R devel - May 2015 - C-API: check whether R has been initialized ?

Beside the possible argumentation that with an API elegance and convenience
might sometimes be superior to necessity, the suggested pattern ("every
program, including R itself, keeping its own flag") does no work too well
when the nested embedding of R is involved.

A concrete example is:
```
$ R -q
> library('rPython'); python.exec('import rpy2.robjects')
R is already initialized
```

https://bitbucket.org/rpy2/rpy2/issue/278/r-in-python-via-rpy2-in-r-via-rpython#comment-17843761

2015-05-03 18:12 GMT-04:00 Duncan Murdoch <murdoch.duncan at gmail.com>:
> On 03/05/2015 4:34 PM, Laurent Gautier wrote:
> > Hi,
> >
> > There appear to be no way to check whether R has already been
> initialized.
> >
> > Could a function like "Rf_isinitialized" be added to the API
?
> >
>
> Surely any program that needs to know that could keep its own flag.
> You'll need to give a much longer argument about why this is necessary
> for what you're doing.
>
> Duncan Murdoch
>
>
	[[alternative HTML version deleted]]

On 03/05/2015 7:02 PM, Laurent Gautier wrote:
> Beside the possible argumentation that with an API elegance and
> convenience might sometimes be superior to necessity, the suggested
> pattern ("every program, including R itself, keeping its own
flag") does
> no work too well when the nested embedding of R is involved.
> 
> A concrete example is:
> ```
> $ R -q
>> library('rPython'); python.exec('import rpy2.robjects')
> R is already initialized
> ```
I don't know rPython at all, but surely this is an rPython bug.  When
the package is loaded by "library('rPython')", R is obviously
initialized.  You don't need to query it to ask that.

The standard R front-ends don't need a flag to know if it is
initialized.  They initialize, then go into the read-eval-print loop.
If they are in that loop, R is initialized.  If it failed to initialize,
they would never get to that loop.

Other front-ends may do other things besides run R, so they do need to
know if it is initialized, but surely they can keep a flag telling them
whether they've succeeded in initializing it.

Duncan Murdoch

rPython appears to provide an interface from R to Python by embedding
Python and I'd think that it can safely assume that R has been initialized,
but might not be the point here.

The issue is that a Python package embedding itself R (here rpy2) appears
to have no way to know that earlier in the life of the process R was
initialized.

2015-05-03 19:48 GMT-04:00 Duncan Murdoch <murdoch.duncan at gmail.com>:
> On 03/05/2015 7:02 PM, Laurent Gautier wrote:
> > Beside the possible argumentation that with an API elegance and
> > convenience might sometimes be superior to necessity, the suggested
> > pattern ("every program, including R itself, keeping its own
flag") does
> > no work too well when the nested embedding of R is involved.
> >
> > A concrete example is:
> > ```
> > $ R -q
> >> library('rPython'); python.exec('import
rpy2.robjects')
> > R is already initialized
> > ```
>
> I don't know rPython at all, but surely this is an rPython bug.  When
> the package is loaded by "library('rPython')", R is
obviously
> initialized.  You don't need to query it to ask that.
>
> The standard R front-ends don't need a flag to know if it is
> initialized.  They initialize, then go into the read-eval-print loop.
> If they are in that loop, R is initialized.  If it failed to initialize,
> they would never get to that loop.
>
> Other front-ends may do other things besides run R, so they do need to
> know if it is initialized, but surely they can keep a flag telling them
> whether they've succeeded in initializing it.
>
> Duncan Murdoch
>
	[[alternative HTML version deleted]]