From Newsgroup: comp.lang.fortran

A program was going wrong, and found
FRACTION(3.75)=0.9375 and FRACTION(2.5)=0.9375
--- Synchronet 3.21a-Linux NewsLink 1.2

From Newsgroup: comp.lang.fortran

Woozy Song <suzyw0ng@outlook.com> schrieb:

A program was going wrong, and found
FRACTION(3.75)=0.9375 and FRACTION(2.5)=0.9375

Do you have a self-contained example (something that can be compiled)?
--
This USENET posting was made without artificial intelligence,
artificial impertinence, artificial arrogance, artificial stupidity,
artificial flavorings or artificial colorants.
--- Synchronet 3.21a-Linux NewsLink 1.2

From Newsgroup: comp.lang.fortran

On 8/13/2025 10:21 PM, Woozy Song wrote:

A program was going wrong, and found
FRACTION(3.75)=0.9375 and FRACTION(2.5)=0.9375

Have you tried upgrading to a newer version of gfrotran ?

Gfortran is now at version 15.2 ???
https://gcc.gnu.org/

Version 12 of Gfortran was no longer supported at version 12.5 ???
https://gcc.gnu.org/gcc-12/

Lynn

--- Synchronet 3.21a-Linux NewsLink 1.2

From Newsgroup: comp.lang.fortran

On 8/13/2025 9:21 PM, Woozy Song wrote:

A program was going wrong, and found
FRACTION(3.75)=0.9375 and FRACTION(2.5)=0.9375

I tried it with gfortran version 15, just for fun, and I got the same
result. A search for the fraction intrinsic turned up this:

https://stackoverflow.com/questions/27686204/wrong-output-of-the-intrinsic-function-fraction-in-fortran

Louis
--- Synchronet 3.21a-Linux NewsLink 1.2

From Newsgroup: comp.lang.fortran

Louis Krupp <lkrupp@invalid.pssw.com.invalid> schrieb:

On 8/13/2025 9:21 PM, Woozy Song wrote:

A program was going wrong, and found
FRACTION(3.75)=0.9375 and FRACTION(2.5)=0.9375

I tried it with gfortran version 15, just for fun, and I got the same result.

For

print *,FRACTION(3.75)
print *,FRACTION(2.5)
end program

I get the same result for gfortran 12, 13, 14, 15 and current trunk:

0.937500000
0.625000000

So, please provide a failing test case that actually compiles.
--
This USENET posting was made without artificial intelligence,
artificial impertinence, artificial arrogance, artificial stupidity,
artificial flavorings or artificial colorants.
--- Synchronet 3.21a-Linux NewsLink 1.2

From Newsgroup: comp.lang.fortran

On Thu, 14 Aug 2025 11:21:00 +0800, Woozy Song wrote:

A program was going wrong, and found
FRACTION(3.75)=0.9375 and FRACTION(2.5)=0.9375

Broken as designed, according to the spec.

Why on Earth does FRACTION have this meaning?
--- Synchronet 3.21a-Linux NewsLink 1.2

From Newsgroup: comp.lang.fortran

On Fri, 15 Aug 2025 22:24:52 +0000, Lawrence D’Oliveiro wrote:

On Thu, 14 Aug 2025 11:21:00 +0800, Woozy Song wrote:

A program was going wrong, and found
FRACTION(3.75)=0.9375 and FRACTION(2.5)=0.9375

Broken as designed, according to the spec.

Why on Earth does FRACTION have this meaning?

To what spec are you referring?

gfortran is giving the correct answer (although
I suspect OP has a transcription problem with
the second example).
--
steve
--- Synchronet 3.21a-Linux NewsLink 1.2

From Newsgroup: comp.lang.fortran

On 8/16/2025 7:02 PM, Steven G. Kargl wrote:

On Fri, 15 Aug 2025 22:24:52 +0000, Lawrence D’Oliveiro wrote:

On Thu, 14 Aug 2025 11:21:00 +0800, Woozy Song wrote:

A program was going wrong, and found
FRACTION(3.75)=0.9375 and FRACTION(2.5)=0.9375

Broken as designed, according to the spec.

Why on Earth does FRACTION have this meaning?

To what spec are you referring?

gfortran is giving the correct answer (although
I suspect OP has a transcription problem with
the second example).

I think most casual, non-math, non-floating point experts would expect this:

"The fractional part of a real number is the decimal portion of the
number, excluding the integer part. It represents the difference between
the real number and its integer part. This is often denoted by curly
braces {x} or frac(x).
In more detail:
Real Number Representation:
Every real number can be expressed as the sum of an integer and a
fractional part. For example, 3.14 can be written as 3 + 0.14, where 3
is the integer part and 0.14 is the fractional part.
Fractional Part Function:
The fractional part function, denoted by {x}, is defined as the
difference between the real number x and its integer part.
Formula:
{x} = x - ⌊x⌋, where ⌊x⌋ is the floor function (the greatest integer less than or equal to x).
Example:
If x = 3.14, then {x} = 3.14 - ⌊3.14⌋ = 3.14 - 3 = 0.14.
Range:
The fractional part of a number is always between 0 and 1 (inclusive of
0, exclusive of 1). For integers, the fractional part is 0. "

So, they would expect fraction(3.0) to be 0, not 0.75.
--- Synchronet 3.21a-Linux NewsLink 1.2

From Newsgroup: comp.lang.fortran

On Sat, 16 Aug 2025 20:40:13 -0500, Gary Scott wrote:

On 8/16/2025 7:02 PM, Steven G. Kargl wrote:

On Fri, 15 Aug 2025 22:24:52 +0000, Lawrence D’Oliveiro wrote:

On Thu, 14 Aug 2025 11:21:00 +0800, Woozy Song wrote:

A program was going wrong, and found
FRACTION(3.75)=0.9375 and FRACTION(2.5)=0.9375

Broken as designed, according to the spec.

Why on Earth does FRACTION have this meaning?

To what spec are you referring?

gfortran is giving the correct answer (although
I suspect OP has a transcription problem with
the second example).

I think most casual, non-math, non-floating point experts would expect this:

For someone programming in the Fortran language, I would expect
that they would consult some form of documentation to learn why
their expectation might be wrong. For example, the gfortran
documentation (which comes with the compiler) contains


8.119 ‘FRACTION’ -- Fractional part of the model representation ===============================================================

_Synopsis_:
‘Y = FRACTION(X)’

_Description_:
‘FRACTION(X)’ returns the fractional part of the model
representation of ‘X’.

_Class_:
Elemental function

_Arguments_:
X The type of the argument shall be a ‘REAL’.

_Return value_:
The return value is of the same type and kind as the argument. The
fractional part of the model representation of ‘X’ is returned; it
is ‘X * RADIX(X)**(-EXPONENT(X))’.

_Example_:
program test_fraction
real :: x
x = 178.1387e-4
print *, fraction(x), x * radix(x)**(-exponent(x))
end program test_fraction
--
steve
--- Synchronet 3.21a-Linux NewsLink 1.2

From Newsgroup: comp.lang.fortran

On Sat, 16 Aug 2025 20:40:13 -0500, Gary Scott wrote:

I think most casual, non-math, non-floating point experts would expect
this:

"The fractional part of a real number is the decimal portion of the
number, excluding the integer part. It represents the difference between
the real number and its integer part.

Sure. Except that computers typically do not represent floating-point
numbers in decimal, but in binary. This is the significance of the “model representation” phrase in the language spec.

The way you describe is obviously the most natural interpretation, these
days. And most other languages do it that way. But not Fortran.
--- Synchronet 3.21a-Linux NewsLink 1.2

From Newsgroup: comp.lang.fortran

On 8/17/2025 2:15 AM, Lawrence D’Oliveiro wrote:

On Sat, 16 Aug 2025 20:40:13 -0500, Gary Scott wrote:

I think most casual, non-math, non-floating point experts would expect
this:

"The fractional part of a real number is the decimal portion of the
number, excluding the integer part. It represents the difference between
the real number and its integer part.

Sure. Except that computers typically do not represent floating-point
numbers in decimal, but in binary. This is the significance of the “model representation” phrase in the language spec.

The way you describe is obviously the most natural interpretation, these days. And most other languages do it that way. But not Fortran.

Yes, which is abhorrent.
--- Synchronet 3.21a-Linux NewsLink 1.2

From Newsgroup: comp.lang.fortran

On Sun, 17 Aug 2025 07:15:43 +0000, Lawrence D’Oliveiro wrote:

On Sat, 16 Aug 2025 20:40:13 -0500, Gary Scott wrote:

I think most casual, non-math, non-floating point experts would expect
this:

"The fractional part of a real number is the decimal portion of the
number, excluding the integer part. It represents the difference between
the real number and its integer part.

Sure. Except that computers typically do not represent floating-point numbers in decimal, but in binary. This is the significance of the “model representation” phrase in the language spec.

The way you describe is obviously the most natural interpretation, these days. And most other languages do it that way. But not Fortran.

Most modern languages likely expect a user to do
something like 'x = x - (int)x;' or 'x = x - int(x)'
or some variation thereof if a user wants the decimal
fraction.

Let's see. Hmmm, C23 offer 'y = frexp(x,&n)', which
returns the fractional part of the floating point
representation. That would be 'x = y * 2**n' with y
in [0.5,1) for a binary foating point number. Fortran
2023 happens to have 'y = fraction(x)' with y in [0.5,1)
for binary FP and does not have a side effect.

C23 has 'y = frexpd64(x,&n)' but I doubt that the user
uses '_Decimal64 x'. Not to mention, that frexpd64()
returns 'x = y * 10**n' with y in [1/10,1), which is not
the decimal fractional part that one gets with
'x = x - (int)x;' Fortran 2023 allows a processor to
support a base-10 REAL, and because 'fraction()' is a
generic subprogram 'y = fraction(x)' would return
y in [1/10,1) for a base-10 REAL x.

Perhaps, you're referring to C23's modf(x, &y) function.
But, C23's modf() does something entirely different than
Fortran 2023 mod() (and modulo()) function. If Fortran
had something like modf(), it would be a subroutine to
avoid a function with a side effect.

Finally, if x is an array, in Fortran one can do
'x = x - (int)x'. C23 would be
'for (int i = 0; i < imax; i++) x[i] = x[i] - (int)x[i];'
One of these is a bit more elegant.
--
steve


--- Synchronet 3.21a-Linux NewsLink 1.2

From Newsgroup: comp.lang.fortran

Am 8/17/25 um 05:18 schrieb Steven G. Kargl:

On Sat, 16 Aug 2025 20:40:13 -0500, Gary Scott wrote:

On 8/16/2025 7:02 PM, Steven G. Kargl wrote:

On Fri, 15 Aug 2025 22:24:52 +0000, Lawrence D’Oliveiro wrote:

On Thu, 14 Aug 2025 11:21:00 +0800, Woozy Song wrote:

A program was going wrong, and found
FRACTION(3.75)=0.9375 and FRACTION(2.5)=0.9375

Broken as designed, according to the spec.

Why on Earth does FRACTION have this meaning?

To what spec are you referring?

gfortran is giving the correct answer (although
I suspect OP has a transcription problem with
the second example).

I think most casual, non-math, non-floating point experts would expect this:

For someone programming in the Fortran language, I would expect
that they would consult some form of documentation to learn why
their expectation might be wrong. For example, the gfortran
documentation (which comes with the compiler) contains

8.119 ‘FRACTION’ -- Fractional part of the model representation ===============================================================

_Synopsis_:
‘Y = FRACTION(X)’

_Description_:
‘FRACTION(X)’ returns the fractional part of the model
representation of ‘X’.

_Class_:
Elemental function

_Arguments_:
X The type of the argument shall be a ‘REAL’.

_Return value_:
The return value is of the same type and kind as the argument. The
fractional part of the model representation of ‘X’ is returned; it
is ‘X * RADIX(X)**(-EXPONENT(X))’.

_Example_:
program test_fraction
real :: x
x = 178.1387e-4
print *, fraction(x), x * radix(x)**(-exponent(x))
end program test_fraction

Thank you, Steve.

The following program, adapted from the Example in the documentation you cite:

program test_fraction
real :: x
x = 3.75
print *, fraction(x), x
print *, x * radix(x)**(-exponent(x))
end program test_fraction

prints:
0.937500000 3.75000000
0.00000000

but I was expecting it to print:
0.937500000 3.75000000
0.937500000

because the documentation says that fraction(x) is the same as x * radix(x)**(-exponent(x)) .

I suggest that the documentation is wrong, and should say
' ... model representation of ‘X’ is returned; it is ‘X * REAL(RADIX(X))**(-EXPONENT(X))’.

Concerning the OP's finding that FRACTION(2.5) is printed as 0.9375:
I cannot reproduce that with gfortran 11.5.0 or 13.3.1 or 14.2.1 which all correctly calculate 0.625 .

So to me this is a documentation bug, and possibly a bug in gfortran 12.2 .

Hope this helps,
Kay


--- Synchronet 3.21a-Linux NewsLink 1.2

From Newsgroup: comp.lang.fortran

On Tue, 19 Aug 2025 12:15:37 +0200, Kay Diederichs wrote:

Am 8/17/25 um 05:18 schrieb Steven G. Kargl:

On Sat, 16 Aug 2025 20:40:13 -0500, Gary Scott wrote:

On 8/16/2025 7:02 PM, Steven G. Kargl wrote:

On Fri, 15 Aug 2025 22:24:52 +0000, Lawrence D’Oliveiro wrote:

On Thu, 14 Aug 2025 11:21:00 +0800, Woozy Song wrote:

A program was going wrong, and found
FRACTION(3.75)=0.9375 and FRACTION(2.5)=0.9375

Broken as designed, according to the spec.

Why on Earth does FRACTION have this meaning?

To what spec are you referring?

gfortran is giving the correct answer (although
I suspect OP has a transcription problem with
the second example).

I think most casual, non-math, non-floating point experts would expect this:

For someone programming in the Fortran language, I would expect
that they would consult some form of documentation to learn why
their expectation might be wrong. For example, the gfortran
documentation (which comes with the compiler) contains


8.119 ‘FRACTION’ -- Fractional part of the model representation
===============================================================

_Synopsis_:
‘Y = FRACTION(X)’

_Description_:
‘FRACTION(X)’ returns the fractional part of the model
representation of ‘X’.

_Class_:
Elemental function

_Arguments_:
X The type of the argument shall be a ‘REAL’.

_Return value_:
The return value is of the same type and kind as the argument. The
fractional part of the model representation of ‘X’ is returned; it >> is ‘X * RADIX(X)**(-EXPONENT(X))’.

_Example_:
program test_fraction
real :: x
x = 178.1387e-4
print *, fraction(x), x * radix(x)**(-exponent(x))
end program test_fraction

Thank you, Steve.

The following program, adapted from the Example in the documentation you cite:

program test_fraction
real :: x
x = 3.75
print *, fraction(x), x
print *, x * radix(x)**(-exponent(x))
end program test_fraction

prints:
0.937500000 3.75000000
0.00000000

but I was expecting it to print:
0.937500000 3.75000000
0.937500000

because the documentation says that fraction(x) is the same as x * radix(x)**(-exponent(x)) .

I suggest that the documentation is wrong, and should say
' ... model representation of ‘X’ is returned; it is ‘X * REAL(RADIX(X))**(-EXPONENT(X))’.

Concerning the OP's finding that FRACTION(2.5) is printed as 0.9375:
I cannot reproduce that with gfortran 11.5.0 or 13.3.1 or 14.2.1 which all correctly calculate 0.625 .

So to me this is a documentation bug, and possibly a bug in gfortran 12.2 .

Looks like you've found a documentation bug. The Fortran standard
actually has

Result Value. The result has the value $X \times b^{-e}$,
where $b$ and $e$ are as defined in 16.4 for the representation
of X in the extended real model for the kind of X. ...

where I have marked-up the sentence in LaTeX. IOW, when the gfortran documentation was written, someone transcripted the equation
$X \times b^{-e}$ into Fortran without realizing RADIX returns an
integer.
--
steve


--- Synchronet 3.21a-Linux NewsLink 1.2