Re: Random Code

I used to generate apparent 'random' by having a function like x = (x + static-largishprimenumber) mod maxrange (eg x = (129123 + 12391) mod 1,000,000,000 That always seemed to work for me, and it would take 1000,000 tries to reuse the values, wouldnt it?? I'd prefer to not go AAAA0000 AAAA0001; the other way to go around this, is to IN FACT increment one step at a tme, but the "numbersystem" is not 0-9,A-Z, but J0Z31AQC5NUWSL7DBKRFOEXVTI294MGYPO68H. Of course, only one digit will change a time... Siggy ----- Original Message ----- From: "Svensson, B.A.T. " <B.A.T.Svensson@lumc.nl> To: "Sigurd Magnusson" <phobo@paradise.net.nz>; <php-windows@lists.php.net> Sent: Wednesday, January 24, 2001 8:55 AM Subject: RE: [PHP-WIN] Random Code > Indeed a very interesting question you put forward (and probably OOT too :-) > > In programming there does not exist "The Best Way". "Best" always depends on > what you consider to be important and less important - that's why > programmers most often writes same piece of code in different ways, and > that's why one should be considerate about to think an extra time before one > classifies some other programmers code as "crap shit". :) > (For ex there might be a reasons like "The Best Way" why M$ made there OS > code as they did - though it hard for use normal mortal none-devine > programmers to perceive it... :) > > There exists a class of 1-1 functions that behaves "randomly", you can uses > one of these. They are well know from analytic math. In principle they work > like this: in the beginning (x close to zero) the curve f(x) behave > relatively nice like any other curve, but as one progress further away from > zero (towards +oo) f(x) starts to get a shape that reminds of a truly random > distributed curve, but f(x) is of course still functional dependent - that > is, "FOR-EVERY(a) : IF f(a)=Y1 AND f(a)=Y2 THEN Y1=Y2", read: for each time > you execute the function with an input, it will yield the same output as the > previous calculation with the very same input. In other words, the function > result is determent in advance (that is not randomly generated) - but still: > it is bloody hard to predict its output in without actually having done the > calculation in advance. ;) > > The "only" thing you then have to do is two things: > > 1) create a mapping function for the output, for ex: 1 == 'AAAA-0000', 2 == > 'AAAA-0001', etc, etc, or what ever way you like it to be. > > 2) keep track of "used" numbers. > > > One simple, but naive, approach to 2) would be to simply starting at a > specific predetermined number (x-value), and then just increment x with a > fixed factor, for ex 0,1 (= 0.1, in U.S. notation), or Pi, or > whatever-you-feel-for, whenever you need to produce a new number However, > this does not guarantee uniqueness, which means that you need to check the > current output value against all previous output values. This is made simple > by a brute force algorithm which, obviously, scans the previous number > whether an output of this value has been or not has been produced before. > > This is the "best" way if you want to write a simple algorithm, but it is > not the best way considering CPU load on the system. For this, there exist > hashing techniques to keep track of large number of values, this might be > fast finders of used numbers. This could be considered "best" from an > CPU-minimization approach, but the programming gets a little bit more > tricky. > > But happily enough, as this forum implicitly is about databases, one can > "cheat" and store the numbers in the database(!), index the table that > stores the values, and then get rid of the problem to have to write the > hashing algorithms - since it already done in the DBMS. > > > >-----Original Message----- > >From: Sigurd Magnusson [mailto:phobo@paradise.net.nz] > >Sent: Tuesday, January 23, 2001 9:59 AM > >To: php-windows@lists.php.net > >Subject: [PHP-WIN] Random Code > > > > > >Whats the best way to make a completely random, unique code in > >the format "ABCD-1234" where each of the 8 characters can be > >alphanumerc; 0-9,A-Z... I thought about generating a 36^4 > >number, converting to to base '36', and repeating the process. > >What's a good way to ensure the code is not repeated, though? > > > >Siggy > > >