cell/source/nota.h

#ifndef NOTA_H
#define NOTA_H

#define NOTA_BLOB 0x00 // C 0 0 0
#define NOTA_TEXT 0x10 // C 0 0 1
#define NOTA_ARR 0x20 // C 0 1 0
#define NOTA_REC 0x30 // C 0 1 1
#define NOTA_FLOAT 0x40 // C 1 0
#define NOTA_INT 0x60 // C 1 1 0
#define NOTA_SYM 0x70 // C 1 1 1

#define NOTA_NULL 0x00
#define NOTA_FALSE 0x02
#define NOTA_TRUE 0x03
#define NOTA_INF 0x03
#define NOTA_PRIVATE 0x08
#define NOTA_SYSTEM 0x09

// Returns the type NOTA_ of the byte at *nota
int nota_type(char *nota);

// Functions take a pointer to a buffer *nota, read or write the value, and then return a pointer to the next byte of the stream

// Pass NULL into the read in variable to skip over it

char *nota_read_blob(long long *len, char **blob, char *nota);
// ALLOCATES! Uses strdup to return it via the text pointer
char *nota_read_text(char **text, char *nota);
char *nota_read_array(long long *len, char *nota);
char *nota_read_record(long long *len, char *nota);
char *nota_read_float(double *d, char *nota);
char *nota_read_int(long long *l, char *nota);
char *nota_read_sym(int *sym, char *nota);

char *nota_write_blob(unsigned long long n, char *data, char *nota);
char *nota_write_text(const char *s, char *nota);
char *nota_write_array(unsigned long long n, char *nota);
char *nota_write_record(unsigned long long n, char *nota);
char *nota_write_number(double n, char *nota);
char *nota_write_sym(int sym, char *nota);

void print_nota_hex(char *nota);

#ifdef NOTA_IMPLEMENTATION

#include "stdio.h"
#include "math.h"
#include "string.h"
#include "stdlib.h"
#include "limits.h"
#include "kim.h"

#define NOTA_CONT 0x80
#define NOTA_DATA 0x7f
#define NOTA_INT_DATA 0x07
#define NOTA_INT_SIGN(CHAR) (CHAR & (1<<3))
#define NOTA_SIG_SIGN(CHAR) (CHAR & (1<<3))
#define NOTA_EXP_SIGN(CHAR) (CHAR & (1<<4))
#define NOTA_TYPE 0x70
#define NOTA_HEAD_DATA 0x0f
#define CONTINUE(CHAR) (CHAR>>7)

#define UTF8_DATA 0x3f

/* define this to use native string instead of kim. Bytes are encoded instead of runes */
#define NOTA_UTF8

int nota_type(char *nota) { return *nota & NOTA_TYPE; }

char *nota_skip(char *nota)
{
  while (CONTINUE(*nota))
    nota++;

  return nota+1;
}

char *nota_read_num(long long *n, char *nota)
{
  if (!n)
    return nota_skip(nota);
    
  *n = 0;
  *n |= (*nota) & NOTA_HEAD_DATA;
  
  while (CONTINUE(*(nota++)))
    *n = (((*n<<7) | *nota) & NOTA_DATA);

  return nota;
}

// Given a number n, and bits used in the first char sb, how many bits are needed
int nota_bits(long long n, int sb)
{
  if (n == 0) return sb;
  int bits = sizeof(n)*CHAR_BIT - __builtin_clzll(n);
  bits-=sb;
  int needed = ((bits + 6) / 7)*7 + sb;
  return needed;
}

// write a number from n into *nota, with sb bits in the first char
char *nota_continue_num(long long n, char *nota, int sb)
{
  int bits = nota_bits(n, sb);
  bits -= sb;
  if (bits > 0)
    nota[0] |= NOTA_CONT;
  else
    nota[0] &= ~NOTA_CONT;

  int shex = (~0) << sb;
  nota[0] &= shex; /* clear shex bits */
  nota[0] |= (~shex) & (n>>bits);

  int i = 1;
  while (bits > 0) {
    bits -= 7;
    int head = bits == 0 ? 0 : NOTA_CONT;
    nota[i] = head | (NOTA_DATA & (n >> bits));
    i++;
  }

  return &nota[i];
}


void print_nota_hex(char *nota)
{
  do {
    printf("%02X ", (unsigned char)(*nota));
  } while(CONTINUE(*(nota++)));
  printf("\n");
  
  return;
  long long chars = 0;
  if (!((*nota>>4 & 0x07) ^ NOTA_TEXT>>4))
    nota_read_num(&chars, nota);

  if ((*nota>>5) == 2 || (*nota>>5) == 6)
    chars = 1;

  for (int i = 0; i < chars+1; i++) {
    do {
      printf("%02X ", (unsigned char)(*nota));
    } while(CONTINUE(*(nota++)));
  }
    
  printf("\n");
}

char *nota_write_int(long long n, char *nota)
{
  char sign = 0;
  
  if (n < 0) {
    sign = 0x08;
    n *= -1;
  }
  
  *nota = NOTA_INT | sign;  

  return nota_continue_num(n, nota, 3);
}

#define NOTA_DBL_PREC 6
#define xstr(s) str(s)
#define str(s) #s

#include <stdio.h>
#include <math.h>

void extract_mantissa_coefficient(double num, long *coefficient, long *exponent)
{
  if (num == 0.0) {
    *coefficient = 0;
    *exponent = 0;
    return;
  }

    // Optional: handle sign separately if you want 'coefficient' always positive.
    // For simplicity, let's just let atol(...) parse the sign if it's there.
    
    // 1) Slightly round the number to avoid too many FP trailing digits:
    //    Example: Round to 12 decimal places.
  double rounded = floor(fabs(num) * 1e12 + 0.5) / 1e12;
  if (num < 0) {
    rounded = -rounded;
  }

    // 2) Convert to string with fewer digits of precision so we do NOT get
    //    the long binary-fraction expansions (like 98.599999999999994).
    char buf[64];
    snprintf(buf, sizeof(buf), "%.14g", rounded);

    // 3) Look for scientific notation
    char *exp_pos = strpbrk(buf, "eE");
    long exp_from_sci = 0;
    if (exp_pos) {
        exp_from_sci = atol(exp_pos + 1);
        *exp_pos = '\0'; // Truncate the exponent part from the string
    }

    // 4) Find decimal point
    char *dec_point = strchr(buf, '.');
    int digits_after_decimal = 0;

    if (dec_point) {
        digits_after_decimal = (int)strlen(dec_point + 1);
        // Remove the '.' by shifting the remainder left
        memmove(dec_point, dec_point + 1, strlen(dec_point));
    }

    // 5) Now the string is just an integer (possibly signed), so parse it
    long long coeff_ll = atoll(buf);  // support up to 64-bit range
    *coefficient = (long)coeff_ll;

    // 6) The final decimal exponent is whatever came from 'e/E'
    //    minus however many digits we removed by removing the decimal point.
    *exponent = exp_from_sci - digits_after_decimal;
}

char *nota_write_decimal_str(const char *decimal, char *nota)
{
    // Handle negative sign
    int neg = (decimal[0] == '-');
    if (neg) decimal++; // Skip the '-' if present

    // Parse integer part
    long coef = 0;
    long exp = 0;
    int decimal_point_seen = 0;
    const char *ptr = decimal;
    int has_exponent = 0;

    // First pass: calculate coefficient up to 'e' or 'E'
    while (*ptr && *ptr != 'e' && *ptr != 'E') {
        if (*ptr == '.') {
            decimal_point_seen = 1;
            ptr++;
            continue;
        }
        
        if (*ptr >= '0' && *ptr <= '9') {
            coef = coef * 10 + (*ptr - '0');
            if (decimal_point_seen) {
                exp--; // Each digit after decimal point decreases exponent
            }
        }
        ptr++;
    }

    // Parse exponent part if present
    if (*ptr == 'e' || *ptr == 'E') {
        has_exponent = 1;
        ptr++; // Skip 'e' or 'E'
        
        int exp_sign = 1;
        if (*ptr == '-') {
            exp_sign = -1;
            ptr++;
        } else if (*ptr == '+') {
            ptr++;
        }
        
        long explicit_exp = 0;
        while (*ptr >= '0' && *ptr <= '9') {
            explicit_exp = explicit_exp * 10 + (*ptr - '0');
            ptr++;
        }
        exp += exp_sign * explicit_exp;
    }

    // If no decimal point and no exponent, treat as integer
    if (!decimal_point_seen && !has_exponent) {
        return nota_write_int(coef * (neg ? -1 : 1), nota);
    }

    // Remove trailing zeros from coefficient
    while (coef > 0 && coef % 10 == 0 && exp < 0) {
        coef /= 10;
        exp++;
    }

    // Handle zero case
    if (coef == 0) {
        return nota_write_int(0, nota);
    }

    // Set up the notation format similar to nota_write_float
    int expsign = exp < 0 ? ~0 : 0;
    exp = llabs(exp);

    nota[0] = NOTA_FLOAT;
    nota[0] |= (expsign & 1) << 4;  // Exponent sign bit
    nota[0] |= (neg & 1) << 3;     // Number sign bit
    
    char *c = nota_continue_num(exp, nota, 3);
    return nota_continue_num(coef, c, 7);
}

char *nota_write_float(double n, char *nota)
{
  int neg = (n < 0);
  long coef;
  long exp;
  extract_mantissa_coefficient(n, &coef, &exp);

  // Store integer if exponent is zero
  if (exp == 0)
    return nota_write_int(coef * (neg ? -1 : 1), nota);

  int expsign = exp < 0 ? ~0 : 0;
  exp = labs(exp);

  nota[0] = NOTA_FLOAT;
  nota[0] |= (expsign & 1) << 4;
  nota[0] |= (neg & 1) << 3;

  char *c = nota_continue_num(exp, nota, 3);
  return nota_continue_num(labs(coef), c, 7);
}

char *nota_read_float_str(char **d, char *nota)
{
    // Extract sign bits from the first byte
    int neg   = NOTA_SIG_SIGN(*nota);  // bit 3 => mantissa sign
    int esign = NOTA_EXP_SIGN(*nota);  // bit 4 => exponent sign

    // Read the exponent’s lower 3 bits from the first byte
    long long e = (*nota) & NOTA_INT_DATA;  // NOTA_INT_DATA = 0x07

    // Count exponent bytes and accumulate value
    int e_bytes = 1;
    while (CONTINUE(*nota)) {
        nota++;
        e = (e << 7) | ((*nota) & NOTA_DATA); // NOTA_DATA = 0x7F
        e_bytes++;
    }

    // Move past the last exponent byte
    nota++;

    // Read the mantissa
    long long sig = (*nota) & NOTA_DATA;
    int sig_bytes = 1;
    while (CONTINUE(*nota)) {
        nota++;
        sig = (sig << 7) | ((*nota) & NOTA_DATA);
        sig_bytes++;
    }

    // Move past the last mantissa byte
    nota++;

    // Apply sign bits
    if (neg)   sig = -sig;
    if (esign) e   = -e;

    // Calculate digits in mantissa (sig) and exponent (e)
    int sig_digits = (sig == 0) ? 1 : (int)log10(llabs(sig)) + 1;
    int e_digits = (e == 0) ? 1 : (int)log10(llabs(e)) + 1;

    // Calculate total string size:
    // - Mantissa: sign (1), digits, decimal (1), 2 decimal places
    // - Exponent: 'e', sign (1), digits
    // - Null terminator (1)
    int size = 1 + sig_digits + 1 + 2 + 1 + 1 + e_digits + 1;
    if (neg) size++;  // Extra space for negative mantissa
    if (esign) size++; // Extra space for negative exponent

    // Allocate the string
    char *result = (char *)malloc(size);
    if (!result) {
        *d = NULL;
        return nota; // Return current position even on failure
    }

    // Format the string as "xey" (e.g., "1.23e4")
    double value = (double)sig * pow(10.0, (double)e);
    snprintf(result, size, "%.*fe%lld", 2, value/pow(10.0, (double)e), e);

    // Set the output pointer and return
    *d = result;
    return nota;
}

char *nota_read_float(double *d, char *nota)
{
    // If the caller passed NULL for d, just skip over the float encoding
    if (!d) {
        return nota_skip(nota);
    }

    // Extract sign bits from the first byte
    int neg   = NOTA_SIG_SIGN(*nota);  // bit 3 => mantissa sign
    int esign = NOTA_EXP_SIGN(*nota);  // bit 4 => exponent sign

    // Read the exponent’s lower 3 bits from the first byte
    long long e = (*nota) & NOTA_INT_DATA;  // NOTA_INT_DATA = 0x07

    // While the continuation bit is set, advance and accumulate exponent
    while (CONTINUE(*nota)) {
        nota++;
        e = (e << 7) | ((*nota) & NOTA_DATA); // NOTA_DATA = 0x7F
    }

    // Move past the last exponent byte
    nota++;

    // Now read the mantissa in the same variable-length style
    long long sig = (*nota) & NOTA_DATA;
    while (CONTINUE(*nota)) {
        nota++;
        sig = (sig << 7) | ((*nota) & NOTA_DATA);
    }

    // Move past the last mantissa byte
    nota++;

    // Apply sign bits
    if (neg)   sig = -sig;
    if (esign) e   = -e;

    // Finally compute the double value: mantissa * 10^exponent
    *d = (double)sig * pow(10.0, (double)e);

    // Return the pointer to wherever we ended
    return nota;
}

char *nota_write_number(double n, char *nota)
{
  if (n < (double)INT64_MIN || n > (double)INT64_MAX) return nota_write_float(n, nota);

  double int_part;
  double frac = modf(n, &int_part);

  if (fabs(frac) < 1e-14)
    return nota_write_int((long long)int_part, nota);
  else
    return nota_write_float(n, nota);
}

char *nota_read_int(long long *n, char *nota)
{
  if (!n)
    return nota_skip(nota);

  *n = 0;
  char *c = nota;
  *n |= (*c) & NOTA_INT_DATA; /* first three bits */
  while (CONTINUE(*(c++)))
    *n = (*n<<7) | (*c & NOTA_DATA);

  if (NOTA_INT_SIGN(*nota)) *n *= -1;

  return c;
}

/* n is the number of bits */
char *nota_write_blob(unsigned long long n, char *data, char *nota)
{
  nota[0] = NOTA_BLOB;
  nota = nota_continue_num(n, nota, 4);
  int bytes = floor((n+7)/8);
  for (int i = 0; i < bytes; i++)
    nota[i] = data[i];
    
  return nota+bytes;
}

char *nota_write_array(unsigned long long n, char *nota)
{
  nota[0] = NOTA_ARR;
  return nota_continue_num(n, nota, 4);
}

char *nota_read_array(long long *len, char *nota)
{
  if (!len) return nota;
  return nota_read_num(len, nota);
}

char *nota_read_record(long long *len, char *nota)
{
  if (!len) return nota;
  return nota_read_num(len, nota);
}

char *nota_read_blob(long long *len, char **blob, char *nota)
{
  if (!len) return nota;
  nota = nota_read_num(len,nota);
  int bytes = floor((*len+7)/8);
  *len = bytes;

  *blob = malloc(bytes);
  memcpy(*blob,nota,bytes);
  
  return nota+bytes;
}

char *nota_write_record(unsigned long long n, char *nota)
{
  nota[0] = NOTA_REC;
  return nota_continue_num(n, nota, 4);
}

char *nota_write_sym(int sym, char *nota)
{
  *nota = NOTA_SYM | sym;
  return nota+1;
}

char *nota_read_sym(int *sym, char *nota)
{
  if (sym) *sym = (*nota) & 0x0f;
  return nota+1;
}

char *nota_read_text(char **text, char *nota)
{
  long long chars;
  nota = nota_read_num(&chars, nota);

  char utf[chars*4]; // enough for the worst case scenario
  char *pp = utf;
  kim_to_utf8(&nota, &pp, chars);
  *pp = 0;
  *text = strdup(utf);  
  
  return nota;
}

char *nota_write_text(const char *s, char *nota)
{
  nota[0] = NOTA_TEXT;
  long long n = utf8_count(s);
  nota = nota_continue_num(n,nota,4);  
  utf8_to_kim(&s, &nota);
  return nota;
}

#endif
#endif