ActiveRecord::Base#sanitize_sql_hash_for_conditions & #sanitize_sql_hash_for_assignment should pass column to #quote_bound_value

I just tried this patch on current master and it does pass all the tests in all three DBs. Can their be tests added for this? It might get acceptance quicker if it did? For instance test that the postgresql quoting for say the bit hexadecimal type?

        elsif value.kind_of?(String) && column && column.sql_type =~ /^bit/
          case value
            when /^[01]*$/
              "B'#{value}'" # Bit-string notation
            when /^[0-9A-F]*$/i
              "X'#{value}'" # Hexadecimal notation

If there is no matching schema in the postgresql_specific_schema.rb file, I'm sure one can be added. What do you think?

This is a really bad design that will result in data corruption every time it is passed a string representing a hexadecimal number that is strictly composed of zeros and ones. It is not possible to determine whether such a string represents a binary or hexadecimal number by scanning its contents. Ruby itself imposes prefix tags to make this distinction: e.g. 0b1010 vs. 0x1010. No ActiveRecord DB driver should pretend to infer the base of an untagged string of ones and zeros.

The solution must either:
* avoid using bitstring column types altogether (not preferred) * pick either binary or hexadecimal, reject the other, and fix the driver code * reject untagged strings and fix the driver code

This code was introduced in this changeset: http://dev.rubyonrails.org/changeset/7329

Bug report created for this: https://rails.lighthouseapp.com/projects/8994-ruby-on-rails/tickets...

To be clear: I'm referring to the code block posted by Ken Collins, not to Erik Bryn's diff (which I have not read).

Man this is old and I have no idea what I was even talking about. I'll just bow out now :)