Lexical scoping is not what I expect

I second Ellison sentiments of "almost never".  One main reason is readability on later
viewing.
Yes, as Duncan says global variables can sometimes be handy and make functions quick to
write, but using a formal argument in the call is always clearer.

Terry Therneau

On 06/27/2013 05:00 AM, r-help-request at r-project.org wrote:

On 13-06-26 6:57 AM, S Ellison wrote:

  -----Original Message-----
  It may be helpful not to worry about the technical details,
  just to look at the source code defining the function:  if it
  is defined in a place where a variable can be seen, it can
  see that variable.

  I too find R's lexical scoping rules straightforward.
  However, I'd say that if your code relies on lexical scoping to find something, you should probably rewrite your code.

  The number of times I've seen new R users get unexpected results because they haven't noticed that their function is referencing a parent environment instead of a locally defined variable or argument is past counting.

  Of course there are times when it's useful and sensible to have globally defined variables that can be accessed within a function. But they are very rare; as a default, I'd recommend avoiding it if at all possible. If your function needs something from outside, pass it as an argument.

I would say the meaning of "probably" in your 2nd sentence depends quite
a bit on the user.  For beginners, it's "almost certainly".  For people
who are comfortable with the concept, it's just "maybe".

I would agree that in most cases the only global objects that functions
should reference are other functions, but small nested functions are
quite safe, and it's sometimes useful to create functions in a local
environment so they have persistent memory.

Duncan Murdoch

On 13-06-27 8:18 AM, Terry Therneau wrote:

I second Ellison sentiments of "almost never".  One main reason is readability on later
viewing.
Yes, as Duncan says global variables can sometimes be handy and make functions quick to
write, but using a formal argument in the call is always clearer.

I didn't say that.  I said that in most cases global variables other than functions 
should not be used.

I would generalize that a little:  global constants are okay, changing globals is 
usually a bad idea.

Duncan Murdoch

Terry Therneau

On 06/27/2013 05:00 AM, r-help-request at r-project.org wrote:

On 13-06-26 6:57 AM, S Ellison wrote:

  -----Original Message-----
  It may be helpful not to worry about the technical details,
  just to look at the source code defining the function:  if it
  is defined in a place where a variable can be seen, it can
  see that variable.

  I too find R's lexical scoping rules straightforward.
  However, I'd say that if your code relies on lexical scoping to find something, 
you should probably rewrite your code.

  The number of times I've seen new R users get unexpected results because they 
haven't noticed that their function is referencing a parent environment instead of a 
locally defined variable or argument is past counting.

  Of course there are times when it's useful and sensible to have globally defined 
variables that can be accessed within a function. But they are very rare; as a 
default, I'd recommend avoiding it if at all possible. If your function needs 
something from outside, pass it as an argument.

I would say the meaning of "probably" in your 2nd sentence depends quite
a bit on the user.  For beginners, it's "almost certainly".  For people
who are comfortable with the concept, it's just "maybe".

I would agree that in most cases the only global objects that functions
should reference are other functions, but small nested functions are
quite safe, and it's sometimes useful to create functions in a local
environment so they have persistent memory.

Duncan Murdoch