(if test then else?) is inconsistent

Description

Issue was "fixed" by adding a comment to the documentation.
I think that's not enough, because actually points to a major inconsistency.
Just have a look at the following code:

;; I know, the definition of y is bad code, and I would
;; never write such rubish, BUT some people do (e.g. the
;; guys at MS, responsible for the sqljdbc4.jar JDBC
;; driver).
(def y (Boolean. false))

(if (= false y)
(println "that's ok"))

(if y
(println "that's inconsistent, because y is " y
" and we proved it with the above if statement"))

Environment

MacOS X 10.10.2

Activity

Show:
Andy Fingerhut
May 5, 2015, 11:50 PM

I'd like to withdraw the last 2 sentences of my previous comment. I think Alex's comment is the shortest accurate answer: It was a choice in Clojure's implementation to do this. As it is right now, there are some compiled 'if' expressions that compare the test expression against both nil (Java null) and Java Boolean.FALSE, and allowing (Boolean. false) to also be treated as false would either require comparing against a third value, or calling a method like Boolean.valueOf() before doing the comparison. Perhaps a micro-optimization, but seems to me like a reasonable one, given the recommendations in Java not to use the Boolean constructors.

import
May 5, 2015, 10:51 PM

Comment made by: jigendaisuke

@Andy Fingerhut

I know all that, but this doesn't make it any more consistent

See, if they won't fix how "if" treats custom made false-values, the next best
thing to do were to manifest this behaviour of "if" into the "=" function (and
yes, that would mean that "=" isn't a mere ".equals" call anymore, because
Boolean false-values would have to be treated in a special way). But then at
least we would have a consistent system - though not exactly in the way I'd
prefer.

Therefore, I can't blame Java. If they had implemented "if" right in the first place,
everything were fine. So this one is on Clojure.

Andy Fingerhut
May 1, 2015, 12:12 AM

These all return 2:

(if nil 1 2)
(if false 1 2)
(if Boolean/FALSE 1 2)

clojure.core/=, like Java .equals, returns true when comparing Boolean/FALSE and (Boolean. false). Blame Java.

import
April 30, 2015, 11:08 PM

Comment made by: jigendaisuke

But, if Boolean/FALSE is the false value, shouldn't

(if (= (Boolean. false) false) true false)

better return false?

Andy Fingerhut
April 30, 2015, 9:41 PM

The longish example/article on ClojureDocs.org might contain additional useful info. Even Java recommends against ever using (Boolean. false): http://clojuredocs.org/clojure.core/if

Declined

Assignee

Unassigned

Reporter

import

Priority

Major

Affects versions