Skip to content

problem with assign and get

4 messages · Jannis, Peter Langfelder, Sarah Goslee +1 more

Original
Jannis
# 
Dear list members,


does anyone have an idea why the following construction does not work 
but gives the following error message:

assign('test', array(1:10, dim=c(10,10)))
dimnames(get('test')) <- list(1:10,1:10)


Error in dimnames(get("test")) <- list(1:10, 1:10) :
   target of assignment expands to non-language object


What could be a way to get this to work?


thanks a lot
Jannis
Peter Langfelder
#
On Mon, Feb 27, 2012 at 9:41 AM, Jannis <bt_jannis at yahoo.de> wrote:
dimnames(test) <- list(1:10,1:10)

Peter
Sarah Goslee
# 
Hi,
On Mon, Feb 27, 2012 at 12:41 PM, Jannis <bt_jannis at yahoo.de> wrote:
There's no object to assign dimnames to. Think about it
this way - what do you expect the dimnames of to be
changed? "test" has no dimnames; get("test") would
print the object named "test", but how could you change
the dimnames of a displayed object?

If you want to change the dimnames of the object named
"test", you first need to use get("test") to assign that object
to a new object, then change the dimnames, then assign
the changed object to the desired name.

myobject <- get("test")
dimnames(myobject) <- list(1:10, 1:10)
assign("test", myobject)

Note: I'm assuming your small reproducible example (thank
you!) is a surrogate for something more complex, so that
dimnames(test) <- list(1:10, 1:10) is not an acceptable
solution. But frequently there are better options than the
use of get/assign for solving particular problems.

Sarah
William Dunlap
# 
Sarah Gosless said
I tend to be more extreme say recommend that you forget
you ever heard of the get and assign functions.

First, out-of-the-local-evaluation-environment assignments
make for confusing code.  You can usually do what you
need to do by having functions that return values.
Replacement functions, defined with
   `something<-` <- function(x, ..., value) { x$something<-value; x }
and used as
   something(x) <- newSomething
can be used to alter an existing object.

Second, if you decide that you need to do an assignment
outside of the current evironment, you ought to know which
environment you want to use.  It is often .GlobalEnv.
Then you can use
   yourEnvironment[["itemName"]]
instead of
   get("itemName", envir=yourEnvironment)
and
   yourEnvironment[["itemName"]] <- itemValue
instead of
   assign("itemName", itemValue, envir=yourEnvironment)
The advantages of the [[ syntax are (a) it forces you
to decide where the data should be and (b) it lets you
use natural R syntax to modify objects.  E.g.,
   rownames(yourEnvironment[["itemName"]]) <- c("R1","R2")
will change the rownames of "itemName" in the environment
called yourEnvironment.

At somepoint you may decide to use a list instead of
an environment to hold your objects, in which case the
[[ syntax still works without any changes.

Bill Dunlap
Spotfire, TIBCO Software
wdunlap tibco.com