Original comment by Anonymous.

Possible fix by editing rpy2py_basic code to read as follows:

#!python

def rpy2py_basic(obj):    
    if hasattr(obj, '__len__'):
        if obj.typeof in [ri.INTSXP, ri.REALSXP, ri.CPLXSXP, 
                            ri.LGLSXP, ri.STRSXP]:
            res = [x for x in obj]
        elif obj.typeof in [ri.VECSXP]:
            if r.names(obj) != []:
                res = dict([(r.names(obj)[i], rpy2py(obj[i])) for i in range(len(obj))])

            else:
                res = [rpy2py(x) for x in obj]
        elif obj.typeof in [ri.LANGSXP]:
            res = Robj(obj)
        else:
            raise ValueError("Invalid type %i (%s) for 'obj'." %(obj.typeof, ri.str_typeint(obj.typeof)))
    else:
        res = Robj(obj)
    return res

from rpy2.

Original comment by Laurent Gautier (Bitbucket: lgautier, GitHub: lgautier).

In R, names are not contrained to be unique (whether we like it or not).

Converting they to dict the way it is proposed can lead to very sneaky problems (as duplicated names will just vanish), and I am seriously considering to break compatibility with rpy here.

This is probably worth being brought to the list for discussion. Should conversion to a dict:

a- make lists of elements associated with each key / name ?

b- fail whenever names are duplicated ?

c- be to an ordered dictionnary (so the order is preserved) ?

(I am currently leaning for c)

from rpy2.

Original comment by Anonymous.

I agree that 'c' is probably the best option. I'm not how rpy overcame the problem (if at all), but it's probably worth looking into further.

from rpy2.

Original comment by Laurent Gautier (Bitbucket: lgautier, GitHub: lgautier).

Option c/ is still holding details to be ironed out.

When a value is queried on its position that's easy, but what will be returned when a value is queried on its key and the key is present at several locations ? If this is a list, then it would be homogeneous to have a list returned in all cases (list of length one if the key is present only once).

PS: AFAIR, rpy will silently override duplicates (to only keep the last one).

from rpy2.

Original comment by Laurent Gautier (Bitbucket: lgautier, GitHub: lgautier).

No reaction on the mailing-list regarding the which way to go.

I made the conversion from R list to Python dict happening under the following two conditions:

The list has names (if not, no conversion is performed)

The list names are unique (if no an exception ValueError is raised).

It committed in:
<<changeset 948:8fdd1ee9ccef>>

from rpy2.

Original comment by Laurent Gautier (Bitbucket: lgautier, GitHub: lgautier).

Removing version: 2.1.0 (automated comment)

from rpy2.