Scilab Website | Contribute with GitLab | Mailing list archives | ATOMS toolboxes

Scilab Online Help
2024.0.0 - English

int8

conversion to one byte integer representation

int16

conversion to 2 bytes integer representation

int32

conversion to 4 bytes integer representation

int64

conversion to 8 bytes integer representation

uint8

conversion to one byte unsigned integer representation

uint16

conversion to 2 bytes unsigned integer representation

uint32

conversion to 4 bytes unsigned integer representation

uint64

conversion to 8 bytes unsigned integer representation

Syntax

y = int8(X)
y = int16(X)
y = int32(X)
y = int64(X)
y = uint8(X)
y = uint16(X)
y = uint32(X)
y = uint64(X)

Arguments

X: matrix of booleans, of encoded integers, or of decimal real or complex numbers. Matrix of strings representing literal big integers for int64 and uint64.
y: matrix of integers encoded on 1, 2, 4 or 8 bytes. Its sizes are X ones.

Description

These functions convert and store data to one, two, four, or eight bytes integers. These data types are especially useful to store big objects such as images, long signals,...

If some X decimal components have a fractional part, it is always trimmed to 0.

Example: int8([-3.6, -2.5, -1.4, 1.3, 3.5, 6.7]) // = [-3, -2, -1, 1, 3, 6].

Values of X components that are out of the targeted interval are wrapped (not saturated). Examples:

int8([127 128 129 130]) // = [127 -128 -127 -126].
uint8([254 255 256 257]) // = [254 255 0 1].

-%inf is always converted into the lower bound of the interval (see below).

%inf is always converted into the upper bound of the interval (see below).

%nan is always converted into 0.

If X is complex encoded, imaginary parts are ignored. Only the real parts are converted and returned.

[%f %t] is always converted into [0 1].

Reading big literal integers with int64 or uint64: Any input like int64(1000000000000000000) implicitly considers 1000000000000000000 as a decimal integer whose mantissa is encoded with only 53 bits < 64 bits. Hence, this mantissa is most often truncated, and the encoded integer we get after the floating point truncation is not the input one. To avoid this effect, the integer can be provided as a string: int64("1000000000000000000") returns the expected integer, without loss of accuracy. For more details about truncation effect for big integers, please refer to the double() page.

Converting [] always keeps it as is, of type 1 (decimal type).

Function	y in	=
int8(X)	[- 2⁷, 2⁷-1]	[-128, 127]
int16(X)	[- 2¹⁵, 2¹⁵-1]	[-32768, 32767]
int32(X)	[- 2³¹, 2³¹-1]	[-2147483648, 2147483647]
int64(X)	[- 2⁶³, 2⁶³-1]	[-9223372036854775808, 9223372036854775807]
uint8(X)	[0, 2⁸-1]	[0, 255]
uint16(X)	[0, 2¹⁶-1]	[0, 65535]
uint32(X)	[0, 2³²-1]	[0, 4294967295]
uint64(X)	[0, 2⁶⁴-1]	[0, 18446744073709551615]

Examples

Rounding to 0 and wrapping, and special values:

int8([-11.7 -8.5 -6.4  6.4 8.5 11.7])
int8([-130 -129 -128 -127 126 127 128 129])
a = -129 + 256*(-3:3)
int8(a)
int8([-%inf %inf %nan])

--> int8([-11.7 -8.5 -6.4  6.4 8.5 11.7])
 ans  =
 -11 -8 -6  6  8  11

--> int8([-130 -129 -128 -127 126 127 128 129])
 ans  =
  126  127 -128 -127  126  127 -128 -127

--> a = -129 + 256*(-3:3)
 a  =
  -897.  -641.  -385.  -129.   127.   383.   639.

--> int8(a)
 ans  =
  127  127  127  127  127  127  127

--> int8([-%inf %inf %nan])
 ans  =
 -128  127  0

About wrapping and upper bounds of unsigned integers:

uint8([-1 %inf])
uint16([-1 %inf])
uint32([-1 %inf])
uint64([-1 %inf])

--> uint8([-1 %inf])
 ans  =
  255  255

--> uint16([-1 %inf])
 ans  =
  65535  65535

--> uint32([-1 %inf])
 ans  =
  4294967295  4294967295

--> uint64([-1 %inf])
 ans  =
  18446744073709551615  18446744073709551615

Converting 64-bits integers into decimal numbers downgrades their accuracy:

i = uint64(2)^63 - 600
i - uint64(double(i))

--> i = uint64(2)^63 - 600
 i  =
  9223372036854775208

--> i - uint64(double(i))
 ans  =
  424

Converting booleans and complex numbers:

int8([%f %t])

int8(-10.2 + 3.4*%i)
uint8(-10.2 + 3.4*%i)

--> int8([%f %t])
 ans  =
  0  1

--> int8(-10.2 + 3.4*%i)
 ans  =
 -10

--> uint8(-10.2 + 3.4*%i)
 ans  =
  246

Reading literal int64 or uint64 integers bigger than 2^53, without floating point truncation effect:

a = [uint64(1000000000000001000) ; uint64("1000000000000001000")]
typeof(a)

a = [int64(-1000000000000001000) ; int64("-1000000000000001000")]
typeof(a)

--> a = [uint64(1000000000000001000) ; uint64("1000000000000001000")]
 a  =
  1000000000000001024
  1000000000000001000

--> typeof(a)
 ans  =
  "uint64"

--> a = [int64(-1000000000000001000) ; int64("-1000000000000001000")]
 a  =
 -1000000000000001024
 -1000000000000001000

--> typeof(a)
 ans  =
  "int64"

[] is not convertible:

i = uint16(3);
i(1,2) = 5.6
typeof(i)

i = uint16([])
i(1,2) = 2.3
typeof(i)

--> i = uint16(3);
--> i(1,2) = 5.6
 i  =
  3  5

--> typeof(i)
 ans  =
 uint16


--> i = uint16([])
 i  =
    []

--> i(1,2) = 2.3
 i  =
   0.   2.3

--> typeof(i)
 ans  =
 constant

History

Version Description

6.0

int64() and uint64() added to Scilab.
Complex input numbers are now accepted.

This	is now	instead of
int8([-%inf, %inf])	[-128, 127]	[0, 0]
int16([-%inf, %inf])	[-32768, 32767]	[0, 0]
int32(%inf)	2147483647	-2147483648
uint8(%inf)	255	0
uint16(%inf)	65535	0
uint32(%inf)	4294967295	0

6.1.2 int64 and uint64 are extended to strings to read big integers > 2⁵³ without truncation.

Report an issue
<< iconvert	Integers	inttype >>

Copyright (c) 2022-2024 (Dassault Systèmes)
Copyright (c) 2017-2022 (ESI Group)
Copyright (c) 2011-2017 (Scilab Enterprises)
Copyright (c) 1989-2012 (INRIA)
Copyright (c) 1989-2007 (ENPC)
with contributors

Last updated:
Tue Oct 24 14:30:03 CEST 2023