Is there a dataset available to fully test a base64 encode/decoder?

Question

I see that there are many base64 implementations available in the opensource and I found multiple internal implementations in a product that I am maintaining.

I'm trying to factor out duplicates but I am not 100% certain that all these implementations give identical output. Therfore I need to have a dataset that tests all possible combinations of input.

Is that somewhere available ? google search did not really report it.

I saw a similar question on stackoverflow but that one has not been fully answered and it is actually just asking for one phrase (in ascii) that would test all 64 chars. It does not handle padding with = for example. So one test string will certainly not fit the bill for a 100% test.

Answer 1

Perhaps something like Base64Test in Bouncy Castle would do what you want?. The tricky part in base64 is handling the padding correctly. It's certainly important to cover that as you mentioned. Accordingly, RFC 4648 specifies these test vectors:

   BASE64("") = ""
   BASE64("f") = "Zg=="
   BASE64("fo") = "Zm8="
   BASE64("foo") = "Zm9v"
   BASE64("foob") = "Zm9vYg=="
   BASE64("fooba") = "Zm9vYmE="
   BASE64("foobar") = "Zm9vYmFy"

Some of your implementations may produce base64 output that differs only by whether they insert line breaks, and where implementations that break lines insert the break and the line termination used. You would have to do additional testing to determine whether you can safely replace an implementation that's using one style with a different one. In particular, a decoder might make assumptions about line length or termination.