From: Tom Sawyer <transami@...>
Date: 2003-01-05T06:03:06+09:00
Subject: Re: dump depth argument

On Saturday 04 January 2003 06:11 am, Dossy wrote:
> When using the version of Marshal#dump that takes anIO as an arg, I
> noticed that it /does/ write to anIO up until the limit is reached.  So,
> if you rescue the exception, what's contained in anIO should be the
> Marshalled result until the limit was hit.  However, this won't do the
> right thing in all cases: it seems to hard-stop at the /first/ member
> variable that exceeds the depth.

yes, i tried it and it bombs out. too bad it dosen't do all 1st level 
instances before going on the 2nd, etc.

> Marshalling out an object with a soft limit is a bad idea, because the
> subsequent Marshal#load won't produce an identical object as the one
> that was Marshal#dump'ed.

that's okay if you know exactly what to expect. so the option wouldn't hurt.

> I always thought that one of the required properties of pushing an
> object through Marshal#dump then Marshal#load is that it produces a deep
> copy of the original object ...

as long as cyclic references stay intact. the problem i was getting, was tthat 
marshal was actually copying cyclic references. according to matz that 
shouldn't happen. and subsequently the problem seems to have gone away. i 
wish i knew what had casued it....

-transami