| Type | Bug | Status | complete | Date | 11-Sep-2010 14:47 |
|---|---|---|---|---|---|
| Version | alpha 107 | Category | Syntax | Submitted by | meijeru |
| Platform | All | Severity | minor | Priority | high |
| Summary | Empty issue not parsed correctly: Single # is parsed with CRLF |
|---|---|
| Description |
Since issue! is now any-word! and not any-string!, an empty issue should no more exist than an empty get-word (:). However, a single # followed by end-of line is parsed as containing three characters. NOTE: As of R3.A108, a single # is defined as the NONE value. -Carl |
| Example code |
>> length? #
== 3
>> to-binary to-string #
== #{230D0A}
|
| Assigned to | n/a | Fixed in | alpha 108 | Last Update | 2-Nov-2010 13:33 |
|---|
| Comments | |
|---|---|
|
(0002490)
BrianH 13-Sep-2010 01:42 |
Issues shouldn't have the same character restrictions as other words, so there is no reason that an empty issue can't continue to exist. It could be useful for some cases, and the syntax supports it (syntax restrictions are the reason for the character restrictions for the other word types). However, an empty issue should be an empty issue, not an issue with a CRLF. |
|
(0002491)
meijeru 13-Sep-2010 11:27 |
There are more things wrong:
>> probe length? #() 6 ;; expected scanning to stop at () >> probe length? #[] ** Syntax error: invalid constructor: [] >> probe length? #/a 5 == /a ;; can't have it both ways: is /a is scanned separately, length should be 1 >> type? probe length? #/a 5 == /a ;; would have expected refinement! as in following example >> type? probe 1 1 == integer! |
|
(0002492)
BrianH 13-Sep-2010 11:58 |
Your second example was legit, an appropriate response to a botched serialized syntax constructor.
For your fourth example, there were two expressions. Try this instead: >> reduce [type? probe length? #/a] 6 ;; this is wrong, should be 0 == [integer! /a] ;; this is right |
|
(0002494)
meijeru 15-Sep-2010 16:18 |
In the fourth example, my version 107 gives 6 for the length, not 5! Plus, it is still the case that the #/a]CRLF combination is scanned twice, once for producing the length 6 issue, and once for producing the second component of the reduced block. But maybe you consider this one and the same error... |
|
(0002495)
BrianH 16-Sep-2010 05:41 |
Comment fixed. This is only one error, in TRANSCODE. The syntax scanner for the index! type is wrong. The second scan of the characters that you mentioned is correct. |
|
(0002500)
Carl 20-Sep-2010 22:42 |
There are actually quite a few test cases needed here to get issue! right, mainly because # is a lead-in to other forms. The problem is that the scanner is not tuned in that way, as can be seen in this clearly ambiguous case:
>> type? #user@example.com == email! In other words, the scanner is ambivalent about the # lead-in, and we probably need to raise its lexical precedence. However, I can easily fix the immediate bug posted here, and I should mention that I'd like to redefine a single # as a single lexical character representing #[none], which is very common in stored blocks, such as those used in micro-databases. No other alternatives stand out as plainly to me. |
|
(0002519)
BrianH 21-Sep-2010 23:38 |
"I'd like to redefine a single # as a single lexical character representing #[none]" - I'm all for that :) |
|
(0002759)
Ladislav 2-Nov-2010 13:33 |
in the core-tests suite |
| Date | User | Field | Action | Change |
|---|---|---|---|---|
| 2-Nov-2010 13:33 | Ladislav | Comment : 0002759 | Added | - |
| 2-Nov-2010 04:37 | BrianH | Status | Modified | tested => complete |
| 20-Oct-2010 04:13 | BrianH | Status | Modified | built => tested |
| 21-Sep-2010 23:38 | BrianH | Comment : 0002519 | Added | - |
| 20-Sep-2010 22:55 | carl | Description | Modified | - |
| 20-Sep-2010 22:55 | carl | Status | Modified | reviewed => built |
| 20-Sep-2010 22:55 | carl | Fixedin | Modified | => alpha 108 |
| 20-Sep-2010 22:55 | carl | Description | Modified | - |
| 20-Sep-2010 22:47 | carl | Comment : 0002500 | Modified | - |
| 20-Sep-2010 22:44 | carl | Comment : 0002500 | Modified | - |
| 20-Sep-2010 22:42 | carl | Comment : 0002500 | Added | - |
| 16-Sep-2010 05:41 | BrianH | Comment : 0002495 | Added | - |
| 16-Sep-2010 05:39 | BrianH | Comment : 0002492 | Modified | - |
| 15-Sep-2010 16:18 | meijeru | Comment : 0002494 | Added | - |
| 13-Sep-2010 11:58 | BrianH | Comment : 0002492 | Modified | - |
| 13-Sep-2010 11:58 | BrianH | Comment : 0002492 | Added | - |
| 13-Sep-2010 11:27 | meijeru | Comment : 0002491 | Added | - |
| 13-Sep-2010 01:42 | BrianH | Comment : 0002490 | Added | - |
| 13-Sep-2010 01:41 | BrianH | Priority | Modified | normal => high |
| 13-Sep-2010 01:41 | BrianH | Status | Modified | submitted => reviewed |
| 13-Sep-2010 01:41 | BrianH | Category | Modified | Datatype => Syntax |
| 13-Sep-2010 01:41 | BrianH | Code | Modified | - |
| 13-Sep-2010 01:41 | BrianH | Summary | Modified | Empty issue should no more exist -- single # is parsed with CRLF, though => Empty issue not parsed correctly: Single # is parsed with CRLF |
| 11-Sep-2010 14:47 | meijeru | Ticket | Added | - |