World: r3wp

how to catch following error? error? try [to-date 29-Feb_2006]

ok, putting date in a string helps ...trying to catch leap year ...

I don't like it - why rebol parser interrupts me so fast?

petr, in the above case (without the quotes), the error happens before 
rebol gets to evaluate the TRY.

you need to put the LOAD or DO that are loading that code in a try 
block.

it does not seem to work:

error? try [load 29-Feb-2006]

I would prefer 29-Feb-2006 error, as well as e.g. error? try [load 
380.250.250] being a bit more relaxed in interpreter ...

how can I catch such syntax errors?

what is your problem?

error? try [load "29-Feb-2006"] ; == true

the problems is uncatchable error

I am not talking about string ....

I am not sure I can meet with such situation in real-life :-) I just 
got asked by Bobik. The thing was, that in sqlite date field there 
can be invalid date. Now I am not sure how is the conversion done, 
if via string, but if you simply type such invalid date in console, 
it can't be recovered, and that is my objection in general ...

shouldn't we be always able to recover from whatever error happens?

To be executed, a script is loaded in a whole. So each values is 
loaded/binded and your error appears at this time, not while the 
expression is evaluated.

do you mean that I can catch the error during load? but why loading 
value itself fails?

what uncatchable error?

try to type in console 29-feb-2006 ....

although it is invalid, I thought that by error? try [29-Feb-2006] 
I will be able to recover eventually, which is not the case ...

>> if error? try [ load form 29-feb-2006 ][print "date format error" 
]
** Syntax Error: Invalid date -- 29-feb-2006

** Near: (line 1) if error? try [ load form 29-feb-2006 ][print "date 
format error" ]
>>

>> if error? try [ load "29-feb-2006" ][print "date format error" 
]
date format error
>>

Doesn't like the 29-Feb-2006 as a date string

date type!

looks like a bug?

Looks like it.

Does not like 32-may-2006 too. Isn't this more a feature than a bug?

error? try [29-Feb-2006]

 is "word! word! [what-is-this?]". Reol has no idea what to do with 
 "error?" at that time.

well, but Volker, anyway, we got to state, where we have non-recoverable 
interpreter state, don't we?

I know the date is invalid, but .... the same goes for tupple, e.g. 
380.250.250

>> 380.250.250
** Syntax Error: Invalid tuple -- 380.250.250
** Near: (line 1) 380.250.250

you are not at interpreter state at that time. think of 'load finding 
an error at compile-time.

I know, I just don't like the fact, console is not safe from such 
errors ..

Thats why ladislav puts the date in a string. compiled later.

how should it work instead?

ok, can we meet with state, where your app returns directly such 
invalid date? e.g. mentioned link to sqlite date field?

how would it work? dunno? a bit more relaxed evaluation? interpreter 
finding string 29-Feb-2006 - it is valid format, not just valid value. 
Under some condition it could be even valid date (leap year). I would 
expect even such error to be catchable ...

ok, so I will have to trust gurus the fact, that technically it can't 
be fixed :-)

how does Rebol not know how to handle 29-feb-2006 as a date! ? I'm 
not sure I understand this, because how then is the unrecoverable 
error generated? from what I see, you'll always need to wrap some 
precautionary measures around handling dates to avoid this. that's 
not good design.

internal its not a string, but a number. you would get 29-feb-2006 
as 1-mar-2006. would that be ok?

anyway ... I don't like current state, that is all. I even don't 
know consequences, just a feeling :-) Maybe Carl could give us his 
opinion, if it is bug or not, and if not, if this is desirable behavior 
...

maybe a hook in 'load, whcih can do some spellchecking?

hmm, help me to imagine, how does interpreter internally works? So 
let's simplify it - it reaches 280.250.250 tuple. It tries to parse 
it. It eventually recognises it as a tupple, otherwise it would not 
return "Invalid tuple", no? But then it finds invalid values ... 
why just doesn't it recognise, it is part of 'try block and does 
not throw error, which would be catchable?

because  it does not try to parse it. it has already.

'load does the parsing. after load has finished, all datatypes are 
in an optimized form.

if load cant handle something, it is helpless. it does not know something 
is inside a try-block. it only knows it has loaded a lot of words 
already and the followingtext is not right.

that means the error happens when the sourcecode with the try is 
laoded, before try as any achance to be executed.

then maybe there could be two phases of load .... parse form, not 
the value first :-) 999.999.999 for a valid tuple! datatype. It is 
too strict, in opposite to:

>> email? to-email "nonsense"
== true

that's why we have the term "loadable", right? Carl talks about this 
when parsing. Otherwise this wouldn't be a problem.

maybe some datatype "weird" with the original string and a type-suggestion? 
but then you get problems with ambiguities.

what is a typo, what ignored? 6-pfu-206 ?