R devel - Aug 2020 - trace creates object in base namespace if called on function argument

Please note that this is documented in ?trace. "fun" is matched to
what,
it is a _name_ of the function to be traced, which is traced in the 
top-level environment. I don't know why it was designed this way, but it 
is documented in detail, and hence the expected behavior.

Debugging is often, and also in R, implemented in the core. Tracing is 
implemented on top without specific support, it thus cannot do some 
things debugging can do.

Tomas


On 8/26/20 3:31 AM, Antoine Fabri wrote:
> Apologies there is one line missing in my last email, the code should be :
>
> foo <- function() "hello"
> trace2 <- function(fun) trace(fun, quote(print("!!!")))
> trace2(foo) # <- THIS LINE WAS MISSING
> base::fun
>
> Best,
>
> Antoine
>
> Le mar. 25 ao?t 2020 ? 22:02, Antoine Fabri <antoine.fabri at
gmail.com> a
> ?crit :
>
>> Dear R-devel,
>>
>> I don't think this is expected :
>>
>> foo <- function() "hello"
>> trace2 <- function(fun) trace(fun, quote(print("!!!")))
>> base::fun
>> # Object with tracing code, class "functionWithTrace"
>> # Original definition:
>> # function() "hello"
>> #
>> # ## (to see the tracing code, look at body(object))
>>
>> `untrace()` has the same behavior.
>>
>> This is inconsistent with how debug works :
>>
>> foo <- function() "hello"
>> debug2 <- function(fun) debug(fun)
>> debug2(foo)
>> isdebugged(foo)
>> # [1] TRUE
>>
>> This can be worked around by defining :
>>
>> trace2 <- function(fun) eval.parent(substitute(trace(fun,
>> quote(print("!!!")))
>>
>> but I believe the current behavior is undesired and it'd be better
to make
>> it behave as `debug()`, or to throw an error.
>>
>> Best,
>>
>> Antoine
>>
> 	[[alternative HTML version deleted]]
>
> ______________________________________________
> R-devel at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel

Hi Tomas,

The doc indeed describes `what` as "the name, possibly quote()d, of a
function to be traced or untraced".
This is a good argument not to change the function and make it behave more
like debug.

However the doc also tells us "A call to trace allows you to insert
debugging code (e.g., a call to browser or recover) at chosen places in any
function" and "The trace function operates by constructing a revised
version of the function (or of the method, if signature is supplied), and
assigning the new object back where the original was found". In the issue I
report a function was created in the base environment, which is not "where
the function was found". It's not clear to me what "where it was
found"
would mean in that case, but I would assume the execution environment of
trace2, or maybe the calling environment (global in that case), or parent
of foo (global here again), but not the base namespace.

It would make sense to me to either have a traced fun in the execution
environment, or to have an error.

Best,

Antoine

Le mer. 26 ao?t 2020 ? 14:17, Tomas Kalibera <tomas.kalibera at gmail.com>
a
?crit :
> Please note that this is documented in ?trace. "fun" is matched
to what,
> it is a _name_ of the function to be traced, which is traced in the
> top-level environment. I don't know why it was designed this way, but
it
> is documented in detail, and hence the expected behavior.
>
> Debugging is often, and also in R, implemented in the core. Tracing is
> implemented on top without specific support, it thus cannot do some
> things debugging can do.
>
> Tomas
>
>
> On 8/26/20 3:31 AM, Antoine Fabri wrote:
> > Apologies there is one line missing in my last email, the code should
be
> :
> >
> > foo <- function() "hello"
> > trace2 <- function(fun) trace(fun, quote(print("!!!")))
> > trace2(foo) # <- THIS LINE WAS MISSING
> > base::fun
> >
> > Best,
> >
> > Antoine
> >
> > Le mar. 25 ao?t 2020 ? 22:02, Antoine Fabri <antoine.fabri at
gmail.com> a
> > ?crit :
> >
> >> Dear R-devel,
> >>
> >> I don't think this is expected :
> >>
> >> foo <- function() "hello"
> >> trace2 <- function(fun) trace(fun,
quote(print("!!!")))
> >> base::fun
> >> # Object with tracing code, class "functionWithTrace"
> >> # Original definition:
> >> # function() "hello"
> >> #
> >> # ## (to see the tracing code, look at body(object))
> >>
> >> `untrace()` has the same behavior.
> >>
> >> This is inconsistent with how debug works :
> >>
> >> foo <- function() "hello"
> >> debug2 <- function(fun) debug(fun)
> >> debug2(foo)
> >> isdebugged(foo)
> >> # [1] TRUE
> >>
> >> This can be worked around by defining :
> >>
> >> trace2 <- function(fun) eval.parent(substitute(trace(fun,
> >> quote(print("!!!")))
> >>
> >> but I believe the current behavior is undesired and it'd be
better to
> make
> >> it behave as `debug()`, or to throw an error.
> >>
> >> Best,
> >>
> >> Antoine
> >>
> >       [[alternative HTML version deleted]]
> >
> > ______________________________________________
> > R-devel at r-project.org mailing list
> > https://stat.ethz.ch/mailman/listinfo/r-devel
>
>
>
	[[alternative HTML version deleted]]

Hi Antoine,

ok, I thought you were reporting that

 > foo <- function() "hello"
 > trace2 <- function(fun) trace(fun, quote(print("!!!")))
 > trace2(foo)
 > base::fun

in error did not trace "foo" in the top-level environment. This is, 
however, expected, because "trace" is called with argument fun, so the
name of the function binding to replace is "fun" (not
"foo"), and "fun"
does not exist in the top-level environment. You cannot wrap "trace" 
like this.

If you meant to report that instead of throwing an error that an object 
named "fun" was not found, this example inserts the instrumented code 
into "base", that I also find surprising. The behavior I would have 
expected is like with primitive tracing

 > foo <- function() "hello"
 > trace2 <- function(fun) trace(fun)
 > trace2(foo)
 > base::fun
Error in get(name, envir = ns, inherits = FALSE) : object 'fun' not
found

If you are interested in this problem, feel free to investigate in more 
detail (validate it is a bug via reading the documentation in detail, 
reading the code, debugging, checking the commit history) and if it 
turns out to be really a bug, feel free to submit a well-tested minimal 
patch, via bugzilla.

Thanks,
Tomas


On 8/26/20 4:39 PM, Antoine Fabri wrote:
> Hi Tomas,
>
> The doc indeed describes `what` as "the name, possibly quote()d, of a 
> function to be traced or untraced".
> This is a good argument not to change the function and make it behave 
> more like debug.
>
> However the doc also tells us "A call to trace allows you to insert 
> debugging code (e.g., a call to browser or recover) at chosen places 
> in any function" and "The trace function operates by constructing
a
> revised version of the function (or of the method, if signature is 
> supplied), and assigning the new object back where the original was 
> found". In the issue I report a function was created in the base 
> environment, which is not "where the function was found".
It's not
> clear to me what "where it was found" would mean in that case,
but I
> would assume the execution environment of trace2, or maybe the calling 
> environment (global in that case), or parent of foo (global here 
> again), but not the base namespace.
>
> It would make sense to me to either have a traced fun in the execution 
> environment, or to have an error.
>
> Best,
>
> Antoine
>
> Le?mer. 26 ao?t 2020 ??14:17, Tomas Kalibera <tomas.kalibera at
gmail.com
> <mailto:tomas.kalibera at gmail.com>> a ?crit?:
>
>     Please note that this is documented in ?trace. "fun" is
matched to
>     what,
>     it is a _name_ of the function to be traced, which is traced in the
>     top-level environment. I don't know why it was designed this way,
>     but it
>     is documented in detail, and hence the expected behavior.
>
>     Debugging is often, and also in R, implemented in the core.
>     Tracing is
>     implemented on top without specific support, it thus cannot do some
>     things debugging can do.
>
>     Tomas
>
>
>     On 8/26/20 3:31 AM, Antoine Fabri wrote:
>     > Apologies there is one line missing in my last email, the code
>     should be :
>     >
>     > foo <- function() "hello"
>     > trace2 <- function(fun) trace(fun,
quote(print("!!!")))
>     > trace2(foo) # <- THIS LINE WAS MISSING
>     > base::fun
>     >
>     > Best,
>     >
>     > Antoine
>     >
>     > Le mar. 25 ao?t 2020 ? 22:02, Antoine Fabri
>     <antoine.fabri at gmail.com <mailto:antoine.fabri at
gmail.com>> a
>     > ?crit :
>     >
>     >> Dear R-devel,
>     >>
>     >> I don't think this is expected :
>     >>
>     >> foo <- function() "hello"
>     >> trace2 <- function(fun) trace(fun,
quote(print("!!!")))
>     >> base::fun
>     >> # Object with tracing code, class
"functionWithTrace"
>     >> # Original definition:
>     >> # function() "hello"
>     >> #
>     >> # ## (to see the tracing code, look at body(object))
>     >>
>     >> `untrace()` has the same behavior.
>     >>
>     >> This is inconsistent with how debug works :
>     >>
>     >> foo <- function() "hello"
>     >> debug2 <- function(fun) debug(fun)
>     >> debug2(foo)
>     >> isdebugged(foo)
>     >> # [1] TRUE
>     >>
>     >> This can be worked around by defining :
>     >>
>     >> trace2 <- function(fun) eval.parent(substitute(trace(fun,
>     >> quote(print("!!!")))
>     >>
>     >> but I believe the current behavior is undesired and it'd
be
>     better to make
>     >> it behave as `debug()`, or to throw an error.
>     >>
>     >> Best,
>     >>
>     >> Antoine
>     >>
>     >? ? ? ?[[alternative HTML version deleted]]
>     >
>     > ______________________________________________
>     > R-devel at r-project.org <mailto:R-devel at r-project.org>
mailing list
>     > https://stat.ethz.ch/mailman/listinfo/r-devel
>
>

	[[alternative HTML version deleted]]