Removing fractional part of a number

funkey_monkey

Is there a way to remove the fractional part of a number without using the mod function or converting to an integer?

I.e. I want to change 2314.67 to 2314.0.


Is should say I am using a very basic tool.


numerator := 2314.67
int denominator := 1
int output := 1

output := 2314.67 / 1

output which was declared as a int (due to lack of decimal in assignment) now takes on the floating point value of the division result.

Would be simple with a mod function or cast to an int. But alas that does not work. The tool does not seem to hold the declared variable in its initial declared type.

In Java / C++ / .......?

funkey_monkey

Nope proprietary test interface.

Was just wondering if there was a mathematical way to do this before I go ask if the owner will add a mod function or request permission for inclusion of a tolerance on my expected results versus the actual returned values.

TheChizler

A bitwise AND?

funkey_monkey

TheChizler wrote: »

A bitwise AND?

The location of the fractional part is not consistent with floating point numbers as the resolution changes as the numbers varies in size. So I don't think bitwise manipulation will work here.
I think the easiest thing is to introduce a tolerance.

Apiarist

It depends on what capabilities are in the tool. You can find the integer number 2314 by employing some search algorithm which compares a test integer to the floating point number. You could convert the float to a string, cut out ".67" and convert back to int. Or may be the tool can do HTML requests? Then create an online "mod" function, send the float to your server and get back an int.

Apiarist

funkey_monkey wrote: »

Nope proprietary test interface....before I go ask if the owner will add a mod function .....

Does it run on a PC? If so, tell them to use some common shell, like Tcl shell.

funkey_monkey

Convert to a string
Parse for '.'
Convert back to num

Yet no modulus operator exists!?! :pac:


So how do I mark this solved?

smacl

funkey_monkey wrote: »

Is there a way to remove the fractional part of a number without using the mod function or converting to an integer?

I.e. I want to change 2314.67 to 2314.0.


Is should say I am using a very basic tool.


numerator := 2314.67
int denominator := 1
int output := 1

output := 2314.67 / 1

output which was declared as a int (due to lack of decimal in assignment) now takes on the floating point value of the division result.

Would be simple with a mod function or cast to an int. But alas that does not work. The tool does not seem to hold the declared variable in its initial declared type.

Looks like Pascal but not. Answer is language dependant, you have functions like floor, ceil and fmod in C or C++ and Round, Frac, Trunc in Pascal. Rounding can also be a minefield, see the following SO question.

Talisman

funkey_monkey wrote: »

numerator := 2314.67
int denominator := 1
int output := 1

output := 2314.67 / 1

If the code is some form of Pascal you can use the div operator. It performs the division operation and discards the fractional part.

2314.67 div 1 => 2314

Alternatively you could just truncate the numerator using the Trunc function.

Trunc(2314.67) => 2314

The Round function will round up for your given example.

Round(2314.67) => 2315

14ned

I'm super late to the reply, sorry, but in C the function for stripping off the fractional part is trunc():

https://en.cppreference.com/w/cpp/numeric/math/trunc

You can implement this using bit masking, just mask off the fractional part. But trunc() does exactly that for you, so no need.

Also, in C a simple cast via an integer would work, but note that larger double values not representable in a 64 bit integer.

smacl

14ned wrote: »

You can implement this using bit masking, just mask off the fractional part.

How exactly would you do that, given the floating point number is represented as exponent and mantissa in binary and not decimal?

14ned

smacl wrote: »

How exactly would you do that, given the floating point number is represented as exponent and mantissa in binary and not decimal?

IEEE754 is ultimately just a bitscan followed by a bitshift of the significand with a few corner case special handling. This is to make it fast to hardwire into transistors, but it can be done by hand quite efficiently. Indeed, back when I began, floating-point hardware was emulated by the CPU.

A quick Google search turns up the algorithm at https://stackoverflow.com/questions/12342926/casting-float-to-int-bitwise-in-c/12343933#12343933. I haven't checked it, but it looks about right.

I would urge people to choose trunc(), or ceil() or floor() instead of writing their own soft floating-point implementation in C. Though, great learning experience of course.

Niall