LogLevelLowerDebug("debug", 5);
    static const std::string LogLevelLowerInfo ("info",  4);
    static const std::string LogLevelLowerWarn ("warn",  4);
    static const std::string LogLevelLowerError("error", 5);

    bool has_line_end = false;
    for (auto& x : f)
    {
        /* check if element is empty */
        if (x.empty())
        {
            continue;
        }

        /* check if element is normal item */
        if (*x.c_str() != '\0')
        {
            os << x;
            continue;
        }

        /* setup the stream format */
        const char * s;
        auto& i = *reinterpret_cast<const format_item*>(x.c_str());
        os  << (i.right_align ? std::right : std::left)
            << std::setfill(i.zero_fill ? '0' : ' ')
            << std::setw(i.width)
            << std::setprecision(i.precision);
        switch (i.token)
        {
            /* level */
            case value_token::level:
                has_line_end = false;
                switch (i.format)
                {
                    case value_format::decimal:
                        os << static_cast<int>(e.level);
                        break;
                    case value_format::string_upper:
                        switch (e.level)
                        {
                            case log_level::debug: os << LogLevelUpperDebug; break;
                            case log_level::info:  os << LogLevelUpperInfo;  break;
                            case log_level::warn:  os << LogLevelUpperWarn;  break;
                            case log_level::error: os << LogLevelUpperError; break;
                            default: break;
                        }
                        break;
                    default:
                        switch (e.level)
                        {
                            case log_level::debug: os << LogLevelLowerDebug; break;
                            case log_level::info:  os << LogLevelLowerInfo;  break;
                            case log_level::warn:  os << LogLevelLowerWarn;  break;
                            case log_level::error: os << LogLevelLowerError; break;
                            default: break;
                        }
                        break;
                }
                break;

            /* uptime */
            case value_token::uptime:
                has_line_end = false;
                switch (i.format)
                {
                    case value_format::decimal:
                        os  << std::chrono::duration_cast<duration_u>(e.uptime.time_since_epoch()).count();
                        break;
                    default:
                        os  << std::fixed
                            << std::chrono::duration_cast<duration_f>(e.uptime.time_since_epoch()).count()
                            << std::dec;
                        break;
                }
                break;

            /* runtime */
            case value_token::runtime:
                has_line_end = false;
                switch (i.format)
                {
                    case value_format::decimal:
                        os  << std::chrono::duration_cast<duration_u>(e.uptime - start_time).count();
                        break;
                    default:
                        os  << std::fixed
                            << std::chrono::duration_cast<duration_f>(e.uptime - start_time).count()
                            << std::dec;
                        break;
                }
                break;

            /* systemtime */
            case value_token::systime:
                has_line_end = false;
                switch (i.format)
                {
                    case value_format::decimal:
                        os  << e.systime;
                        break;
                    default:
                        os  << std::fixed
                            << e.systime
                            << std::dec;
                        break;
                }
                break;

            /* sender */
            case value_token::sender:
                has_line_end = false;
                switch (i.format)
                {
                    case value_format::decimal:
                        os  << reinterpret_cast<uintptr_t>(e.sender);
                        break;
                    case value_format::hex_lower:
                        os  << std::hex
                            << reinterpret_cast<uintptr_t>(e.sender)
                            << std::dec;
                        break;
                    default:
                        os  << std::hex
                            << std::uppercase
                            << reinterpret_cast<uintptr_t>(e.sender)
                            << std::nouppercase
                            << std::dec;
                        break;
                }
                break;

            /* sender_type */
            case value_token::sender_type:
                has_line_end = false;
                os << e.sender_type;
                break;

            /* thread */
            case value_token::thread:
                has_line_end = false;
                switch (i.format)
                {
                    case value_format::decimal:
                        os  << e.thread;
                        break;
                    case value_format::hex_lower:
                        os  << std::hex
                            << e.thread
                            << std::dec;
                        break;
                    default:
                        os  << std::hex
                            << std::uppercase
                            << e.thread
                            << std::nouppercase
                            << std::dec;
                        break;
                }
                break;

            /* filename */
            case value_token::filename:
                has_line_end = false;
                s = e.file;
                if (s)
                {
                    auto tmp = strrchr(s, '/');
                    if (tmp)
                        s = tmp + 1;
                }
                else
                    s = "unknown";
                os << s;
                break;

            /* filepath */
            case value_token::filepath:
                has_line_end = false;
                os << (e.file ? e.file : "unknown");
                break;

            /* line */
            case value_token::line:
                has_line_end = false;
                switch (i.format)
                {
                    case value_format::hex_lower:
                        os  << std::hex
                            << e.line
                            << std::dec;
                        break;
                    case value_format::hex_upper:
                        os  << std::hex
                            << std::uppercase
                            << e.line
                            << std::nouppercase
                            << std::dec;
                        break;
                    default:
                        os  << e.line;
                        break;
                }
                break;

            /* name */
            case value_token::name:
                if (!e.name.empty())
                {
                    os << e.name;
                    has_line_end = (e.name.back() == '\n');
                }
                break;

            /* message */
            case value_token::message:
                if (!e.message.empty())
                {
                    os << e.message;
                    has_line_end = (e.message.back() == '\n');
                }
                break;

            /* newline */
            case value_token::newline:
                has_line_end = true;
                os  << std::endl;
                break;

            default:
                break;
        }
    }

    if (!has_line_end)
        os << std::endl;
}