Fixed overflow when creating hash #12
Conversation
This fixes an overflow when creating the hash for type Int64 (long). The result is calculated before the cast, to ensure that the resulting integer isn't negative before the modulo is applied, which can occur when the value is sufficiently large.. By downcasting the Int64 before the modulo operation, significant information is lost, the value will likely not result in the expected index, resulting in an erroneous letter being pulled from the alphabet, or an IndexOutOfRangeException being thrown due to negative values.
|
Nice catch, do you have a case where the bug is showing so we can construct a test for the bug? |
|
I've been running for-loops encoding+decoding+verify of longs. I've started at |
|
Yeah, I'm terrible sorry about this. I have been super busy and I've been meaning to circle back with you on this. It's not just the casting that's the issue. This also happens with large quantities of longs. I believe I had something on the order of 50,000 large values, but I don't recall. I am in the process of coming up with a very concise, sanitized example that should help in creating a test. |
|
Here's a gist that, while seemingly contrived, does hit close to the size of the data. Now, we could argue that maybe this isn't a good use of Hashids with a dataset this size, but that's neither here nor there right now. ;-) git@gist.github.com:62de4246b3953f5e8c3d.git |
|
cool, I'll have a look |
This fixes an overflow when creating the hash for type Int64 (long). The
result is calculated before the cast, to ensure that the resulting
integer isn't negative before the modulo is applied, which can occur
when the value is sufficiently large..
By downcasting the Int64 before the modulo operation, significant
information is lost, the value will likely not result in the expected
index, resulting in an erroneous letter being pulled from the alphabet,
or an IndexOutOfRangeException being thrown due to negative values.