is.finite {base}

R Documentation

Finite, Infinite and NaN Numbers

Description

is.finite and is.infinite return a vector of the same length as x, indicating which elements are finite or not.

Inf and -Inf are positive and negative ‘infinity’ whereas NaN means “Not a Number”.

Usage

is.finite(x)
is.infinite(x)
Inf
NaN
is.nan(x)

Arguments

x	(numerical) object to be tested.

Details

is.finite returns a vector of the same length as x the jth element of which is TRUE if x[j] is finite (i.e. it is not one of the values NA, NaN, Inf or -Inf).

is.infinite returns a vector of the same length as x the jth element of which is TRUE if x[j] is infinite (i.e. equal to one of Inf or -Inf).

Note

In R, basically all mathematical functions (including basic Arithmetic), are supposed to work properly with +/- Inf and NaN as input or output.

The basic rule should be that calls and relations with Infs really are statements with a proper mathematical limit, see the many examples below.

References

ANSI/IEEE 754 Floating-Point Standard.

Currently (6/1999), Bill M.'s billm@melbpc.org.au tutorial and examples at
http://www.linuxsupportline.com/~billm/

See Also

NA, ‘Not Available’ which is not a number as well, however usually used for missing values.

Examples

pi / 0 ## = Inf a non-zero number divided by zero creates infinity
0 / 0  ## =  NaN

1/0 + 1/0# Inf
1/0 - 1/0# NaN

stopifnot(
    1/0 == Inf,
    1/Inf == 0
)
exp(-Inf) == 0
##              (actually, the last one seems to give NA on not-very-new
##               versions of Linux, which is a Linux bug and seems to be
##               corrected in newer 'libc6' based Linuxen).

stopifnot(
    is.na(0/0),
    !is.na(Inf),
    is.nan(0/0),

    !is.nan(NA)  &&  !is.infinite(NA)  && !is.finite(NA),
     is.nan(NaN) &&  !is.infinite(NaN) && !is.finite(NaN),
    !is.nan(c(1,NA)),
    c(FALSE,TRUE,FALSE) == is.nan(c   (1,NaN,NA)),
    c(FALSE,TRUE,FALSE) == is.nan(list(1,NaN,NA))#-> FALSE in older versions
)


lgamma(Inf) == Inf
Inf + Inf == Inf
Inf - Inf == NaN # NA --- should test with  'is.nan()

(1/0) * (1/0)# Inf
(1/0) / (1/0)# NaN

pm <- c(-1,1) # 'pm' = plus/minus

log(0) == - 1/0
exp(-Inf) == 0

sin(Inf)
cos(Inf)
tan(Inf)
all(atan(Inf*pm) == pm*pi/2) # TRUE


x <- c(100,-1e-13,Inf,-Inf, NaN, pi, NA)
x #  1.000000 -3.000000       Inf      -Inf        NA  3.141593        NA
names(x) <- formatC(x, dig=3)
is.finite(x)
##-   100 -1e-13 Inf -Inf NaN 3.14 NA
##-     T      T   .    .   .    T  .
is.na(x)
##-   100 -1e-13 Inf -Inf NaN 3.14 NA
##-     .      .   .    .   T    .  T
which(is.na(x) & !is.nan(x))# only 'NA': 7

is.na(x) | is.finite(x)
##-   100 -1e-13 Inf -Inf NaN 3.14 NA
##-     T      T   .    .   T    T  T
is.infinite(x)
##-   100 -1e-13 Inf -Inf NaN 3.14 NA
##-     .      .   T    T   .    .  .

##-- either  finite or infinite  or  NA:
all(is.na(x) != is.finite(x) | is.infinite(x)) # TRUE
all(is.nan(x) != is.finite(x) | is.infinite(x)) # FALSE: have 'real' NA

##--- Integer
(ix <- structure(as.integer(x),names= names(x)))
##-   100 -1e-13    Inf   -Inf    NaN   3.14     NA 
##-   100      0     NA     NA     NA      3     NA 
all(is.na(ix) != is.finite(ix) | is.infinite(ix)) # TRUE (still)

ix[3] == (iI <- as.integer(Inf))#> warning: NAs introduced by coercion 
ix[4] == (imI<- as.integer(-Inf))
iI  == .Machine$integer.max # TRUE
imI == -.Machine$integer.max # TRUE

##--- Overflow in simple integer arithmetic:
as.integer(2)*iI #  -2
as.integer(3)*iI #  2147483645
as.integer(3)*iI == iI-2 # TRUE

storage.mode(ii <- -3:5)
storage.mode(zm <- outer(ii,ii, FUN="*"))# integer
storage.mode(zd <- outer(ii,ii, FUN="/"))# double
range(zd, na.rm=TRUE)# -Inf  Inf
zd[,ii==0]

(storage.mode(print(1:1 / 0:0)))# Inf "double"
(storage.mode(print(1:1 / 1:1)))# 1 "double"
(storage.mode(print(1:1 + 1:1)))# 2 "integer"
(storage.mode(print(2:2 * 2:2)))# 4 "integer"

---