On Feb 24, 3:17 pm, William Stein <wst...@gmail.com> wrote: [reason for current design] > But what was it? Since he is MIA, we might not ever know...
When you look a little at how Python's random.Random is designed, his design seems a fairly clean solution. Whether it leads to a convenient interface is another question, of course. In particular from his description here: http://www.sagemath.org/doc/reference/sage/misc/randstate.html#generating-random-numbers-in-library-code it is clear that he was intending to get to a point where the "random state" of all of sage's components could be reflected in a single object. This allows not just a convenient way of globally forcing the system into a well-defined seed state, but also allows saving and restoring that state and make context managers (he already did) that allow random number generation without affecting the seed state outside. Implementing this consistently requires some discipline on the side of sage's library, as he discusses. Python's random.Random objects essentially already implement the infrastructure for this. So to capture python's random state, the obvious solution is to make a random.Random object part of sage's randstate. Sage's "global" current_randstate() is simply an instance of Sage's general randstate. Python's random module maintains its own "global" instance in random._inst and random.randint is simply a rebinding of random._inst.randint. Sage's current_randstate().python_random() is not that object. If you absolutely want to use random.randint *AND* ensure that set_random_seed affects it, you could include a random.setstate in set_random_seed. In that case, you probably want to ensure that you also set _current_randstate._python_random = random._inst #you can't do this literally This probably requires that the randstate constructor takes a flag parameter "please make the python random component the global python one". That way, the global sage randstate would really incorporate the global python random._inst. -- To post to this group, send an email to sage-devel@googlegroups.com To unsubscribe from this group, send an email to sage-devel+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URL: http://www.sagemath.org
