4:51 p.m.
Hi,
totally different small issue:
? eval int($huge)
-2.1474836e+009
... as opposed to floor($huge) and round($huge), which work as expected.
BTW, int(-$huge) is also negative, the same number.
thanks
sven
6:25 p.m.
On 1.12.2022 17:51, Sven Schreiber wrote:
Hi,
totally different small issue:
? eval int($huge)
-2.1474836e+009
Hmm,
if ($huge==1e+100) I would't say it is an issue, since it's far beyond
32 bits integer.
Marcin
--
Marcin Błażejowski
Wyższa Szkoła Bankowa w Toruniu (WSB University in Toruń)
ul. Młodzieżowa 31a
87-100 Toruń
pok. 116 (room)
1:38 p.m.
Am 01.12.2022 um 19:25 schrieb Marcin Błażejowski:
On 1.12.2022 17:51, Sven Schreiber wrote:
>
> ? eval int($huge)
> -2.1474836e+009
Hmm,
if ($huge==1e+100) I would't say it is an issue, since it's far beyond
32 bits integer.
I know this example was a bit ridiculous, but AFAIK nowhere in our
documentation does it say what the usable range of (pseudo-) integers in
gretl is. All I'm saying is that we need a sentence like "don't use this
function outside the range -2147483648 to 2147483647". And is that even
correct? Or are we implying an even smaller range?
thanks
sven
11:46 a.m.
Am 02.12.22 um 14:38 schrieb Sven Schreiber:
Am 01.12.2022 um 19:25 schrieb Marcin Błażejowski:
> On 1.12.2022 17:51, Sven Schreiber wrote:
>>
>> ? eval int($huge)
>> -2.1474836e+009
>
> Hmm,
>
> if ($huge==1e+100) I would't say it is an issue, since it's far beyond
> 32 bits integer.
Yes, may be a valid point (let's hear Allin's final word). But what
about Sven's 2nd case?
I would expect this to work:
<hansl>
assert(int($huge) == 2147483647)
assert(int(-$huge) == -2147483647)
</hansl>
but currently either int($huge) or int(-$huge) returns the wrong sign,
right?
Artur
11:49 a.m.
Am 02.12.22 um 14:38 schrieb Sven Schreiber:
Am 01.12.2022 um 19:25 schrieb Marcin Błażejowski:
> On 1.12.2022 17:51, Sven Schreiber wrote:
>>
>> ? eval int($huge)
>> -2.1474836e+009
>
> Hmm,
>
> if ($huge==1e+100) I would't say it is an issue, since it's far beyond
> 32 bits integer.
>
I know this example was a bit ridiculous, but AFAIK nowhere in our
documentation does it say what the usable range of (pseudo-) integers in
gretl is. All I'm saying is that we need a sentence like "don't use this
function outside the range -2147483648 to 2147483647". And is that even
correct? Or are we implying an even smaller range?
Btw, the actual range is -2147483648 to 2147483648. This works:
<hansl
assert(int(-$huge) == 2147483648)
assert(int($huge) == -2147483648)
</hansl>
even though I still think the signs are wrong.
Artur
11:52 a.m.
Am 03.12.22 um 12:49 schrieb Artur T.:
Am 02.12.22 um 14:38 schrieb Sven Schreiber:
> Am 01.12.2022 um 19:25 schrieb Marcin Błażejowski:
>> On 1.12.2022 17:51, Sven Schreiber wrote:
>>>
>>> ? eval int($huge)
>>> -2.1474836e+009
>>
>> Hmm,
>>
>> if ($huge==1e+100) I would't say it is an issue, since it's far
>> beyond 32 bits integer.
>>
> I know this example was a bit ridiculous, but AFAIK nowhere in our
> documentation does it say what the usable range of (pseudo-) integers
> in gretl is. All I'm saying is that we need a sentence like "don't use
> this function outside the range -2147483648 to 2147483647". And is
> that even correct? Or are we implying an even smaller range?
Btw, the actual range is -2147483648 to 2147483648. This works:
<hansl
assert(int(-$huge) == 2147483648)
assert(int($huge) == -2147483648)
</hansl>
even though I still think the signs are wrong.
Ah, no you were right, Sven. The actual value range is -2147483648 to
2147483647.
See for instance here:
https://study.com/academy/lesson/16-32-64-128-bit-integers.html
Artur
10:56 a.m.
Am 01.12.2022 um 17:51 schrieb Sven Schreiber:
totally different small issue:
? eval int($huge)
-2.1474836e+009
... as opposed to floor($huge) and round($huge), which work as
expected. BTW, int(-$huge) is also negative, the same number.
Off-list discussion revealed that the overflow issue with int() is not a
surprise, although it remains debatable whether the average user is
expected to know when to stop using int(). The alternative being to
return NA.
An additional sentence in the documentation which states that there are
some subtle differences between the behaviors of int(), round(), and
floor(), was also seen as useful.
cheers
sven
11:41 a.m.
Am 01.12.22 um 17:51 schrieb Sven Schreiber:
Hi,
totally different small issue:
? eval int($huge)
-2.1474836e+009
... as opposed to floor($huge) and round($huge), which work as expected.
BTW, int(-$huge) is also negative, the same number.
It should be added that the help states
<help>
By default this is 1.0E100
</help>
Artur
503
days inactive
505
days old
7 comments
3 participants
participants (3)
-
Artur T. -
Marcin Błażejowski -
Sven Schreiber