be made after padding each string with spaces to the same length.
That means that using a standard MySQL collation, the two strings
'foo' and 'foo ' are equal.
The problem (which just bit me after all these years) is essentially:
query_name = 'foo'
Thing.create(:name => 'foo ')
thing = Thing.where(:name => query_name).first # returns created thing
thing.name == query_name # => false ## ouch
That result seems to violate the POLS to me, but I'd like to hear any
opinions about if and/or where this behavior should be fixed before I
start working on a patch to the MySQL2 gem
Given that is how the SQL specification works, the MySQL2 adapter is working correctly. While I would consider the above extra space a surprise, certainly, I would consider it *more* of a surprise if the adapter returned anything *but* exactly what the dbms fed it. Going down the path of patching the driver to give something else is a slippery slope.
1) It's the gem developer's decision to use `=` without specifying a
BINARY comparison, which would *not* return equal, or LIKE
without wildcards which would also not return equal. So it's not
a cut-and-dried matter of "working correctly".
2) It breaksActiveRecord DB-agnosticism to have the same data
return different results, regardless of who's "right" about adhering
to which standards. The role of an ORM should be to isolate the
app from the vagaries of storage engines.
If I receive an object based on a requested attribute, and then turn
around and find out that that object's attribute IS NOT EQUAL to the
one I requested -- I don't see how that's anything but broken.
However, while I can't actually envision a use case where the trailing
space was significant, I could see a case for making the gem behavior
configurable. Does that make a patch more palatable?