[PATCH] A false value for a boolean field fails validates_presence_of

That happens because validates_presence_of checks attributes with "blank?", nil, false and empty strings are all considered blank. Not sure though if this validation should not be used at all or if it should actually check only that the attribute would not be nil.

I noticed this recently, and without looking at the source I assumed it just tests "if variable" rather than "if !variable.nil?". I suppose if the database has a not-null constraint, and the interface only presents true and false options (and not N/A) then you would never be able to insert an invalid value. On the other hand I don't like not being able to specify validates_presence_of just for completeness.

I ran into this problem today. I think it is worth distinguishing between TRUE, FALSE, and NIL for booleans and allowing validates_presence_of. While the select box should always provide a valid True/False value, a bug in the code somewhere could try to save a model to the database without setting a field that should have a guaranteed TRUE or FALSE. Without validates_presence_of available for booleans, it seems that this could slip by.

Yes, this happens also in mysql, for the record.

This is problem becomes even worse if the field's column is defined to be non-null (at least in mysql) because then the field gets initialized with a value, but the user may not have selected one yet.

The same problem applies to floats, and ints.

As far as I know, there is no work around for boolean and floats to validate that they were entered with non-null columns.

This is intended behavior and documented with the validation method.

From the documentation:

If you want to validate the presence of a boolean field (where the real values are true and false), you will want to use validates_inclusion_of :field_name, :in => [true, false]. This is due to the way Object#blank? handles boolean values. false.blank? # => true}

manfred, are you sure it's the intended behaviour?

The documentation snippet seems to be saying: "validates_presence_of does not work because it's implemented in the wrong way". But why can't it be fixed?

One possible fix is this:

Index: lib/active_record/validations.rb
===================================================================
--- lib/active_record/validations.rb	(revision 7353)
+++ lib/active_record/validations.rb	(working copy)
@@ -79,7 +79,8 @@
     def add_on_blank(attributes, msg = @@default_error_messages[:blank])
       for attr in [attributes].flatten
         value = @base.respond_to?(attr.to_s) ? @base.send(attr.to_s) : @base[attr.to_s]
-        add(attr, msg) if value.blank?
+        add(attr, msg) if value.blank? && value != false
       end
     end

But it makes test_validate_presences test fail because the test sets a serialized attribute to a string that contains only spaces. This string, when sent through YAML::load will be turned into 'false' which validates_presence_of with the fix does let through.

So does that mean that validates_presence_of can not properly validate boolean values because it would break serialized attributes that use validates_presence_of and that get set (for no apparent reason) an empty string as a value? Why should an empty string, when unserialized be turned into 'false' instead of 'nil' or perhaps even, an empty string?

I primarily closed the ticket because it's documented behaviour, not specifically because I think it's the right behaviour. I remember some sort of problem with changing the validation, but I can't really recall what it was. Maybe somebody else's memory about this is better than mine?

Ah, I remember why you're proposed solution doesn't work. If you allow false as a non-blank value, you also allow false on other column types like text and integer.

Assigning false to an integer column seems to convert it to 0 so that does not seem to be a problem but assigning it to a text column:

007:0> x.first_name=false
false
008:0> x.first_name
false
009:0> x.valid?
NoMethodError: undefined method `size' for false:FalseClass
	from /usr/lib/ruby/gems/1.8/gems/activerecord-1.15.3/lib/active_record/validations.rb:493:in `validates_size_of'

(NOTE this is with rails 1.2.3 without my proposed change)

So integer is not really a problem or if it is then integer columns themselves behave in a wrong way (by accepting false and converting it to 0). And text columns, well, validates_presence_of may work on them if they get set to false, but validates_size_of certainly does not, so whats the point? And actually there really is no way to actually set a false value to a text column (as a browser can only send text) so I can't see why this is something that should be specifically validated cause it can basically only be an error in the application code not in what the user submits.

more logical as assignment to integer converts true/false into 1/0 .. assignment to string should convert it into "true" and "false" and reader methods somehow back to true/false.

this patch modifies the validates_presence_of for boolean attributes to accept the false value by specifying...

validates_presence_of :is_published, :allow_false => true

Test case added for the modification. Tested for mysql, all existing test cases and the new are running fine.

resubmitting the updated path and tests as the ticket 9392 is fixed.

added updated patch in context with the ticket 9392