On Tue, Apr 17, 2012 at 10:11 AM, Igor Stasenko <[email protected]> wrote:
> I think that we missing a way to determine to which package some > global belongs to. > When you loading a package it can declare new global, but there is no > reflective mechanisms in system to tell, to which package it belongs. > If we could have that, then clearly, if you serialize the package, you > would want to serialize its globals as well, > and if not, you can just leave a reference to it, which will be > rewired at materialization. > +1. Some construct that defines a module, analogous to a class definition? > > On 17 April 2012 18:20, Eliot Miranda <[email protected]> wrote: > > > > > > On Tue, Apr 17, 2012 at 3:50 AM, Mariano Martinez Peck > > <[email protected]> wrote: > >> > >> Hi guys. In Fuel, we need to identity which globals present in Smalltalk > >> globals are NOT classes or traits, because we need to manage them as > >> globals. So far we have a hardcoded list: > >> > >> defaultGlobalSymbols > >> > >> ^ #(#Smalltalk #SourceFiles #Transcript #Undeclared #Display > >> #TextConstants #ActiveWorld #ActiveHand #ActiveEvent #Sensor #Processor > >> #ImageImports #SystemOrganization #World) > >> > >> So...as every hardcoded stuff, this is not nice because as soon as > globals > >> change, we need to fix this. So it would be nice if the system (pharo) > can > >> provide us this information. > >> > >> What do you think? > > > > > > What I think is that the set of globals depends on usage. Some time I > might > > actually want to serialize the SourceFiles Array, sometimes I might not > want > > to. So this should be configurable, with a sensible default. Just as > when > > one serializes a set of classes, references to classes outside the set > are > > not serialized, serializing some subset of global objects in Smalltalk > > should serialize those objects, not references. So design some > convenient > > API that allows one to specify which objects are not globals, and have > the > > default be e.g. all global objects in Smalltalk. > > > >> > >> > >> > >> -- > >> Mariano > >> http://marianopeck.wordpress.com > >> > > > > > > > > -- > > best, > > Eliot > > > > > > -- > Best regards, > Igor Stasenko. > > -- best, Eliot
