nim-stint/stint/io.nim

492 lines
17 KiB
Nim
Raw Normal View History

# Stint
# Copyright 2018-2023 Status Research & Development GmbH
# Licensed under either of
#
# * Apache License, version 2.0, ([LICENSE-APACHE](LICENSE-APACHE) or http://www.apache.org/licenses/LICENSE-2.0)
# * MIT license ([LICENSE-MIT](LICENSE-MIT) or http://opensource.org/licenses/MIT)
#
# at your option. This file may not be copied, modified, or distributed except according to those terms.
import
2020-06-13 14:44:13 +00:00
# Standard library
2023-06-14 06:22:52 +00:00
std/[typetraits, algorithm, hashes],
2020-06-13 14:44:13 +00:00
# Internal
./private/datatypes,
2020-06-13 14:44:13 +00:00
./uintops, ./endians2
2023-06-14 06:22:52 +00:00
from stew/byteutils import toHex
2023-06-20 07:59:26 +00:00
# Helpers
# --------------------------------------------------------
{.push raises: [], gcsafe.}
2023-06-20 07:59:26 +00:00
template leastSignificantWord*(a: SomeBigInteger): Word =
2023-06-14 15:26:44 +00:00
mixin limbs
a.limbs[0]
template mostSignificantWord*(a: SomeBigInteger): Word =
2023-06-14 15:26:44 +00:00
mixin limbs
a.limbs[^1]
template signedWordType*(_: type SomeBigInteger): type =
SignedWord
template wordType*(_: type SomeBigInteger): type =
Word
2023-06-20 07:59:26 +00:00
template hash*(num: StUint|StInt): Hash =
# TODO:
# `hashData` is not particularly efficient.
# Explore better hashing solutions in nim-stew.
hashData(unsafeAddr num, sizeof num)
{.pop.}
# Constructors
# --------------------------------------------------------
{.push raises: [], inline, gcsafe.}
func stuint*[T: SomeInteger](n: T, bits: static[int]): StUint[bits] {.inline.}=
## Converts an integer to an arbitrary precision integer.
when sizeof(n) > sizeof(Word):
2023-06-14 06:22:52 +00:00
result.limbs[0] = Word(n and Word.high.T)
2023-06-13 01:13:39 +00:00
result.limbs[1] = Word(n shr WordBitWidth)
else:
2023-06-13 12:03:43 +00:00
result.limbs[0] = Word(n)
2023-06-13 12:03:43 +00:00
func stint*[T: SomeInteger](n: T, bits: static[int]): StInt[bits] {.inline.}=
## Converts an integer to an arbitrary precision signed integer.
when T is SomeUnsignedInt:
2023-06-15 07:31:59 +00:00
result.impl = stuint(n, bits)
2023-06-13 12:03:43 +00:00
else:
if n < 0:
2023-06-14 06:22:52 +00:00
if n == low(T):
# special case, bug #92 workaround
2023-06-15 07:31:59 +00:00
result.impl = stuint(high(T), bits) + stuint(1, bits)
2023-06-14 06:22:52 +00:00
else:
2023-06-15 07:31:59 +00:00
result.impl = stuint(-n, bits)
2023-06-13 12:03:43 +00:00
result.negate
else:
2023-06-15 07:31:59 +00:00
result.impl = stuint(n, bits)
2023-06-14 02:52:56 +00:00
func to*(a: SomeInteger, T: typedesc[StInt]): T =
2023-06-13 12:03:43 +00:00
stint(a, result.bits)
func to*(a: SomeUnsignedInt, T: typedesc[StUint]): T =
stuint(a, result.bits)
2023-06-20 07:59:26 +00:00
{.pop.}
# Conversions
# --------------------------------------------------------
{.push raises: [], gcsafe.}
2023-06-20 07:59:26 +00:00
2023-06-14 15:26:44 +00:00
func truncate*(num: StUint, T: typedesc[SomeInteger]): T {.inline.}=
2019-10-08 13:11:08 +00:00
## Extract the int, uint, int8-int64 or uint8-uint64 portion of a multi-precision integer.
2018-10-08 19:04:00 +00:00
## Note that int and uint are 32-bit on 32-bit platform.
## For unsigned result type, result is modulo 2^(sizeof T in bit)
## For signed result type, result is undefined if input does not fit in the target type.
2023-06-19 11:27:49 +00:00
when T is SomeSignedInt and sizeof(T) <= sizeof(Word):
2023-06-19 10:20:54 +00:00
result = T(num.leastSignificantWord() and Word(T.high))
else:
result = T(num.leastSignificantWord())
when sizeof(T) > sizeof(Word):
result = result or (T(num.limbs[1]) shl WordBitWidth)
2018-10-08 19:04:00 +00:00
2023-06-14 15:26:44 +00:00
func truncate*(num: StInt, T: typedesc[SomeInteger]): T {.inline.}=
## Extract the int, uint, int8-int64 or uint8-uint64 portion of a multi-precision integer.
## Note that int and uint are 32-bit on 32-bit platform.
## For unsigned result type, result is modulo 2^(sizeof T in bit)
## For signed result type, result is undefined if input does not fit in the target type.
let n = num.abs
when sizeof(T) > sizeof(Word):
2023-06-20 07:59:26 +00:00
result = T(n.leastSignificantWord())
2023-06-14 15:26:44 +00:00
else:
result = T(n.leastSignificantWord() and Word(T.high))
if num.isNegative:
when T is SomeUnsignedInt:
raise newException(OverflowDefect, "cannot truncate negative number to unsigned integer")
elif sizeof(T) <= sizeof(Word):
2023-06-14 15:26:44 +00:00
if n.leastSignificantWord() == Word(T.high) + 1:
result = low(T)
else:
result = -result
else:
if n == stint(T.high, num.bits) + 1'u:
result = low(T)
else:
#result = result or (T(num.limbs[1]) shl WordBitWidth)
result = -result
else:
when sizeof(T) > sizeof(Word):
result = result or (T(num.limbs[1]) shl WordBitWidth)
2023-06-14 15:26:44 +00:00
func stuint*(a: StUint, bits: static[int]): StUint[bits] {.inline.} =
## unsigned int to unsigned int conversion
## smaller to bigger bits conversion will have the same value
## bigger to smaller bits conversion, the result is truncated
2023-06-20 07:59:26 +00:00
when bits <= a.bits:
for i in 0 ..< result.len:
result[i] = a[i]
else:
for i in 0 ..< a.len:
result[i] = a[i]
func stuint*(a: StInt, bits: static[int]): StUint[bits] {.inline.} =
## signed int to unsigned int conversion
## bigger to smaller bits conversion, the result is truncated
if a.isNegative:
raise newException(OverflowDefect, "Cannot convert negative number to unsigned int")
2023-06-20 07:59:26 +00:00
stuint(a.impl, bits)
func smallToBig(a: StInt, bits: static[int]): StInt[bits] =
if a.isNegative:
result.impl = stuint(a.neg.impl, bits)
result.negate
else:
result.impl = stuint(a.impl, bits)
func stint*(a: StInt, bits: static[int]): StInt[bits] =
2023-06-20 07:59:26 +00:00
## signed int to signed int conversion
## will raise exception if input does not fit into destination
when a.bits < bits:
if a.isNegative:
result.impl = stuint(a.neg.impl, bits)
result.negate
else:
result.impl = stuint(a, bits)
elif a.bits > bits:
template checkNegativeRange() =
const dmin = smallToBig((type result).low, a.bits)
if a < dmin: raise newException(RangeDefect, "value out of range")
template checkPositiveRange() =
const dmax = smallToBig((type result).high, a.bits)
if a > dmax: raise newException(RangeDefect, "value out of range")
if a.isNegative:
checkNegativeRange()
result.impl = stuint(a.neg.impl, bits)
result.negate
else:
checkPositiveRange()
result.impl = stuint(a, bits)
else:
result = a
func stint*(a: StUint, bits: static[int]): StInt[bits] {.inline.} =
## signed int to unsigned int conversion
## will raise exception if input does not fit into destination
const dmax = stuint((type result).high, a.bits)
if a > dmax: raise newException(RangeDefect, "value out of range")
result.impl = stuint(a, bits)
{.pop.}
# Serializations to/from string
# --------------------------------------------------------
{.push gcsafe.}
func readHexChar(c: char): int8 {.inline.}=
## Converts an hex char to an int
case c
of '0'..'9': result = int8 ord(c) - ord('0')
of 'a'..'f': result = int8 ord(c) - ord('a') + 10
of 'A'..'F': result = int8 ord(c) - ord('A') + 10
else:
raise newException(ValueError, $c & "is not a hexadecimal character")
func skipPrefixes(current_idx: var int, str: string, radix: range[2..16]) {.inline.} =
## Returns the index of the first meaningful char in `hexStr` by skipping
## "0x" prefix
if str.len < 2:
return
2019-03-14 02:43:51 +00:00
doAssert current_idx == 0, "skipPrefixes only works for prefixes (position 0 and 1 of the string)"
if str[0] == '0':
if str[1] in {'x', 'X'}:
doAssert radix == 16, "Parsing mismatch, 0x prefix is only valid for a hexadecimal number (base 16)"
current_idx = 2
elif str[1] in {'o', 'O'}:
doAssert radix == 8, "Parsing mismatch, 0o prefix is only valid for an octal number (base 8)"
current_idx = 2
2023-07-07 05:56:30 +00:00
elif str[1] in {'b', 'B'}:
if radix == 2:
current_idx = 2
elif radix == 16:
# allow something like "0bcdef12345" which is a valid hex
current_idx = 0
else:
doAssert false, "Parsing mismatch, 0b prefix is only valid for a binary number (base 2), or hex number"
func nextNonBlank(current_idx: var int, s: string) {.inline.} =
## Move the current index, skipping white spaces and "_" characters.
const blanks = {' ', '_'}
inc current_idx
while current_idx < s.len and s[current_idx] in blanks:
inc current_idx
func readDecChar(c: range['0'..'9']): int {.inline.}=
## Converts a decimal char to an int
# specialization without branching for base <= 10.
ord(c) - ord('0')
2023-06-20 07:59:26 +00:00
func parse*[bits: static[int]](input: string,
T: typedesc[StUint[bits]],
radix: static[uint8] = 10): T =
## Parse a string and store the result in a Stint[bits] or StUint[bits].
2019-03-14 02:43:51 +00:00
static: doAssert (radix >= 2) and radix <= 16, "Only base from 2..16 are supported"
# TODO: use static[range[2 .. 16]], not supported at the moment (2018-04-26)
# TODO: we can special case hex result/input as an array of bytes
# and be much faster
const base = radix.uint8.stuint(bits)
var curr = 0 # Current index in the string
skipPrefixes(curr, input, radix)
while curr < input.len:
# TODO: overflow detection
when radix <= 10:
result = result * base + input[curr].readDecChar.stuint(bits)
else:
result = result * base + input[curr].readHexChar.stuint(bits)
nextNonBlank(curr, input)
2023-06-20 07:59:26 +00:00
func parse*[bits: static[int]](input: string,
T: typedesc[StInt[bits]],
2023-07-07 06:22:22 +00:00
radix: static[uint8] = 10): T =
2023-06-13 13:39:22 +00:00
## Parse a string and store the result in a Stint[bits] or StUint[bits].
2023-06-13 13:39:22 +00:00
static: doAssert (radix >= 2) and radix <= 16, "Only base from 2..16 are supported"
# TODO: use static[range[2 .. 16]], not supported at the moment (2018-04-26)
# TODO: we can special case hex result/input as an array of bytes
# and be much faster
2023-06-13 13:39:22 +00:00
# For conversion we require overflowing operations (for example for negative hex numbers)
const base = radix.int8.stuint(bits)
2023-06-13 13:39:22 +00:00
var
curr = 0 # Current index in the string
isNeg = false
noOverflow: StUint[bits]
2023-06-13 13:39:22 +00:00
if input[curr] == '-':
doAssert radix == 10, "Negative numbers are only supported with base 10 input."
isNeg = true
inc curr
else:
skipPrefixes(curr, input, radix)
2023-06-13 13:39:22 +00:00
while curr < input.len:
# TODO: overflow detection
when radix <= 10:
noOverflow = noOverflow * base + input[curr].readDecChar.stuint(bits)
else:
noOverflow = noOverflow * base + input[curr].readHexChar.stuint(bits)
nextNonBlank(curr, input)
2023-06-15 07:31:59 +00:00
result.impl = noOverflow
2023-06-13 13:39:22 +00:00
if isNeg:
result.negate
2023-06-20 14:39:37 +00:00
func fromHex*(T: typedesc[StUint|StInt], s: string): T {.inline.} =
## Convert an hex string to the corresponding unsigned integer
parse(s, type result, radix = 16)
2023-06-20 14:39:37 +00:00
func fromDecimal*(T: typedesc[StUint|StInt], s: string): T {.inline.} =
parse(s, type result, radix = 10)
func hexToUint*[bits: static[int]](hexString: string): StUint[bits] {.inline.} =
## Convert an hex string to the corresponding unsigned integer
parse(hexString, type result, radix = 16)
func toString*[bits: static[int]](num: StUint[bits], radix: static[uint8] = 10): string =
## Convert a Stint or StUint to string.
## In case of negative numbers:
## - they are prefixed with "-" for base 10.
## - if not base 10, they are returned raw in two-complement form.
static: doAssert (radix >= 2) and radix <= 16, "Only base from 2..16 are supported"
# TODO: use static[range[2 .. 16]], not supported at the moment (2018-04-26)
const hexChars = "0123456789abcdef"
const base = radix.uint8.stuint(bits)
result = ""
var (q, r) = divmod(num, base)
while true:
when bits <= 64:
result.add hexChars[r.leastSignificantWord()]
else:
result.add hexChars[r.truncate(int)]
if q.isZero:
break
(q, r) = divmod(q, base)
reverse(result)
2023-06-13 12:03:43 +00:00
func toString*[bits: static[int]](num: StInt[bits], radix: static[int8] = 10): string =
## Convert a Stint or StUint to string.
## In case of negative numbers:
## - they are prefixed with "-" for base 10.
## - if not base 10, they are returned raw in two-complement form.
let isNeg = num.isNegative
2023-06-13 15:17:43 +00:00
if radix == 10 and isNeg:
2023-06-15 07:31:59 +00:00
"-" & toString(num.neg.impl, radix)
2023-06-13 12:03:43 +00:00
else:
2023-06-15 07:31:59 +00:00
toString(num.impl, radix)
2023-06-13 12:03:43 +00:00
func `$`*(num: StInt or StUint): string {.inline.}=
toString(num, 10)
2023-06-12 14:07:15 +00:00
func toHex*[bits: static[int]](num: StInt[bits] or StUint[bits]): string {.inline.}=
## Convert to a hex string.
## Output is considered a big-endian base 16 string.
## Leading zeros are stripped. Use dumpHex instead if you need the in-memory representation
toString(num, 16)
2023-06-12 14:07:15 +00:00
func dumpHex*(a: StInt or StUint, order: static[Endianness] = bigEndian): string =
2020-06-13 14:44:13 +00:00
## Stringify an int to hex.
## Note. Leading zeros are not removed. Use toString(n, base = 16)/toHex instead.
##
## You can specify bigEndian or littleEndian order.
## i.e. in bigEndian:
## - 1.uint64 will be 00000001
## - (2.uint128)^64 + 1 will be 0000000100000001
##
## in littleEndian:
## - 1.uint64 will be 01000000
## - (2.uint128)^64 + 1 will be 0100000001000000
let bytes = a.toBytes(order)
result = bytes.toHex()
2023-06-20 07:59:26 +00:00
{.pop.}
# Serializations to/from bytes
# --------------------------------------------------------
{.push raises: [], inline, noinit, gcsafe.}
2023-06-21 08:25:29 +00:00
export fromBytes, toBytes, toBytesLE, toBytesBE
2023-06-12 14:07:15 +00:00
func readUintBE*[bits: static[int]](ba: openArray[byte]): StUint[bits] {.noinit, inline.}=
## Convert a big-endian array of (bits div 8) Bytes to an UInt[bits] (in native host endianness)
## Input:
2022-02-24 19:09:53 +00:00
## - a big-endian openArray of size (bits div 8) at least
## Returns:
## - A unsigned integer of the same size with `bits` bits
result = (typeof result).fromBytesBE(ba)
2023-06-12 14:07:15 +00:00
func toByteArrayBE*[bits: static[int]](n: StUint[bits]): array[bits div 8, byte] {.noinit, inline.}=
2023-06-20 12:45:49 +00:00
## Convert a Uint[bits] to to a big-endian array of bits div 8 bytes
## Input:
## - an unsigned integer
## Returns:
## - a big-endian array of the same size
result = n.toBytesBE()
2023-06-20 12:45:49 +00:00
func fromBytesBE*(T: type StUint, ba: openArray[byte]): T {.noinit, inline.}=
result = readUintBE[T.bits](ba)
2023-06-20 12:45:49 +00:00
template initFromBytesBE*(x: var StUint, ba: openArray[byte]) =
x = fromBytesBE(type x, ba)
func readUintLE*[bits: static[int]](ba: openArray[byte]): StUint[bits] {.noinit, inline.}=
## Convert a lettle-endian array of (bits div 8) Bytes to an UInt[bits] (in native host endianness)
## Input:
## - a little-endian openArray of size (bits div 8) at least
## Returns:
## - A unsigned integer of the same size with `bits` bits
result = (typeof result).fromBytesLE(ba)
func toByteArrayLE*[bits: static[int]](n: StUint[bits]): array[bits div 8, byte] {.noinit, inline.}=
## Convert a Uint[bits] to to a little-endian array of bits div 8 bytes
## Input:
## - an unsigned integer
## Returns:
## - a little-endian array of the same size
result = n.toBytesLE()
func fromBytesLE*(T: type StUint, ba: openArray[byte]): T {.noinit, inline.}=
result = readUintLE[T.bits](ba)
template initFromBytesLE*(x: var StUint, ba: openArray[byte]) =
x = fromBytesLE(type x, ba)
#---------------Byte Serialization of Signed Integer ---------------------------
func readIntBE*[bits: static[int]](ba: openArray[byte]): StInt[bits] {.noinit, inline.}=
## Convert a big-endian array of (bits div 8) Bytes to an Int[bits] (in native host endianness)
## Input:
## - a big-endian openArray of size (bits div 8) at least
## Returns:
## - A signed integer of the same size with `bits` bits
result.impl = (typeof result.impl).fromBytesBE(ba)
func fromBytesBE*(T: type StInt, ba: openArray[byte]): T {.noinit, inline.}=
result = readIntBE[T.bits](ba)
template initFromBytesBE*(x: var StInt, ba: openArray[byte]) =
x = fromBytesBE(type x, ba)
func readIntLE*[bits: static[int]](ba: openArray[byte]): StInt[bits] {.noinit, inline.}=
## Convert a lettle-endian array of (bits div 8) Bytes to an Int[bits] (in native host endianness)
## Input:
## - a little-endian openArray of size (bits div 8) at least
## Returns:
## - A signed integer of the same size with `bits` bits
result.impl = (typeof result.impl).fromBytesLE(ba)
func toByteArrayLE*[bits: static[int]](n: StInt[bits]): array[bits div 8, byte] {.noinit, inline.}=
## Convert a Int[bits] to to a little-endian array of bits div 8 bytes
## Input:
## - an signed integer
## Returns:
## - a little-endian array of the same size
result = n.impl.toBytesLE()
func fromBytesLE*(T: type StInt, ba: openArray[byte]): T {.noinit, inline.}=
result = readIntLE[T.bits](ba)
template initFromBytesLE*(x: var StInt, ba: openArray[byte]) =
x = fromBytesLE(type x, ba)
2023-06-20 07:59:26 +00:00
2023-06-21 08:25:29 +00:00
template toBytesLE*[bits: static[int]](n: StInt[bits]): array[bits div 8, byte] =
result = n.impl.toBytesLE()
template toBytesBE*[bits: static[int]](n: StInt[bits]): array[bits div 8, byte] =
result = n.impl.toBytesBE()
2023-06-20 07:59:26 +00:00
{.pop.}
2023-07-07 06:22:22 +00:00
func getRadix(s: static string): uint8 {.compileTime.} =
if s.len <= 2:
return 10
# maybe have prefix have prefix
if s[0] != '0':
return 10
if s[1] == 'b':
return 2
if s[1] == 'o':
return 8
if s[1] == 'x':
return 16
func customLiteral*(T: type SomeBigInteger, s: static string): T =
when s.len == 0:
doAssert(false, "customLiteral cannot accept param with zero length")
const radix = getRadix(s)
parse(s, T, radix)