What can I do to get Oracle and Java hash codes to match?

[lozina]

I want to be able to take an identical string and get the same hash code for it from both Oracle and Java.

If I use ora_hash ( string ) in Oracle
and String.hashCode() in Java

I get a different result

Do I have another option?

 

[DaveSimmons]

Does Oracle have any functions for a standard / known hashing scheme like MD5, or just their unknown hash code?

If not, can you write your own c++ / java / etc. extension / function for Oracle to add one? (I have no idea how extensible Oracle dbs are.)

 

[Cogman]

What david said, there are literally 1000's of hashing algorithms out there. To get it to match, you need to be using the same hashing algorithm and the same seed.

 

[Train]

Check the docs, the common ones today, IIRC, are SHA-1 and SHA-2, or possibly MD5 like someone mentioned above.

Also make sure the "string" in Oracle is treated the same as a string in Java, i.e. is your Oracle string unicode and your Java string UTF? That could also make a difference

 

[Crusty]

IIRC Java uses something like for a string hash.

for i = 0 -> n - 1
hash += string * 31 ^(n-1-i)

It should be pretty trivial to implement that in Oracle. I'm sure you could create some sort of UDF that does your own hashing instead of relying on the built-in one.

 

[boran]

Nobody has mentioned yet here that a string hash in Java is NOT guaranteed to be unique. Its only function is for use in a hash table to place strings in buckets.
I recommend like other posters to use a hashing algorithm that is designed to generate unique hashes.

 

[Train]

Nobody has mentioned yet here that a string hash in Java is NOT guaranteed to be unique. Its only function is for use in a hash table to place strings in buckets.
I recommend like other posters to use a hashing algorithm that is designed to generate unique hashes.

NO language or hashing algorithm can guarantee uniqueness. Hash collisions are something that need to be handled.

The only way to guarantee a 1:1 seed:hash is to have an exact match within the hash space. i.e. an 8 bit hash for an 8 bit input, wherein the algo never repeats its output. Of course this is only feasible for small inputs, Once you go to something like variable length strings, your hashspace is usually 128 bits or more, which is next to impossible to guarantee a 1:1 mapping

 

[Chris27]

I recommend like other posters to use a hashing algorithm that is designed to generate unique hashes.

Tell me when you find one that works in the general case.

 

[DaveSimmons]

I recommend like other posters to use a hashing algorithm that is designed to generate unique hashes.

We aren't recommending a unique hash, we're recommending a documented hash that follows some standard like MD5 or SHA-1.

Crusty's approach of coding your own for Oracle and possibly also Java will work too if done carefully (making sure large values don't overflow differently for example).

 

[degibson]

Tell me when you find one that works in the general case.

Yes, I'll give you $150 -- for exclusive rights.

 

[ForumMaster]

Or, just use the built in package:

link

 

[degibson]

Another thread rises from the grave.

 

[Cogman]

Whats with this trend all the sudden?

 

[EagleKeeper]

Another thread rises from the grave.

Allowed only because the post contains useful info and this forum is not normally frequented by troll types., but by those that want to assist their fellow crazys()

Common Courtesy
AT Admin