| Type | Issue | Status | reviewed | Date | 26-Jul-2009 20:08 |
|---|---|---|---|---|---|
| Version | alpha 76 | Category | Native | Submitted by | Sunanda |
| Platform | All | Severity | minor | Priority | normal |
| Summary | SORT order changed for datatypes |
|---|---|
| Description |
May be related to CC#415. And may be a deliberate break in compatibility with R2 that is worth documenting: R2: sort reduce [1 2 3 unset!] == [unset! 1 2 3] R3: sort reduce [1 2 3 unset!] == [1 2 3 unset!] |
| Example code |
| Assigned to | n/a | Fixed in | - | Last Update | 18-Jun-2010 01:55 |
|---|
| Comments | |
|---|---|
|
(0001426)
BrianH 26-Jul-2009 20:33 |
This is related to #415, but in a surprising way.
There is no sensible answer to the question of whether 1 is less than #[unset!] - the concept simply doesn't apply. The most REBOL can hope to do is to just pick an answer. The answer R3 picked is that when the datatypes of the values are different, the datatypes are compared instead. The result is that the order follows the sorting order of datatypes: The order that the datatypes are listed in system/catalog/datatypes. I expect that R2 also followed this model, but had a different order of datatypes. This is probably not even an error, but I've marked it as an Issue since this clearly needs some documentation. |
|
(0001429)
Sunanda 26-Jul-2009 20:58 |
There's a further, related, issue:
MAXIMUM-OF in R3 is a mezzanine and error!s on: maximum-of reduce [1 2 3 unset!] That is perhaps acceptable behavior if such values are incommensurate. But in R2, MAXIMUM-OF is a native. And (I guess) uses the same internal method that SORT does when comparing disparate data types. It does not fail (or not in the same way as R3 :-): maximum-of reduce [1 2 3 unset!] == [3 unset!] **** The wider issue perhaps is: Which of the lines of code below should: a) fail b) return a consistent result c) return the same consistent result as R2 sort reduce [unset! 1] unset! > 1 unset! = 1 maximum-of reduce [unset! 1] |
|
(0001431)
BrianH 26-Jul-2009 21:18 |
That is because of this:
>> greater? () 1 ** Script error: greater? does not allow unset! for its value1 argument Or more precisely, this: >> () > 1 ** Script error: the comparison is not valid Those errors are proper and, especially the latter one, descriptive. R2 *not* failing is arguably an error. |
|
(0002394)
Ladislav 18-Jun-2010 01:55 |
My opinion is, that we should not mistake #[datatype! unset!] (in the description) with #[unset!] |
| Date | User | Field | Action | Change |
|---|---|---|---|---|
| 18-Jun-2010 01:55 | Ladislav | Comment : 0002394 | Added | - |
| 26-Jul-2009 21:18 | BrianH | Comment : 0001431 | Added | - |
| 26-Jul-2009 20:58 | sunanda | Comment : 0001429 | Added | - |
| 26-Jul-2009 20:33 | BrianH | Comment : 0001426 | Added | - |
| 26-Jul-2009 20:14 | BrianH | Description | Modified | - |
| 26-Jul-2009 20:14 | BrianH | Category | Modified | Unspecified => Native |
| 26-Jul-2009 20:14 | BrianH | Status | Modified | submitted => reviewed |
| 26-Jul-2009 20:14 | BrianH | Type | Modified | Bug => Issue |
| 26-Jul-2009 20:08 | sunanda | Ticket | Added | - |