www.digitalmars.com         C & C++   DMDScript  

digitalmars.D - bug? floating point precision with std.format

reply Steven Schveighoffer <schveiguy yahoo.com> writes:
It appears that the precision parameter in std.format differs from its 
meaning in printf. Is that expected behavior?

Example:

import std.stdio;
import core.stdc.stdio;

void main()
{
     auto f = 20.66666;
     writeln(f);
     writefln("%0.3s", f);
     printf("%0.3f\n", f);
}

prints:
20.6667
20.7
20.667

It appears that the precision specifier is dictating the total number of 
digits on *both sides* of the decimal place. Whereas, in C, it's only 
the number of digits *after* the decimal place.

I'm trying to specify 3 places of precision after the decimal. How do I 
do this easily?

I'm having a hard time believing this behavior has never been reported, 
but I can't find anything about it in bugzilla. Tested all the way back 
to 2.040.

-Steve
Jun 05 2017
parent reply Seb <seb wilzba.ch> writes:
On Monday, 5 June 2017 at 15:37:42 UTC, Steven Schveighoffer 
wrote:

 It appears that the precision parameter in std.format differs 
 from its meaning in printf. Is that expected behavior?

 Example:

 import std.stdio;
 import core.stdc.stdio;

 void main()
 {
     auto f = 20.66666;
     writeln(f);
     writefln("%0.3s", f);
     printf("%0.3f\n", f);
 }

 prints:
 20.6667
 20.7
 20.667

 It appears that the precision specifier is dictating the total 
 number of digits on *both sides* of the decimal place. Whereas, 
 in C, it's only the number of digits *after* the decimal place.

 I'm trying to specify 3 places of precision after the decimal. 
 How do I do this easily?

 I'm having a hard time believing this behavior has never been 
 reported, but I can't find anything about it in bugzilla. 
 Tested all the way back to 2.040.

 -Steve


You do realize that you have used "s" in the D version?
This works as expected:

writefln("%0.3f", f); // 20.667
printf("%0.3f\n", f); // 20.667

This is a bit more interesting:

writefln("%0.3s", f); // 20.7
printf("%0.3s\n", f); // 20.
Jun 05 2017
next sibling parent reply "H. S. Teoh via Digitalmars-d" <digitalmars-d puremagic.com> writes:
On Mon, Jun 05, 2017 at 04:29:06PM +0000, Seb via Digitalmars-d wrote:

 On Monday, 5 June 2017 at 15:37:42 UTC, Steven Schveighoffer wrote:

 It appears that the precision parameter in std.format differs from its
 meaning in printf. Is that expected behavior?
 
 Example:
 
 import std.stdio;
 import core.stdc.stdio;
 
 void main()
 {
     auto f = 20.66666;
     writeln(f);
     writefln("%0.3s", f);




[...]

That should be "%0.3f", not "%0.3s".

If you use the "%s" specifier, precision is interpreted differently,
i.e., as "maximum number of characters", as per "%s" in C's printf.


T

-- 
If it tastes good, it's probably bad for you.
Jun 05 2017
parent Steven Schveighoffer <schveiguy yahoo.com> writes:
On 6/5/17 12:53 PM, H. S. Teoh via Digitalmars-d wrote:

 On Mon, Jun 05, 2017 at 04:29:06PM +0000, Seb via Digitalmars-d wrote:

 On Monday, 5 June 2017 at 15:37:42 UTC, Steven Schveighoffer wrote:

 It appears that the precision parameter in std.format differs from its
 meaning in printf. Is that expected behavior?

 Example:

 import std.stdio;
 import core.stdc.stdio;

 void main()
 {
     auto f = 20.66666;
     writeln(f);
     writefln("%0.3s", f);




 [...]

 That should be "%0.3f", not "%0.3s".

 If you use the "%s" specifier, precision is interpreted differently,
 i.e., as "maximum number of characters", as per "%s" in C's printf.


Interesting. I thought s just stood for "interpret based on the type", 
and would automatically switch to floating point 'f'. I see in the docs 
now, it uses 'g', something I've never used.

Curious that 'f' isn't used, I thought it would have been the default.

In any case, I have a fix for my code, move along :)

-Steve
Jun 05 2017
prev sibling parent reply Steven Schveighoffer <schveiguy yahoo.com> writes:
On 6/5/17 12:29 PM, Seb wrote:

 You do realize that you have used "s" in the D version?


Yes, I thought it was a stand in for "use the type to determine the 
specifier", and I mistakenly assumed that would be 'f', since that's 
what I've always used for floating point. Apparently it is 'g', which 
behaves as I have shown.


 This is a bit more interesting:

 writefln("%0.3s", f); // 20.7
 printf("%0.3s\n", f); // 20.


That is really bad, because %s means interpret the parameter as a char * 
string. So the memory pointed at by the bit pattern of 20.66666 cast to 
a pointer, has the first 3 bytes '2', '0', and '.', and then a null 
character (or a bunch of unprintable characters, followed by a null).

Bizarre...

-Steve
Jun 05 2017
parent Jonathan M Davis via Digitalmars-d <digitalmars-d puremagic.com> writes:
On Monday, June 05, 2017 13:23:38 Steven Schveighoffer via Digitalmars-d 
wrote:

 On 6/5/17 12:29 PM, Seb wrote:

 You do realize that you have used "s" in the D version?


 Yes, I thought it was a stand in for "use the type to determine the
 specifier", and I mistakenly assumed that would be 'f', since that's
 what I've always used for floating point. Apparently it is 'g', which
 behaves as I have shown.


I always assumed that it just meant "convert to string" and that it did
basically the same thing that to!string would do, in which case, doing
something like passing a number to it like you did would not be legal.
Clearly, I never read the docs. :)

- Jonathan M Davis
Jun 05 2017