ule/serialize.cpp

#include <fast_float/fast_float.h>

#include "types.h"
#include "serialize.h"
#include "string.h"
#include "print.h"


static inline const char* getFormatStringOut(u8     v) { TYPES_H_FTAG; return "%hu\n"; }
static inline const char* getFormatStringOut(u16    v) { TYPES_H_FTAG; return "%hu\n"; }
static inline const char* getFormatStringOut(u32    v) { TYPES_H_FTAG; return "%u\n"; }
static inline const char* getFormatStringOut(u64    v) { TYPES_H_FTAG; return "%llu\n"; }

static inline const char* getFormatStringOut(s8     v) { TYPES_H_FTAG; return "%hd\n"; }
static inline const char* getFormatStringOut(s16    v) { TYPES_H_FTAG; return "%hd\n"; }
static inline const char* getFormatStringOut(s32    v) { TYPES_H_FTAG; return "%d\n"; }
static inline const char* getFormatStringOut(s64    v) { TYPES_H_FTAG; return "%lld\n"; }

static inline const char* getFormatStringOut(float  v) { TYPES_H_FTAG; return "%f\n"; }
static inline const char* getFormatStringOut(double v) { TYPES_H_FTAG; return "%f\n"; }

// important constraint - strings need to be wrapped in double-quotes.
// the sentinel value 'null' without quotations is used to denote null values, which means
// if strings were not wrapped in double quotes, you would not be able to distinguish null
// values from the literal string "null".
static inline const char* getFormatStringOut(char*       v) { TYPES_H_FTAG; return "\"%s\"\n"; }
static inline const char* getFormatStringOut(const char* v) { TYPES_H_FTAG; return "\"%s\"\n"; }

#ifdef _USING_GLM_TYPES__
static inline const char* getFormatStringOut(glm::vec<2, float, (glm::qualifier) 3> v) { TYPES_H_FTAG; return "%f %f\n"; }
static inline const char* getFormatStringOut(glm::vec<3, float, (glm::qualifier) 3> v) { TYPES_H_FTAG; return "%f %f %f\n"; }
static inline const char* getFormatStringOut(glm::vec<4, float, (glm::qualifier) 3> v) { TYPES_H_FTAG; return "%f %f %f %f\n"; }

static inline const char* getFormatStringOut(glm::mat<2, 2, float, (glm::qualifier) 3> v) { TYPES_H_FTAG; return "%f %f %f %f\n"; }
static inline const char* getFormatStringOut(glm::mat<3, 3, float, (glm::qualifier) 3> v) { TYPES_H_FTAG; return "%f %f %f %f %f %f %f %f %f\n"; }
static inline const char* getFormatStringOut(glm::mat<4, 4, float, (glm::qualifier) 3> v) { TYPES_H_FTAG; return "%f %f %f %f %f %f %f %f %f %f %f %f %f %f %f %f\n"; }

#endif

#define SERIALIZE_H_FUNC_BODY str->appendf(getFormatStringOut(v), v);
void serialize(String* str, u8     v) { TYPES_H_FTAG; SERIALIZE_H_FUNC_BODY }
void serialize(String* str, u16    v) { TYPES_H_FTAG; SERIALIZE_H_FUNC_BODY }
void serialize(String* str, u32    v) { TYPES_H_FTAG; SERIALIZE_H_FUNC_BODY }
void serialize(String* str, u64    v) { TYPES_H_FTAG; SERIALIZE_H_FUNC_BODY }
void serialize(String* str, s8     v) { TYPES_H_FTAG; SERIALIZE_H_FUNC_BODY }
void serialize(String* str, s16    v) { TYPES_H_FTAG; SERIALIZE_H_FUNC_BODY }
void serialize(String* str, s32    v) { TYPES_H_FTAG; SERIALIZE_H_FUNC_BODY }
void serialize(String* str, s64    v) { TYPES_H_FTAG; SERIALIZE_H_FUNC_BODY }
void serialize(String* str, float  v) { TYPES_H_FTAG; SERIALIZE_H_FUNC_BODY }
void serialize(String* str, double v) { TYPES_H_FTAG; SERIALIZE_H_FUNC_BODY }

template<typename T> // do I really need a template for this?
static inline void deserializeInteger(char** buffer, T* v) {
    TYPES_H_FTAG;
    char* _buffer = *buffer;
    T value = 0;
    
    // skip leading whitespace - all our functions do this, so gotta
    while (String::isAsciiWhitespace(*_buffer)) {
        _buffer++; 
    }
    
    // check the first character for a negative sign
    bool negative = false;
    if (_buffer[0] == '-') {
        negative = true;
        _buffer++;
    }

    // parse the digits of the number. doesn't account for overflow
    while (String::isDigit(*_buffer)) {
        value = 10 * value + (*_buffer++ - '0');
    }

    // make it negative if the first character was '-'
    if (negative) value *= -1;

    // store the value back into v
    *v = value;

    // if the original pointer is the same as the current, we didn't parse anything
    massert(_buffer != *buffer, "tried to parse an integer, and found nothing");

    // report back how far we advanced the buffer
    *buffer = _buffer;
}

// just for integers, signed/unsigned including size_t
#define SERIALIZE_H_DESERIALIZE_FUNC_BODY deserializeInteger(buffer, v);
void deserialize(char** buffer, u8*     v) { SERIALIZE_H_DESERIALIZE_FUNC_BODY }
void deserialize(char** buffer, u16*    v) { SERIALIZE_H_DESERIALIZE_FUNC_BODY }
void deserialize(char** buffer, u32*    v) { SERIALIZE_H_DESERIALIZE_FUNC_BODY }
void deserialize(char** buffer, u64*    v) { SERIALIZE_H_DESERIALIZE_FUNC_BODY }
void deserialize(char** buffer, s8*     v) { SERIALIZE_H_DESERIALIZE_FUNC_BODY }
void deserialize(char** buffer, s16*    v) { SERIALIZE_H_DESERIALIZE_FUNC_BODY }
void deserialize(char** buffer, s32*    v) { SERIALIZE_H_DESERIALIZE_FUNC_BODY }
void deserialize(char** buffer, s64*    v) { SERIALIZE_H_DESERIALIZE_FUNC_BODY }

// base value = ceil(log10(2^64)) or ceil(log10(2^32))
// is this right? lemire's paper mentions you only need 17 digits for 64 bit floats: https://lemire.me/en/publication/arxiv2101/
// +1 for 'e' 
// +1 for '.' 
// +1 for '^'
// +1 for '-'/'+'
static const u32 BINARY32_MAX_CHARS = 14;
static const u32 BINARY64_MAX_CHARS = 24;

void deserialize(char** buffer, float*  v) { 
    TYPES_H_FTAG; 
    char* _buffer = *buffer;
    while (String::isAsciiWhitespace(*_buffer)) _buffer++; 
    fast_float::from_chars_result result = fast_float::from_chars(_buffer, _buffer + BINARY32_MAX_CHARS, *v);
    massert(result.ec == std::errc(), "failed to parse a float");
    *buffer = (char*) result.ptr;
}
void deserialize(char** buffer, double* v) { 
    TYPES_H_FTAG;  
    char* _buffer = *buffer;
    while (String::isAsciiWhitespace(*_buffer)) _buffer++; 
    fast_float::from_chars_result result = fast_float::from_chars(_buffer, _buffer + BINARY64_MAX_CHARS, *v);
    massert(result.ec == std::errc(), "failed to parse a double");
    *buffer = (char*) result.ptr;
}

#ifndef _WIN32
// win32 doesn't treat size_t as different than a u64, which causes ambiguous function calls
static inline const char* getFormatStringOut(size_t v) { TYPES_H_FTAG; return "%lu\n"; }
void serialize(String* str, size_t v) { SERIALIZE_H_FUNC_BODY }
void deserialize(char** buffer, size_t* v) { SERIALIZE_H_DESERIALIZE_FUNC_BODY }
#endif

// STRING STUFF
void serialize(String* str, char* v) {
    TYPES_H_FTAG; 
    SERIALIZE_HANDLE_NULL(str, v);
    SERIALIZE_H_FUNC_BODY;
}
void serialize(String* str, const char* v) {
    TYPES_H_FTAG; 
    SERIALIZE_HANDLE_NULL(str, v);
    SERIALIZE_H_FUNC_BODY;
}

#define SERIALIZE_SCRATCH_BUFFER_SIZE 2048
// make this number be the largest number of bytes you can expect to deserialize when deserializing a single string, in the case when the memory allocation must be done by the deserialization code.
// meaning, you don't have a pre-allocated buffer of a fixed, known size, like a char name[32] or something.
//
// if you call `deserialize(char** buffer, char* v)`, it's expected that you know what you're doing,
// and the |buffer| won't contain more characters than you can fit in |v|.
//
// in the case where you call `deserialize(char** buffer, char** v)`, it's assumed the size of the string
// is mostly unknown, but some buffer still must be allocated in order to be filled with the deserialized data.
// we use the SERIALIZE_SCRATCH_BUFFER for this duty, and Strcpy out of it once done to avoid wasting space.
//
// the ideal size of this buffer is use-case dependent, and you will have to change that yourself.
//
// there are asserts in the codepaths that use it for overflowing it, if you are unsure.
//
static char SERIALIZE_SCRATCH_BUFFER[SERIALIZE_SCRATCH_BUFFER_SIZE];

static s32 deserializeString(char** buffer, char* v, s32 vSize) {
    TYPES_H_FTAG; 
    char* _buffer = *buffer;

    while (String::isAsciiWhitespace(*_buffer)) _buffer++; 

    massert(_buffer[0] == '"', "expecting to deserialize a string, but found something other than a double quote");
    _buffer++; // skip over the first quote

    s32 i = 0;
    while (_buffer[i] != '"') {
        // @TODO handle escaped quotes?
        v[i] = _buffer[i];
        i++;
        massert(i <= (vSize - 1), "deserializing a string bigger than the reported size of the buffer we are trying to fill!");
    }

    v[i] = '\0';
    *buffer = _buffer + i + 1; // +i for the number of chars in string, +1 for final quotemark
    return i;
}
static s32 deserializeString(char** buffer, char* v) {
    TYPES_H_FTAG; 
    char* _buffer = *buffer;
    while (String::isAsciiWhitespace(*_buffer)) _buffer++;
    massert(_buffer[0] == '"', "expecting to deserialize a string, but found something other than a double quote");
    _buffer++; // skip over the first quote
    
    s32 i = 0;
    while (_buffer[i] != '"') {
        // @TODO handle escaped quotes?
        v[i] = _buffer[i];
        i++;
    }
   
    v[i] = '\0';
    *buffer = _buffer + i + 1; // +i for the number of chars in string, +1 for final quotemark
    return i;
}
void deserialize(char** buffer, char* v) {
    TYPES_H_FTAG; 
    deserializeString(buffer, v);
}
void deserialize(char** buffer, const char* v) {
    TYPES_H_FTAG; 
    deserializeString(buffer, (char*) v);
}
void deserialize(char** buffer, char** v) {
    TYPES_H_FTAG; 
    
    DESERIALIZE_HANDLE_NULL(buffer, v);
    
    s32 i = deserializeString(buffer, SERIALIZE_SCRATCH_BUFFER, SERIALIZE_SCRATCH_BUFFER_SIZE);
    massert(i >= 0, "didn't deserialize anything when expecting a string!");
    *v = String::cpy(SERIALIZE_SCRATCH_BUFFER, (u32) i);
}
void deserialize(char** buffer, const char** v) {
    TYPES_H_FTAG; 
    
    DESERIALIZE_HANDLE_NULL(buffer, (char*) v); // error: readonly variable is not assignable
    
    s32 i = deserializeString(buffer, SERIALIZE_SCRATCH_BUFFER, SERIALIZE_SCRATCH_BUFFER_SIZE);
    massert(i >= 0, "didn't deserialize anything when expecting a string!");
    *v = String::cpy(SERIALIZE_SCRATCH_BUFFER, (u32) i);
}

#ifdef _USING_GLM_TYPES__
// I have no fucking idea why, but the declarations of glm types here resolve to a type,
// that has a template parameter == 0, but all other instances of those types in my program
// have that template parameter == 3, so everything below becomes unresolved symbols if 
// I don't do the nasty template garbage here
void serialize(String* str, glm::vec<2, float, (glm::qualifier) (glm::qualifier) 3> v) { 
    TYPES_H_FTAG; 
    str->appendf(getFormatStringOut(v), v[0], v[1]); 
}
void serialize(String* str, glm::vec<3, float, (glm::qualifier) (glm::qualifier) 3> v) { 
    TYPES_H_FTAG; 
    str->appendf(getFormatStringOut(v), v[0], v[1], v[2]); 
}
void serialize(String* str, glm::vec<4, float, (glm::qualifier) 3> v) { 
    TYPES_H_FTAG; 
    str->appendf(getFormatStringOut(v), v[0], v[1], v[2], v[3]); 
}

void serialize(String* str, glm::mat<2, 2, float, (glm::qualifier) 3> v) {
    TYPES_H_FTAG; 
    str->appendf(getFormatStringOut(v)
        , v[0][0], v[0][1]
        , v[1][0], v[1][1]);
}
void serialize(String* str, glm::mat<3, 3, float, (glm::qualifier) 3> v) {
    TYPES_H_FTAG; 
    str->appendf(getFormatStringOut(v)
        , v[0][0], v[0][1], v[0][2]
        , v[1][0], v[1][1], v[1][2]
        , v[2][0], v[2][1], v[2][2]);
}
void serialize(String* str, glm::mat<4, 4, float, (glm::qualifier) 3> v) {
    TYPES_H_FTAG; 
    str->appendf(getFormatStringOut(v)
        , v[0][0], v[0][1], v[0][2], v[0][3]
        , v[1][0], v[1][1], v[1][2], v[1][3]
        , v[2][0], v[2][1], v[2][2], v[2][3]
        , v[3][0], v[3][1], v[3][2], v[3][3]);
}

void deserialize(char** buffer, glm::vec<2, float, (glm::qualifier) 3>* v) {
    TYPES_H_FTAG; 
    float* _v = (float*) v;
    for (u32 i = 0; i < 2; i++) {
        deserialize(buffer, _v + i);
    }
}
void deserialize(char** buffer, glm::vec<3, float, (glm::qualifier) 3>* v) {
    TYPES_H_FTAG; 
    float* _v = (float*) v;
    for (u32 i = 0; i < 3; i++) {
        deserialize(buffer, _v + i);
    }
}
void deserialize(char** buffer, glm::vec<4, float, (glm::qualifier) 3>* v) {
    TYPES_H_FTAG; 
    float* _v = (float*) v;
    for (u32 i = 0; i < 4; i++) {
        deserialize(buffer, _v + i);
    }
}

void deserialize(char** buffer, glm::mat<2, 2, float, (glm::qualifier) 3>* v) {
    TYPES_H_FTAG; 
    float* m = (float*) v;
    for (u32 i = 0; i < 4; i++) {
        deserialize(buffer, m + i);
    }
}
void deserialize(char** buffer, glm::mat<3, 3, float, (glm::qualifier) 3>* v) {
    TYPES_H_FTAG; 
    float* m = (float*) v;
    for (u32 i = 0; i < 9; i++) {
        deserialize(buffer, m + i);
    }
}
void deserialize(char** buffer, glm::mat<4, 4, float, (glm::qualifier) 3>* v) {
    TYPES_H_FTAG; 
    float* m = (float*) v;
    for (u32 i = 0; i < 16; i++) {
        deserialize(buffer, m + i);
    }
}

#undef SERIALIZE_H_FUNC_BODY
#undef SERIALIZE_H_DESERIALIZE_FUNC_BODY

#endif