Random character generator apex code#

You may be pondering “I have never seen base 34 as a commonly used base” and you would be correct. SQL> select to_char(pk) from t where rownum = 1ģ select base34(pk) from t where rownum = 1ĥ select to_char(dec34(base34(pk))) from t where rownum = 1 SQL> select base34(pk) from t where rownum Ģ function dec34(p_str varchar2) return number isħ l_num := l_num * 34 + instr(l_dig,upper(substr(p_str,i,1)))-1 Here is a simple example where we convert the identifier to base 34.Ģ function base34(p_num number) return varchar2 isģ l_dig varchar2(34) := 'AB0CD1EF2GH3JK4LM5NP6QR7ST8UV9WXYZ' Ĩ l_str := substr(l_dig,mod(l_num,34)+1,1) || l_str A solution is to pass them the value as reduced in size by converting the number to a larger base than decimal. Ideally we want to give them something smaller and simpler to remember. Whilst we have a solution for our random and unique identifier requirement, that of course is now a very large number for them to remember and/or note down. This would suggest that this identifier will be potentially provided to users directly, for example, when they register with our application they may be told “Congratulations, your customer ID is 123456” etc. Of course I started this post discussing the need for an identifier that reveals, for instance, a customer account. And similarly I would contend that this identifier is “always” unique because in order to encounter a violation we would need to generate the same sequence number as it cycles around, along with the same random value from our random number generator, and chance upon both of those repeated values at the same fraction of a second within a one minute. Running this demo you can still see that there is a slight semblance of a pattern due in the values due to the sequence but I would contend that this is still an impossible to guess identifier. Duplicate random numbers are not an issue, because the sequence element guarantees uniqueness.

And just to add a little more unpredictability I will suffix the value with a call to DBMS_RANDOM. For example, a timestamp/sequence pair of “123-87” could clash with “12-387” if they were not both fixed width. I am choosing fixed width because if not, if there is a “bleed” of values between the timestamp values and the sequence values, the potential for duplicates would be increased. I will grab the seconds and fractions of a second from the current timestamp, and also I append to that a fixed width sequence to guarantee the uniqueness. Only a sequence or a GUID is a guarantee of uniqueness so an identifier must at a minimum contain a usage of one of those. So to generate a identifier that is both random and therefore hard to guess but unique as well, we need to combine some of the techniques above. SQL> select pk from t where rownum select pk from t The demo runs to completion but we can easily detect that some duplicates occurred.ģ insert into t values (abs(dbms_random.random))

Random character generator apex code#

The lack of uniqueness when using a random number generator randomness this is easy to prove by doing a drop in replacement of our GUID code with DBMS_RANDOM.

For a random number generator to be truly random, it could easily generate the same number 20 times in a row and still be valid in terms of randomness.

However the very definition of “random” runs counter to that of uniqueness. Maybe a solution is to use the random number generation facilities inside the database. It is almost hard to even distinguish that these values are unique, but I’ve highlighted the critical nybble. Similarly, if we consider just the raw data type it also in this demo appears to be random:īut all is not as it seems because the generation of the value for SYS_GUID is platform dependent and when I run the same demo on Linux then suddenly our unique identifier suffers the same issue as the sequence namely it could potentially be predicted by a malicious party. SQL> select pk from t where rownum <= 20 Any solutions to this will be helpful.3 insert into t values (to_number(sys_guid(), 'XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX')) I do get 100% code coverage, but I feel like the test is not executed properly. How can we unit test a random string generator class of this given code.