[R] Problem with object inside a function

[Duncan Murdoch]

On 20/04/2008 3:04 PM, Ronaldo Reis Junior wrote:
> Em Dom 20 Abr 2008, Duncan Murdoch escreveu:
>> On 20/04/2008 12:32 PM, Ronaldo Reis Junior wrote:
>>> Em Dom 20 Abr 2008, Duncan Murdoch escreveu:
>>>> On 20/04/2008 9:40 AM, Ronaldo Reis Junior wrote:
>>>>> Hi all,
>>>>>
>>>>> I have a problem with an object inside a formula in a function that I
>>>>> make.
>>>>>
>>>>> I create an object named qvartemp. This object is OK, look the levels:
>>>>>> print(levels(qvartemp))
>>>>> [1] "baixomedio" "alto"
>>>>>
>>>>> Now I create a new object with a new formula:
>>>>>> new.form <-
>>>>>> as.formula(gsub(qvar,"qvartemp",as.expression(mma$formula)))
>>>>> Look this new object:
>>>>>> print(new.form)
>>>>> Riqueza ~ Biomassa * qvartemp
>>>>>
>>>>> Now here is the problem, I try to update an existing model (mma)
>>>>> changing the formula by the new.form
>>>>>
>>>>> mmaa <- update(mma,new.form)
>>>>>
>>>>> But my function is aborted with this message:
>>>>>
>>>>> Error in eval(expr, envir, enclos) : object "qvartemp" not found
>>>>>
>>>>> If I create this object qvartemp in R section manually it my function
>>>>> work, because it get the object from my workspace (but it is usefull
>>>>> just for testing), but my function dont find this object qvartemp
>>>>> created inside the function.
>>>>>
>>>>> If I run all my function line by line without use the function it work,
>>>>> I think that is because in this case the object is created in my
>>>>> workspace. But when I run a function the object is created only in
>>>>> memory (not in my workspece) and in this case update dont find this
>>>>> object.
>>>>>
>>>>> Anybody can suggest any Idea to fix it? I necessary I send my function
>>>>> and a example.
>>>> Formulas have attached environments; that's where they look for the
>>>> variables they use.  You created new.form, but didn't give it the
>>>> environment of the old one.
>>>>
>>>> I think all you need is
>>>>
>>>> environment(new.form) <- environment(mma$formula)
>>>>
>>>> Duncan Murdoch
>>> Hi Duncan,
>>>
>>> thanks for your quick reply. I think that is not the problem, because
>>> with glm or lm the formula inside the new.form work, only on update it
>>> dont work.
>>>
>>> But I try this solution, I put new.form enclosed in an environment:
>>>
>>> environment(new.form) <-
>>> environment(as.formula(gsub(qvar,"qvartemp",as.expression(mma$formula))))
>>>
>>> But the error is the same.
>> Of course it is.  That operation does nothing, since the value of
>> new.form was already
>> as.formula(gsub(qvar,"qvartemp",as.expression(mma$formula))).
>>
>> If you want any more help, you're going to have to provide a
>> reproducible example.
>>
>> Duncan Murdoch
> 
> Duncan,
> 
> sorry, I dont understand how to make this.
> 
> I send for you a full code with an example that work and other that dont work.

This is a little more complicated than I thought.  The problem is that 
update() uses the environment of the old formula, and ignores the new 
environment.  So you need to sneak your qvartemp into visibility in the 
old formula:

       new.form <- 
as.formula(gsub(qvar,"qvartemp",as.expression(mma$formula)))
       environment(mma$formula) <- new.env(parent = 
environment(mma$formula))
       environment(mma$formula)$qvartemp <- qvartemp

       print(new.form)

       mmaa <- update(mma,new.form)

This appears to work.  It looks ugly; I don't know if there's a cleaner 
way to achieve what you want, i.e. to update a formula to use a variable 
that's defined locally in a function.

Duncan Murdoch