from stdint import *
import t, c
import stdio


# String copy function
def strcpy(dest: str, src: str) -> str:
    if dest == None or src == None: return None
    original_dest: str = dest
    for s in src:
        dest[0] = s
        dest += 1
    dest[0] = 0
    return original_dest


# String concatenate function
def strcat(dest: str, src: str) -> str:
    if dest == None or src == None: return dest
    original_dest: str = dest
    # Find end of dest
    while dest[0] != '\0':
        dest += 1
    # Copy src
    for s in src:
        dest[0] = s
        dest += 1
    dest[0] = 0
    return original_dest


# String copy function with limited length
def strncpy(dest: str, src: str, n: t.CSizeT) -> str:
    if dest == None or src == None or n == 0: return dest
    original_dest: str = dest
    i: t.CSizeT = 0
    for s in src:
        if i >= n: break
        dest[0] = s
        dest += 1
        i += 1
    # Fill remaining space with '\0'
    while i < n:
        dest[0] = 0
        dest += 1
        i += 1
    return original_dest


# String concatenate function with limited length
def strncat(dest: str, src: str, n: t.CSizeT) -> str:
    if dest == None or src == None or n == 0: return dest
    original_dest: str = dest
    # Find end of dest
    while dest[0] != '\0':
        dest += 1
    # Copy up to n chars from src
    i: t.CSizeT = 0
    for s in src:
        if i >= n: break
        dest[0] = s
        dest += 1
        i += 1
    dest[0] = 0
    return original_dest


# String length function
export def strlen(src: str) -> t.CSizeT:
    length: t.CSizeT = 0
    for _ in src:
        length += 1
    return length

# String comparison function
# 注意：str1[0] - str2[0] 是 i8 减法会回绕（0xFF 表示 -1），返回前必须
# int() 提升到 i32 再相减（与下方 memcmp 的 int(p1[i]) - int(p2[i]) 一致），
# 否则 strcmp("abc","abd") 返回 255 而非 -1（StringTest Test 4 根因）。
def strcmp(str1: str, str2: str) -> int:
    while str1[0] != '\0' and str2[0] != '\0':
        if str1[0] != str2[0]:
            return int(str1[0]) - int(str2[0])
        str1 += 1
        str2 += 1
    return int(str1[0]) - int(str2[0])

def samestr(str1: str, str2: str) -> bool:
    return strcmp(str1, str2) == 0


# String compare function with limited length
def strncmp(str1: str, str2: str, n: t.CSizeT) -> int:
    i: t.CSizeT = 0
    while str1[0] != '\0' and str2[0] != '\0' and i < n:
        if str1[0] != str2[0]:
            return int(str1[0]) - int(str2[0])
        str1 += 1
        str2 += 1
        i += 1
    if i < n:
        return int(str1[0]) - int(str2[0])
    return 0


# Memory compare function
def memcmp(ptr1: t.CVoid | t.CPtr, ptr2: t.CVoid | t.CPtr, n: t.CSizeT) -> int:
    p1: BYTEPTR = BYTEPTR(ptr1)
    p2: BYTEPTR = BYTEPTR(ptr2)
    i: t.CSizeT = 0
    while i < n:
        if p1[i] != p2[i]:
            return int(p1[i]) - int(p2[i])
        i += 1
    return 0


# String find function
def strchr(s: str, cr: int) -> str:
    while s[0] != '\0':
        if s[0] == t.CChar(cr):
            return str(s)
        s += 1
    if cr == '\0':
        return str(s)
    return None

# String reverse find function (find last occurrence of character)
def strrchr(s: str, cr: int) -> str:
    last: str = None
    while s[0] != '\0':
        if s[0] == t.CChar(cr):
            last = s
        s += 1
    if cr == '\0': return s
    return last


# Substring search function (find first occurrence of needle in s, return pointer to that position, or None if not found)
def strstr(s: str, needle: str) -> str:
    if needle[0] == 0:
        return s
    needle_len: t.CSizeT = strlen(needle)
    if needle_len == 0:
        return s
    i: t.CSizeT = 0
    s_len: t.CSizeT = strlen(s)
    while i + needle_len <= s_len:
        if strncmp(s + i, needle, needle_len) == 0:
            return s + i
        i += 1
    return None

# String skip function
def strspn(s: str, skip: str = " ") -> int:
    si: int = 0
    while s[si] != 0:
        found: int = 0
        sj: int = 0
        while skip[sj] != 0:
            if s[si] == skip[sj]:
                found = 1
                break
            sj += 1
        if found == 0:
            break
        si += 1
    return si

# Memory set function
export def memset(ptr: t.CPtr, value: int, num: t.CSizeT) -> t.CPtr:
    # Original implementation:
    # p: BYTEPTR = BYTEPTR(ptr)
    # i: t.CSizeT
    # for i in range(num):
    #     p[i] = t.CUnsignedChar(value)
    if num == 0: return ptr
    c.Asm(f"""mov rdi, {c.AsmInp(ptr, t.ASM_DESCR.REG_ANY)}
              mov eax, {c.AsmInp(value, t.ASM_DESCR.REG_ANY)}
              mov rcx, {c.AsmInp(num, t.ASM_DESCR.REG_ANY)}
              rep stosb""",
          op=[t.ASM_DESCR.CLOBBER_MEMORY, t.ASM_DESCR.CLOBBER_RAX,
              t.ASM_DESCR.CLOBBER_RCX, t.ASM_DESCR.CLOBBER_RDI])
    return ptr


def memset32(ptr: t.CVoid | t.CPtr, value: t.CUInt32T, count: t.CSizeT) -> t.CVoid | t.CPtr:
    # Original implementation:
    # p: UINT32PTR = UINT32PTR(ptr)
    # i: t.CSizeT
    # for i in range(count):
    #     p[i] = value
    if count == 0: return ptr
    c.Asm(f"""mov rdi, {c.AsmInp(ptr, t.ASM_DESCR.REG_ANY)}
              mov eax, {c.AsmInp(value, t.ASM_DESCR.REG_ANY)}
              mov rcx, {c.AsmInp(count, t.ASM_DESCR.REG_ANY)}
              rep stosd""",
          op=[t.ASM_DESCR.CLOBBER_MEMORY, t.ASM_DESCR.CLOBBER_RAX,
              t.ASM_DESCR.CLOBBER_RCX, t.ASM_DESCR.CLOBBER_RDI])
    return ptr


# Memory copy function
export def memcpy(dest: t.CVoid | t.CPtr, src: t.CVoid | t.CPtr, num: t.CSizeT) -> t.CPtr:
    # Original implementation:
    # d: BYTEPTR = BYTEPTR(dest)
    # s: BYTEPTR = BYTEPTR(src)
    # i: t.CSizeT
    # for i in range(num):
    #     d[i] = s[i]
    if num == 0: return dest
    c.Asm(f"""mov rdi, {c.AsmInp(dest, t.ASM_DESCR.REG_ANY)}
              mov rsi, {c.AsmInp(src, t.ASM_DESCR.REG_ANY)}
              mov rcx, {c.AsmInp(num, t.ASM_DESCR.REG_ANY)}
              rep movsb""",
          op=[t.ASM_DESCR.CLOBBER_MEMORY, t.ASM_DESCR.CLOBBER_RCX,
              t.ASM_DESCR.CLOBBER_RDI, t.ASM_DESCR.CLOBBER_RSI])
    return dest


# 内存移动函数
def memmove(dest: t.CVoid | t.CPtr, src: t.CVoid | t.CPtr, num: t.CSizeT) -> t.CVoid | t.CPtr:
    # 原始实现:
    # d: BYTEPTR = BYTEPTR(dest)
    # s: BYTEPTR = BYTEPTR(src)
    # i: t.CSizeT
    # if d < s:
    #     for i in range(num):
    #         d[i] = s[i]
    # elif d > s:
    #     for i in range(num, 0, -1):
    #         d[i - 1] = s[i - 1]
    if num == 0: return dest
    c.Asm(f"""mov rdi, {c.AsmInp(dest, t.ASM_DESCR.REG_ANY)}
              mov rsi, {c.AsmInp(src, t.ASM_DESCR.REG_ANY)}
              mov rcx, {c.AsmInp(num, t.ASM_DESCR.REG_ANY)}
              cmp rdi, rsi
              jb 2f
              std
              add rdi, rcx
              dec rdi
              add rsi, rcx
              dec rsi
              rep movsb
              cld
              jmp 3f
              2:
              rep movsb
              3:""",
          op=[t.ASM_DESCR.CLOBBER_MEMORY, t.ASM_DESCR.CLOBBER_RCX,
              t.ASM_DESCR.CLOBBER_RDI, t.ASM_DESCR.CLOBBER_RSI])
    return dest


def atoi(src: str) -> int:
    num: int = 0
    sign: int = 1
    s: str = src
    if s == None: return 0
    # 跳过空白
    while s[0] == ' ' or s[0] == '\t' or s[0] == '\n' or s[0] == '\r':
        s += 1
    # 符号
    if s[0] == '-':
        sign = -1
        s += 1
    elif s[0] == '+':
        s += 1
    for ch in s:
        if ch < '0' or ch > '9': break
        digit: int = ch - '0'
        num = num * 10 + digit
    return num * sign


def atoll(src: str) -> t.CInt64T:
    num: t.CInt64T = 0
    sign: t.CInt64T = 1
    s: str = src
    if s == None: return 0
    while s[0] == ' ' or s[0] == '\t' or s[0] == '\n' or s[0] == '\r':
        s += 1
    if s[0] == '-':
        sign = -1
        s += 1
    elif s[0] == '+':
        s += 1
    # 十六进制前缀 0x / 0X
    is_hex: int = 0
    if s[0] == '0' and (s[1] == 'x' or s[1] == 'X'):
        is_hex = 1
        s += 2
    if is_hex:
        for ch in s:
            d: int = 0
            if ch >= '0' and ch <= '9':
                d = ch - '0'
            elif ch >= 'a' and ch <= 'f':
                d = ch - 'a' + 10
            elif ch >= 'A' and ch <= 'F':
                d = ch - 'A' + 10
            else:
                break
            digit: t.CInt64T = t.CInt64T(d)
            num = num * 16 + digit
    else:
        for ch in s:
            if ch < '0' or ch > '9': break
            d: int = ch - '0'
            digit: t.CInt64T = t.CInt64T(d)
            num = num * 10 + digit
    return num * sign


# ============================================================
# 十进制字面量 → double 的正确舍入转换（N4）
#
# 旧实现用"逐位累加 + 反复乘/除 10"：小数位超过 16 位后每步都舍入
# （e 偏 +1 ULP、ln2 偏 −1 ULP），指数按 while exp>0: result*=10.0 累乘
# 308 次（DBL_MAX 偏 −6 ULP）。
# 新实现：尾数装进 u64（≤19 位；第 20 位做一次舍入），用 10 的**精确幂**
# 在**双 double（hi+lo，约 106 位）**下做一次缩放，最后一次性舍入。
# 算法先在 Python 里 1:1 对拍 float()（正确舍入）通过：20 万随机样例 +
# 边界样例在**正规数范围**内 bit-exact；次正规/溢出仍走旧逐步缩放兜底
# （语义与旧实现一致，无回归）。
# 约定：不用全局表——10^0..10^22 由循环乘 10 得到（该区间每一步都可精确
# 表示）；p<0 时在正规区算 10^p×2^600 再精确缩回，避免双 double 低位在
# ~1e-292 以下掉进次正规而丢有效位。
# ============================================================

# _dd_split_hi - Dekker 分裂的上半（hi = 取高 26 位）
#
# 含量级缩尺：|a| > 6.7e292 时先乘 2^-28 再算（否则 a*(2^27+1) 溢出为 inf）；
# |a| 极小（< 1e-292）时先乘 2^28（否则 lo = a-hi 掉进次正规丢位）。
# 缩尺因子均为 2 的幂 ⇒ 对正规数**无损**；调用方用 lo = a - hi 取下半部分。
def _dd_split_hi(a: double) -> double:
    aa: double = a
    big: int = 0
    small: int = 0
    if a > 6.7e292 or a < -6.7e292:
        aa = a * 3.725290298461914e-09
        big = 1
    elif (a > 0.0 and a < 1.0e-292) or (a < 0.0 and a > -1.0e-292):
        aa = a * 268435456.0
        small = 1
    tt: double = aa * 134217729.0
    hi: double = tt - (tt - aa)
    if big != 0:
        return hi * 268435456.0
    if small != 0:
        return hi * 3.725290298461914e-09
    return hi


# _dd_mul - 双 double 乘法：out = a × b（out[0]=hi, out[1]=lo）
def _dd_mul(ah: double, al: double, bh: double, bl: double,
            out: t.CArray[double, 2]):
    p: double = ah * bh
    a_hi: double = _dd_split_hi(ah)
    a_lo: double = ah - a_hi
    b_hi: double = _dd_split_hi(bh)
    b_lo: double = bh - b_hi
    e: double = ((a_hi * b_hi - p) + a_hi * b_lo + a_lo * b_hi) + a_lo * b_lo
    e = e + (ah * bl + al * bh)
    s: double = p + e
    out[0] = s
    out[1] = e - (s - p)


# _dd_add - 双 double 加法：out = a + b
def _dd_add(ah: double, al: double, bh: double, bl: double,
            out: t.CArray[double, 2]):
    s: double = ah + bh
    bb: double = s - ah
    e: double = (ah - (s - bb)) + (bh - bb)
    e = e + (al + bl)
    r0: double = s + e
    out[0] = r0
    out[1] = e - (r0 - s)


# _dd_div - 双 double 除法：out = a / b（q0 + 一步残差修正）
def _dd_div(ah: double, al: double, bh: double, bl: double,
            out: t.CArray[double, 2], tmp: t.CArray[double, 2]):
    q0: double = ah / bh
    _dd_mul(bh, bl, q0, 0.0, tmp)
    _dd_add(ah, al, -tmp[0], -tmp[1], tmp)
    q1: double = tmp[0] / bh
    _dd_add(q0, 0.0, q1, 0.0, out)


# _dd_from_u64 - u64 → 双 double（精确：hi 为四舍五入值，lo 为精确残差）
def _dd_from_u64(m: t.CUInt64T, out: t.CArray[double, 2]):
    hi: double = double(m)
    lo: double = 0.0
    if m >= t.CUInt64T(9007199254740992):
        lo = double(m - t.CUInt64T(hi))
    s: double = hi + lo
    bb: double = s - hi
    out[0] = s
    out[1] = (hi - (s - bb)) + (lo - bb)


# _dd_pow10 - 10^e 的双 double（e >= 0；按 ≤22 分块，每块可精确表示）
def _dd_pow10(e: int, out: t.CArray[double, 2], tmp: t.CArray[double, 2]):
    out[0] = 1.0
    out[1] = 0.0
    k: int = e
    while k > 0:
        step: int = 22
        if k < 22:
            step = k
        p10: double = 1.0
        i: int = 0
        while i < step:
            p10 = p10 * 10.0
            i += 1
        _dd_mul(out[0], out[1], p10, 0.0, tmp)
        out[0] = tmp[0]
        out[1] = tmp[1]
        k -= step


# _pow2 - 2^n（|n| <= 1000）：循环乘 2/0.5，每步精确
def _pow2(n: int) -> double:
    r: double = 1.0
    i: int = n
    if i >= 0:
        while i > 0:
            r = r * 2.0
            i -= 1
    else:
        while i < 0:
            r = r * 0.5
            i += 1
    return r


# _atof_scale_stepwise - 旧的逐步 ×/÷10 缩放（次正规/溢出场景的兜底）
def _atof_scale_stepwise(x: double, p: int) -> double:
    r: double = x
    e: int = p
    while e > 0:
        r = r * 10.0
        e -= 1
    while e < 0:
        r = r / 10.0
        e += 1
    return r


def atof(src: str) -> double:
    s: str = src
    if s == None: return 0.0
    # 跳过空白
    while s[0] == ' ' or s[0] == '\t' or s[0] == '\n' or s[0] == '\r':
        s += 1
    # 符号
    sign: double = 1.0
    if s[0] == '-':
        sign = -1.0
        s += 1
    elif s[0] == '+':
        s += 1
    # 尾数：有效数字装进 u64（≤19 位），同时维护十进制指数
    mant: t.CUInt64T = 0
    nd: int = 0
    dexp: int = 0
    seen_dot: int = 0
    started: int = 0
    rounded: int = 0
    while s[0] != '\0':
        ch: t.CChar = s[0]
        if ch >= '0' and ch <= '9':
            d: int = ch - '0'
            if started == 0:
                # 前导零不进尾数；小数部分的前导零要占十进制位
                if d != 0:
                    started = 1
                    mant = t.CUInt64T(d)
                    nd = 1
                    if seen_dot != 0:
                        dexp -= 1
                elif seen_dot != 0:
                    dexp -= 1
            elif nd < 19:
                mant = mant * t.CUInt64T(10) + t.CUInt64T(d)
                nd += 1
                if seen_dot != 0:
                    dexp -= 1
            else:
                # 第 20 位起：整数部分丢弃的位让指数 +1；小数部分丢弃的位不影响
                # 指数。只用第 20 位做一次舍入（丢弃量 <= 10^-19 相对误差）。
                if seen_dot == 0:
                    dexp += 1
                elif rounded == 0:
                    if d >= 5:
                        mant += t.CUInt64T(1)
                    rounded = 1
        elif ch == '.':
            seen_dot = 1
        else:
            break
        s += 1
    if started == 0:
        return sign * 0.0
    # 进位：舍入可能使 19 位变 20 位
    if mant >= t.CUInt64T(10000000000000000000):
        mant = mant // t.CUInt64T(10)
        dexp += 1
    # 指数部分
    exp: int = 0
    if s[0] == 'e' or s[0] == 'E':
        s += 1
        exp_sign: int = 1
        if s[0] == '-':
            exp_sign = -1
            s += 1
        elif s[0] == '+':
            s += 1
        while s[0] >= '0' and s[0] <= '9':
            exp = exp * 10 + (s[0] - '0')
            s += 1
        exp = exp * exp_sign
    p: int = dexp + exp
    if p == 0:
        return sign * double(mant)
    # 双 double 缩放：mant × 10^p，最后一次舍入
    ta: t.CArray[double, 2]
    tb: t.CArray[double, 2]
    tc: t.CArray[double, 2]
    tr: t.CArray[double, 2]
    _dd_from_u64(mant, ta)
    v: double = 0.0
    if p > 0:
        _dd_pow10(p, tb, tr)
        _dd_mul(ta[0], ta[1], tb[0], tb[1], tr)
        v = tr[0] + tr[1]
    else:
        # p < 0：在正规区算「2^600 / 10^(-p)」= 10^p × 2^600，最后精确缩回。
        # 不能直接算 10^p（|p| 大时 ~1e-308，双 double 低位掉次正规丢有效位）；
        # 分母按 ≤308 分块累除（10^313 会先溢出成 inf 使结果变 NaN）。
        tb[0] = _pow2(600)
        tb[1] = 0.0
        q: int = -p
        while q > 0:
            step2: int = 308
            if q < 308:
                step2 = q
            _dd_pow10(step2, tc, tr)
            _dd_div(tb[0], tb[1], tc[0], tc[1], tr, tc)
            tb[0] = tr[0]
            tb[1] = tr[1]
            q -= step2
        _dd_mul(ta[0], ta[1], tb[0], tb[1], tr)
        v = (tr[0] + tr[1]) * _pow2(-600)
    # 正规数范围（非 NaN / 非 inf / 非次正规）→ 已正确舍入的结果
    if v == v and v <= 1.7976931348623157e308 and v >= -1.7976931348623157e308 \
            and (v == 0.0 or v >= 2.2250738585072014e-308 or v <= -2.2250738585072014e-308):
        return sign * v
    # 溢出 / 次正规 / NaN → 回退旧逐步缩放（与旧实现语义一致）
    return sign * _atof_scale_stepwise(double(mant), p)


def atof_out(src: str, out: double | t.CPtr) -> int:
    """atof 结果写入 out[0]（输出参数方式，避免跨模块 double 返回值传递 bug）

    返回值仅表示成功与否（1=成功，0=失败），结果值通过 out 指针传递，
    不依赖 double 返回值的跨模块传递。
    """
    if out is None:
        return 0
    v: double = atof(src)
    out[0] = v
    return 1

def split(s: str, delim: str, result: t.CArray[str], cap: int) -> int:
    """按单字符分隔符切分 s，token 指针写入 result[0..]，尾部写 None 哨兵。

    cap: result 数组的**声明容量**（元素个数）。此前无容量参数 → 写
    result[count] 与 result[count]=None 均无上界；token 数超过调用方数组
    容量时越界写（devfs 用 t.CArray[str, 256]，256 个 token 即写穿 2048 字节）。
    现限制 token 数 ≤ cap-1（留一格给结尾哨兵）。
    """
    count: int = 0
    start: str = s
    if not s or not delim or not result:
        return 0
    if cap <= 0:
        return 0
    d: t.CChar = delim[0] # 单分隔符

    while start[0]:
        # 跳过分隔符
        while start[0] == d: start += 1
        if not start[0]: break
        # 容量保护：留最后一格放结尾哨兵
        if count >= cap - 1:
            break

        # 记录token起始
        result[count] = start
        count += 1

        # 找到下一个分隔符，截断
        while start[0] and start[0] != d: start += 1
        if start[0]:
            start[0] = '\0'
            start += 1
    result[count] = None # 结尾标记
    return count


ascii_lowercase: str = "abcdefghijklmnopqrstuvwxyz"
ascii_uppercase: str = "ABCDEFGHIJKLMNOPQRSTUVWXYZ"
ascii_letters: str = "abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ"

digits: str = "0123456789"
hexdigits: str = "0123456789abcdefABCDEF"
octdigits: str = "01234567"
punctuation: str = "!\"#$%&'()*+,-./:;<=>?@[]^_`{|}~"
whitespace: str = " \t\n\r\f\v"
printable: str = "0123456789abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ!\"#$%&'()*+,-./:;<=>?@[\\]^_`{|}~ \t\n\r\f\v"