www.digitalmars.com         C & C++   DMDScript  

digitalmars.D - Negative infinity?

reply Sean Kelly <sean f4.ca> writes:
Am I correct in assuming that D does not support negative infinity?  That was an
assumption I made in unFormat and I thought I'd ask.  Assuming I was wrong and
it does, what about negative zero?


Sean
Sep 22 2004
next sibling parent reply Nick <Nick_member pathlink.com> writes:
In article <cisidv$2shn$1 digitaldaemon.com>, Sean Kelly says...

Am I correct in assuming that D does not support negative infinity? 



By the looks of it, no. See example below.


Assuming I was wrong and it does, what about negative zero?



From what I can see, a float can have the value -0 but it is treated as equal to
0 (which of course is correct behaviour). Consider this example:

# import std.stdio;
# void main()
# {
#   // Produce negative zero
#   float f=-1;
#   for(int i = 0; i < 100; i++)
#     f /= 1000;
#   writefln(f);
#   assert(f==-f);
#
#   // Produce negative infinity
#   f=1/f;
#   writefln(f);
# }

Output is:
-0
-inf

Nick
Sep 22 2004
parent reply Sean Kelly <sean f4.ca> writes:
In article <cismki$2uqa$1 digitaldaemon.com>, Nick says...

In article <cisidv$2shn$1 digitaldaemon.com>, Sean Kelly says...

Am I correct in assuming that D does not support negative infinity? 



By the looks of it, no. See example below.


Assuming I was wrong and it does, what about negative zero?



From what I can see, a float can have the value -0 but it is treated as equal to
0 (which of course is correct behaviour). Consider this example:

# import std.stdio;
# void main()
# {
#   // Produce negative zero
#   float f=-1;
#   for(int i = 0; i < 100; i++)
#     f /= 1000;
#   writefln(f);
#   assert(f==-f);
#
#   // Produce negative infinity
#   f=1/f;
#   writefln(f);
# }

Output is:
-0
-inf



Hrm... so I tred this:

# import std.stdio;
# 
# void main()
# {
#     float f = -0;
#     writefln( f );
#     f = f.infinity * -1;
#     writefln( f );
# }

And it printed:

0
-inf

So while there seems to be an easy way to get negative infinity, I don't know of
one to get negative zero.  I'll update unFormat to handle negative infinity, but
unless someone has a brilliant idea I'm not going to bother with negative zero.


Sean
Sep 22 2004
next sibling parent reply Nick <Nick_member pathlink.com> writes:
In article <cisnr9$2vfj$1 digitaldaemon.com>, Sean Kelly says...

So while there seems to be an easy way to get negative infinity, I don't know of
one to get negative zero.  I'll update unFormat to handle negative infinity, but
unless someone has a brilliant idea I'm not going to bother with negative zero.



Try -1e-1000000000. Maybe there should be a float.negzero property?

Nick
Sep 22 2004
parent reply Sean Kelly <sean f4.ca> writes:
In article <cisp9c$30a6$1 digitaldaemon.com>, Nick says...

In article <cisnr9$2vfj$1 digitaldaemon.com>, Sean Kelly says...

So while there seems to be an easy way to get negative infinity, I don't know of
one to get negative zero.  I'll update unFormat to handle negative infinity, but
unless someone has a brilliant idea I'm not going to bother with negative zero.



Try -1e-1000000000. Maybe there should be a float.negzero property?



No can do.  The compiler errors out with "number is not representable."  A
negzero property would be nice, unless Walter can suggest an alternate method
that doesn't involve a loop?


Sean
Sep 22 2004
parent Nick <Nick_member pathlink.com> writes:
In article <cisrr5$31gf$1 digitaldaemon.com>, Sean Kelly says...

Try -1e-1000000000. Maybe there should be a float.negzero property?



No can do.  The compiler errors out with "number is not representable."  A
negzero property would be nice, unless Walter can suggest an alternate method
that doesn't involve a loop?



Strange, worked for me...

Nick
Sep 23 2004
prev sibling next sibling parent reply Arcane Jill <Arcane_member pathlink.com> writes:
In article <cisnr9$2vfj$1 digitaldaemon.com>, Sean Kelly says...


So while there seems to be an easy way to get negative infinity, I don't know of
one to get negative zero.  I'll update unFormat to handle negative infinity, but
unless someone has a brilliant idea I'm not going to bother with negative zero.

Sean



How about (1/-f.infinity)?
Or ((-1)/f.infinity)?
Or -(1/f.infinity)?

Jill
Sep 23 2004
parent reply Sean Kelly <sean f4.ca> writes:
In article <cittdg$k81$1 digitaldaemon.com>, Arcane Jill says...

How about (1/-f.infinity)?
Or ((-1)/f.infinity)?
Or -(1/f.infinity)?



Darn.  There I go forgetting basic mathematics.  I hope this doesn't mean I'm
over the hill ;)  Thanks Jill!


Sean
Sep 23 2004
parent Arcane Jill <Arcane_member pathlink.com> writes:
In article <ciur0c$1ii1$1 digitaldaemon.com>, Sean Kelly says...


Darn.  There I go forgetting basic mathematics.



The behavior of IEEE floats is pretty bizarre as far as mathematics goes, so I
wouldn't worry about it. In /real/ mathematics, of course, there is only one
zero, and infinity is not even a real number (though the concept has its uses in
limits, set theory and so on).



I hope this doesn't mean I'm
over the hill ;)  Thanks Jill!



You're /never/ over the hill.
Have fun.
Jill
Sep 23 2004
prev sibling parent "Walter" <newshound digitalmars.com> writes:
"Sean Kelly" <sean f4.ca> wrote in message
news:cisnr9$2vfj$1 digitaldaemon.com...

 Hrm... so I tred this:

 # import std.stdio;
 #
 # void main()
 # {
 #     float f = -0;
 #     writefln( f );
 #     f = f.infinity * -1;
 #     writefln( f );
 # }

 And it printed:

 0
 -inf

 So while there seems to be an easy way to get negative infinity, I don't




 one to get negative zero.  I'll update unFormat to handle negative




 unless someone has a brilliant idea I'm not going to bother with negative




Negative 0 does print as just "0". Also, the way to generate a negative 0 is
to write:
    -0.0
writing -0 will not, since the negation happens before the conversion to
double.
Sep 23 2004
prev sibling parent "Walter" <newshound digitalmars.com> writes:
"Sean Kelly" <sean f4.ca> wrote in message
news:cisidv$2shn$1 digitaldaemon.com...

 Am I correct in assuming that D does not support negative infinity?  That




 assumption I made in unFormat and I thought I'd ask.  Assuming I was wrong




 it does, what about negative zero?



Negative infinity as well as negative zero are fully supported in D.
Sep 22 2004