R devel - Oct 2016 - How to assign NULL value to pairlist element while keeping it a pairlist?

Hi Henrik,

It would help to understand your use case for pairlists.

Thanks,
Michael

On Wed, Oct 12, 2016 at 9:40 AM, Michael Lawrence <michafla at gene.com>
wrote:
> The coercion is probably the most viable workaround for now, as it's
> consistent with what happens internally for calls. All pairlists/calls
> are converted to list for subassignment, but only calls are converted
> back. My guess is that the intent was for users to move from using a
> pairlist to the "new" (almost 20 years ago) list. In my opinion,
> consistency trumps "convenience" in this case. If others agree,
I'll
> change it to also coerce back to pairlist.
>
> Michael
>
> On Wed, Oct 12, 2016 at 9:20 AM, Henrik Bengtsson
> <henrik.bengtsson at gmail.com> wrote:
>> Hi, I seem to not be able to assign NULL to an element of a pairlist
>> without causing it to be coerced to a plain list.  For example:
>>
>>> x <- pairlist(1, 2)
>>> class(x)
>> [1] "pairlist"
>>
>>> x[1] <- list(NULL)
>>> class(x)
>> [1] "list"
>>
>> This actually true for all [()<- assignments regardless of list
value, e.g.
>>
>>> x <- pairlist(1, 2)
>>> x[1] <- list(0)
>> [1] "list"
>>
>> I also tried assigning a pairlist(), but still the same problem:
>>
>>> x <- pairlist(1, 2)
>>> x[1] <- pairlist(0)
>> [1] "list"
>>
>> The only workaround I'm aware of is to:
>>
>> x <- as.pairlist(x)
>>
>> at the end.  Any other suggestions?
>>
>> Thanks,
>>
>> Henrik
>>
>> ______________________________________________
>> R-devel at r-project.org mailing list
>> https://stat.ethz.ch/mailman/listinfo/r-devel

Michael, thanks for this info.

I've stumbled upon this in a case where I walk an R expression (the
AST) and (optionally) modifies it (part of the globals package).  In R
expressions, a function definition uses a pairlist to represent the
arguments.  For example,
> expr <- quote(function(x = 1) x)
> str(as.list(expr))
List of 4
 $ : symbol function
 $ :Dotted pair list of 1
  ..$ x: num 1
 $ : symbol x
 $ :Class 'srcref'  atomic [1:8] 1 15 1 29 15 29 1 1
.. ..- attr(*, "srcfile")=Classes 'srcfilecopy',
'srcfile'
<environment: 0x13918b8>

Here the 2nd element is a pairlist:
> str(expr[[2]])
Dotted pair list of 1
 $ x: num 1
> typeof(expr[[2]])
[1] "pairlist"

Now say that I want to update the default value of argument 'x', which
is currently 1, to NULL.  Then I do:
> expr[[2]][1] <- list(x = NULL)
At this step, I end up with an expression 'expr' where the arguments
are no longer represented by a pairlist:
> str(expr[[2]])
List of 1
 $ x: NULL
> typeof(expr[[2]])
[1] "list"

More importantly, at this point 'expr' no longer holds a valid R
expression:
> expr
Error: badly formed function expression

The solution is to make sure we have a pairlist:
> expr[[2]] <- as.pairlist(expr[[2]])
> expr
function(x = NULL) x


I agree it would be nice to fix this for consistency, but if you bump
into major issues, at least I can live with having to use an explicit
as.pairlist().

Thanks

Henrik

On Wed, Oct 12, 2016 at 10:53 AM, Michael Lawrence
<lawrence.michael at gene.com> wrote:
> Hi Henrik,
>
> It would help to understand your use case for pairlists.
>
> Thanks,
> Michael
>
> On Wed, Oct 12, 2016 at 9:40 AM, Michael Lawrence <michafla at
gene.com> wrote:
>> The coercion is probably the most viable workaround for now, as
it's
>> consistent with what happens internally for calls. All pairlists/calls
>> are converted to list for subassignment, but only calls are converted
>> back. My guess is that the intent was for users to move from using a
>> pairlist to the "new" (almost 20 years ago) list. In my
opinion,
>> consistency trumps "convenience" in this case. If others
agree, I'll
>> change it to also coerce back to pairlist.
>>
>> Michael
>>
>> On Wed, Oct 12, 2016 at 9:20 AM, Henrik Bengtsson
>> <henrik.bengtsson at gmail.com> wrote:
>>> Hi, I seem to not be able to assign NULL to an element of a
pairlist
>>> without causing it to be coerced to a plain list.  For example:
>>>
>>>> x <- pairlist(1, 2)
>>>> class(x)
>>> [1] "pairlist"
>>>
>>>> x[1] <- list(NULL)
>>>> class(x)
>>> [1] "list"
>>>
>>> This actually true for all [()<- assignments regardless of list
value, e.g.
>>>
>>>> x <- pairlist(1, 2)
>>>> x[1] <- list(0)
>>> [1] "list"
>>>
>>> I also tried assigning a pairlist(), but still the same problem:
>>>
>>>> x <- pairlist(1, 2)
>>>> x[1] <- pairlist(0)
>>> [1] "list"
>>>
>>> The only workaround I'm aware of is to:
>>>
>>> x <- as.pairlist(x)
>>>
>>> at the end.  Any other suggestions?
>>>
>>> Thanks,
>>>
>>> Henrik
>>>
>>> ______________________________________________
>>> R-devel at r-project.org mailing list
>>> https://stat.ethz.ch/mailman/listinfo/r-devel

Thanks, this was what I expected. There is a desire to eliminate the
usage of pairlist from user code, which suggests the alternative of
allowing for function arguments to be stored in lists. That's a much
deeper change though.

On Wed, Oct 12, 2016 at 12:31 PM, Henrik Bengtsson
<henrik.bengtsson at gmail.com> wrote:
> Michael, thanks for this info.
>
> I've stumbled upon this in a case where I walk an R expression (the
> AST) and (optionally) modifies it (part of the globals package).  In R
> expressions, a function definition uses a pairlist to represent the
> arguments.  For example,
>
>> expr <- quote(function(x = 1) x)
>> str(as.list(expr))
> List of 4
>  $ : symbol function
>  $ :Dotted pair list of 1
>   ..$ x: num 1
>  $ : symbol x
>  $ :Class 'srcref'  atomic [1:8] 1 15 1 29 15 29 1 1
> .. ..- attr(*, "srcfile")=Classes 'srcfilecopy',
'srcfile'
> <environment: 0x13918b8>
>
> Here the 2nd element is a pairlist:
>
>> str(expr[[2]])
> Dotted pair list of 1
>  $ x: num 1
>> typeof(expr[[2]])
> [1] "pairlist"
>
> Now say that I want to update the default value of argument 'x',
which
> is currently 1, to NULL.  Then I do:
>
>> expr[[2]][1] <- list(x = NULL)
>
> At this step, I end up with an expression 'expr' where the
arguments
> are no longer represented by a pairlist:
>
>> str(expr[[2]])
> List of 1
>  $ x: NULL
>> typeof(expr[[2]])
> [1] "list"
>
> More importantly, at this point 'expr' no longer holds a valid R
expression:
>
>> expr
> Error: badly formed function expression
>
> The solution is to make sure we have a pairlist:
>
>> expr[[2]] <- as.pairlist(expr[[2]])
>> expr
> function(x = NULL) x
>
>
> I agree it would be nice to fix this for consistency, but if you bump
> into major issues, at least I can live with having to use an explicit
> as.pairlist().
>
> Thanks
>
> Henrik
>
> On Wed, Oct 12, 2016 at 10:53 AM, Michael Lawrence
> <lawrence.michael at gene.com> wrote:
>> Hi Henrik,
>>
>> It would help to understand your use case for pairlists.
>>
>> Thanks,
>> Michael
>>
>> On Wed, Oct 12, 2016 at 9:40 AM, Michael Lawrence <michafla at
gene.com> wrote:
>>> The coercion is probably the most viable workaround for now, as
it's
>>> consistent with what happens internally for calls. All
pairlists/calls
>>> are converted to list for subassignment, but only calls are
converted
>>> back. My guess is that the intent was for users to move from using
a
>>> pairlist to the "new" (almost 20 years ago) list. In my
opinion,
>>> consistency trumps "convenience" in this case. If others
agree, I'll
>>> change it to also coerce back to pairlist.
>>>
>>> Michael
>>>
>>> On Wed, Oct 12, 2016 at 9:20 AM, Henrik Bengtsson
>>> <henrik.bengtsson at gmail.com> wrote:
>>>> Hi, I seem to not be able to assign NULL to an element of a
pairlist
>>>> without causing it to be coerced to a plain list.  For example:
>>>>
>>>>> x <- pairlist(1, 2)
>>>>> class(x)
>>>> [1] "pairlist"
>>>>
>>>>> x[1] <- list(NULL)
>>>>> class(x)
>>>> [1] "list"
>>>>
>>>> This actually true for all [()<- assignments regardless of
list value, e.g.
>>>>
>>>>> x <- pairlist(1, 2)
>>>>> x[1] <- list(0)
>>>> [1] "list"
>>>>
>>>> I also tried assigning a pairlist(), but still the same
problem:
>>>>
>>>>> x <- pairlist(1, 2)
>>>>> x[1] <- pairlist(0)
>>>> [1] "list"
>>>>
>>>> The only workaround I'm aware of is to:
>>>>
>>>> x <- as.pairlist(x)
>>>>
>>>> at the end.  Any other suggestions?
>>>>
>>>> Thanks,
>>>>
>>>> Henrik
>>>>
>>>> ______________________________________________
>>>> R-devel at r-project.org mailing list
>>>> https://stat.ethz.ch/mailman/listinfo/r-devel