I am sorry about the confusion, I was too hasty.
asInteger(coerceVector(x,INTSXP)) does not work after all. Here are more
details of what I am trying to accomplish: I have a matrix with column
names that are actually known to be integers (because I set them so
myself in the R code, say, colnames(mat) <- 1:10. Of course, they become
converted to character strings.)
The relevant part of my code used to be:
SEXP MyFunction(SEXP mat);
int warn, minY
SEXP rl, cl;
char *rn, *cn;
GetMatrixDimnames(mat, &rl, &cl, &rn, &cn);
minY = IntegerFromString(VECTOR_ELT(cl,0), &warn);
if (warn > 0) error("Names of popmatrix columns are not
integers");
Running some tests it appears that VECTOR_ELT(cl,0) is CHARSXP (which I
wound up using without even knowing it).
I tried replacing the IntegerFromString part with both
asInteger(VECTOR_ELT(cl,0)) and with
asInteger(coerceVector(VECTOR_ELT(cl,0),INTSXP)), but as you surmised,
since VECTOR_ELT(cl,0) is CHARSXP, it does not work.
So, how could I get the actual values in the column names?
Thanks for all your help,
Aniko
-Original Message-
From: Prof Brian Ripley [mailto:[EMAIL PROTECTED]
Sent: Thursday, June 07, 2007 12:51 PM
To: Simon Urbanek
Cc: Douglas Bates; Aniko Szabo
Subject: Re: [Rd] missing IntegerFromString()
On Thu, 7 Jun 2007, Simon Urbanek wrote:
>
> On Jun 7, 2007, at 1:00 PM, Prof Brian Ripley wrote:
>
>> On Thu, 7 Jun 2007, Simon Urbanek wrote:
>>
>>>
>>> On Jun 7, 2007, at 11:33 AM, Douglas Bates wrote:
>>>
>>>> On 6/6/07, Aniko Szabo <[EMAIL PROTECTED]> wrote:
>>>>> Thanks to everybody who responded to my question.
>>>>> asInteger(coerceVector(x,INTSXP)) indeed does what I need. I guess
>>>>> there
>>>>> is a lot I don't understand about type coercion, as I would not
have
>>>>> expected it to work.
>>>>
>>>> It is better to use
>>>>
>>>> asInteger(x)
>>>>
>>>> which will do the coercion if necessary.
>>>
>>> Unfortunately not if it gets a character vector:
>>
>> Yes, if it gets a character *vector*, no if it gets a CHARSXP.
>>
>
> Indeed, I was starting with an assumption that the task at hand is to
get the
> same result as as.integer in R code from a string - mea culpa. I had
the
> impression that it was likely what Aniko wanted (because anything else
would
> work with trivial asInteger which I assumed was not enough). Yes, I
made too
> many unsafe assumptions ;).
You weren't the only one: I think we all are a little confused here
The positive outcome is that I will make asInteger work on CHARSXPs as
well.
Brian
>
> Sorry for the noise,
> Simon
>
>
>>>> .Call("foo","1")
>>> Error: unimplemented type 'character' in 'asInteger'
>>>
>>>
>>>> When you do the coercion
>>>> yourself you should PROTECT the result then UNPROTECT it. Calling
>>>> asInteger directly avoids this overhead without the risk of losing
>>>> data in a garbage collection. asInteger can accomplish this
because
>>>> only the first element of the SEXP x is converted to an integer.
>>>>
>>>
>>> It could, but doesn't ;). That is what the original
IntegerFromString
>>> did, but now you either have to do that yourself or coerce the whole
>>> vector (not as efficient but easier to write :P).
>>
>> But does coerceVector really handle CHARSXPs? There are not vectors
and I
>> don't see it in the code. Consider
>>
>> #include
>>
>> SEXP foo(SEXP x)
>> {
>> return coerceVector(STRING_ELT(x, 0), INTSXP);
>> }
>>
>> SEXP foo2(SEXP x)
>> {
>> Rprintf("%d\n", asInteger(STRING_ELT(x, 0)));
>> return x;
>> }
>>
>>> .Call("foo","1")
>> Error: cannot coerce type char to integer vector
>>> .Call("foo2","1")
>> -2147483648
>> [1] "1"
>>
>> (and that is NA_INTEGER).
>>
>> As I said, if asInteger(coerceVector(x,INTSXP)) works, 'x' is not a
>> CHARSXP. So I have little idea what the actual story here is.
>>
>> Brian
>>
>>
>>> Cheers,
>>> Simon
>>>
>>>
>>>>>
>>>>> Aniko
>>>>>
>>>>> -Original Message-
>>>>> From: Seth Falcon [mailto:[EMAIL PROTECTED]
>>>>> Sent: Tuesday, June 05, 2007 11:24 PM
>
