|
Description:
|
JRuby is not currently consistent with MRI.1.9.3,
and in turn MRI.1.9.3 isn't entirely consistent
between behaviour of finite and infinite step values.
In MRI.1.9.3:
(5.0...5.0).step(8.0) {|v| p v} # yields no values
(5.0...5.0).step(1.0 / 0.0) {|v| pv} # yields 5.0
There is a patch after the Ruby code:
I've tested the patch in JRuby.1.7.0.
Patch brings JRuby up to MRI.1.9.3,
and puts JRuby where I think MRI.1.9.3 ought to be.
def test_float_step(begv, endv, stepv)
rv = "stepv= #{stepv};"
itvs = []; begv.step(endv, stepv) {|v| itvs << v}
rv << "(#{begv}).step(#{endv}, stepv) its: #{itvs.inspect};"
itvs = []; rr = (begv..endv); rr.step(stepv) {|v| itvs << v}
rv << "(#{rr).step(stepv) its: #{itvs.inspect};"
itvs = []; rr = (begv...endv); rr.step(stepv) {|v| itvs << v}
rv << "(#{rr).step(stepv) its: #{itvs.inspect};"
puts rv
end
puts
test_float_step 5.0, 7.0, 8.0
test_float_step 5.0, 7.0, 1.0 / 0.0
test_float_step 5.0, 5.0, 8.0
- MRI problem with (5.0...5.0).step(1.0 / 0.0)
test_float_step 5.0, 5.0, 1.0 / 0.0
test_float_step 5.0, 3.0, 8.0
test_float_step 5.0, 3.0, 1.0 / 0.0
if 1==0
puts
test_float_step 5.0, 7.0, -8.0
test_float_step 5.0, 7.0, -1.0 / 0.0
test_float_step 5.0, 5.0, -8.0
test_float_step 5.0, 5.0, -1.0 / 0.0
test_float_step 5.0, 3.0, -8.0
test_float_step 5.0, 3.0, -1.0 / 0.0
end
puts
*****************************************
RubyNumeric.java
static void floatStep19(ThreadContext context, Ruby runtime, IRubyObject from, IRubyObject to, IRubyObject step, boolean excl, Block block) {
...
if (Double.isInfinite(unit)) {
//- if (unit > 0) block.yield(context, RubyFloat.newFloat(runtime, beg)); //-
//+ To be consistent with MRI.1.9.3:
// if (unit > 0 ? beg <= end : beg >= end) {
// block.yield(context, RubyFloat.newFloat(runtime, beg));
// }
//+ Following patch assumes MRI changes to full consistency:
if (excl ? (unit > 0 ? beg < end : beg > end) : //+
(unit > 0 ? beg <= end : beg >= end)) { //+
block.yield(context, RubyFloat.newFloat(runtime, beg)); //+
} //+
} else {
|
