I am disappointed -- but should not be surprised -- that no one found anything interesting in my code. That's OK, What I do NOT like are the incorrect implications that my code is defective!
It's just another of my many perversions, but I've always been a fan of "nifty one-liners." Need to convert a Gregorian date to day of the week? Why use a loop if there's a nifty one-line macro
In webpages you need to do that sometimes so it’s easy to do it in PHP (code from ChatGPT)
Code:
$date = '2023-05-05'; // The date you want to convert
$day_of_week = date('l', strtotime($date)); // 'l' stands for the full name of the day of the week
If someone has already written a function to do XXX, of course one can invoke that function and it will be a one-liner that does XXX. Duh.
But somebody had to write that function! That somebody could be anyone who finds it FUN to write low-level code. It could be (and often was) me. (There's a variety of consumer devices circa 2000 that run code or hardware algorithms written or designed by me.)
I'm reminded of the time I tried to explain the power of Unix by pointing out that the
rot13 service could be obtained in Unix with simply
tr n-zN-Za-mA-M a-mA-Mn-zN-Z
I got a reply like "So what? Microsoft has the same thing: you just download
rot13.exe."
Talk about missing the point!
C:
#include <stdio.h>
#include <stdlib.h>
#define NBITS 27 // number of bits returned by random()
#define MAXRAND 0x7ffffff
#define NB_TO_M(nb) (((unsigned long long)1 << (nb)) - 1)
// Do NOT try to understand the following arithmetic gibberish.
// Just know that IT WORKS !!!
#define M_TO_NB(m) \
((m) / ((m) % 255 + 1) / 255 % 255 * 8 + 7 - 100 / ((m) % 255 + 14))
In books like the Clean Coder they say code should be self-documenting - so you could have helpful function and variable names even if they are more than ten characters long. I'm not sure why you're using an unsigned long long.... it allows up to 18,446,744,073,709,551,615 and might be slower than a smaller integer type.
I did NOT obfuscate the code I delivered to customers, The code I post HERE at IIDB is often deliberately terse to be more challenging. The M_TO_NB macro is seldom if ever used, but seemed like an interesting example for a post about one-liners.
The reason I use
long long is in case you apply the macro to such a 64-bit integer. Much code will use such integers, if only in the future. Why limit the utility of the macro?
As for "might be slower", the MAXRAND (or similar) argument provided to the macro will be a compile-time constant, so the invoked macro is a compile-time constant also. It will be computed ZERO times during execution. (Some scenarios will execute the macro ONCE during execution.)
Still don't see how the code in your post has any valuem and as usal you don't seem able to explain why.
A normal distribution can have non zero means. Look at the equation. The normalized variate has zero mean and a stAndard deviation of 1. If yOu do not understnd that then you really do not undertsnd and your polar coodrnates and 2d is just hand waving.
I tried to explain that the function returns a 2-D gaussian in polar coordinates. And that I wasn't trying to complicate needlessly, but rather
that it's EASIER to compute the 2D gaussian variates (and in polar coordinates) than to calculate a 1-D variate. (Similarly sometimes integrating a 2D gaussian is easier than integrating a 1D!)
Your typical post. A moving goal post. This is why I don't usually reond to your posts.
Not responding would probably work well.
Making it easier to read.
double ruf(void){
double k1,k2;
unsigned int mask = RAND_MAX>>2;
k1 = sqrt((double)((rand() & mask | 1) ));
k2 = (k1/(double)(mask + 1));
return k2;
}
There's an unrelated
sqrt that somehow got interjected into your version of my code.
sqrt((double)((rand() & mask | 1) )) is ambiguous. You have to specify grouping for the logic.
Wrong. C defines precedence, so the "grouping" here is defined. Most coders don't have a problem with (a*b*c + 3*f + d) and because useless parentheses become distracting, that expression is more readable than ((((a*b)*c)+(3*f))+d). Admittedly, most programmers haven't memorized all the precedences, and omitting the parentheses from ((rand() & mask) | 1) can be called obfuscation, Note that in the wrong-precedence version the "| 1" would have no effect. Thus the precedence can be immediately deduced by anyone who's forgotten it.
sqrt((double)(((rand() & mas) | 1) )) or sqrt((double)((rand() & (mask | 1)) ));
Note that the compiler flags a warning ob the ambiguity. Also nesting built in function calls can lead to problems. The compiler parser may not always implement what yiou intend.
"Nesting built in function calls can lead to problems", while perhaps not completely wrong, is confused. Code either does or does not comply with spec. Mine does.
I actually ensure that my code produces no warnings by default (gcc -W). Perhaps you use compiler flags that led to more stringent warnings. My experience is that "-W" flags possible real problems. If you prefer whining about parentheses, that's fine for you.
Did you mean rand() % mask + 1 instead of rand() & mask | 1 ?
No. I'm deliberately outputting a random x, with 0 < x < 1 where the inequalities are STRICT. Thus (ruf() * N) will be at most N-1. That would NOT be the case if ruf() were allowed to return 1.0. (Debugging might miss this once-in-a-billion problem.)
I ran your code.Using your ruff() the resulting histogram is a negative exponential. Using a unfirm distribution results in a log normal distribution.
Your ruf() does not appear to be a uniform distribution. I checd my resits in C against Scilab using their random uniform distribution.
Code:
double ruf(void){
double k1,k2;
unsigned int mask = RAND_MAX>>2;
k1 = sqrt((double)((rand() & mask | 1) ));
k2 = (k1/(double)(mask + 1));
return k2;
}
Again, where did the sqrt come from? Test with my ACTUAL code please, not with you confused version of it.
