Visualworks - Unique strings object?

> Hi,
>
> I have some memory issues with a VW 5i image.
>
> By instance, I found out that for the class ByteString I have 12 megs
> of memory used and for that 12 megs only 6 megs are unique strings.
>
> Is there a way (as JAVA) to tell Smalltalk to "re-use" string objects?
> I do not think so but I would like to be sure.
>
> Is there something else I should do other less that I may have some
> internal issues with the application code.

> Hi,
>
> I have some memory issues with a VW 5i image.
>
> By instance, I found out that for the class ByteString I have 12 megs
> of memory used and for that 12 megs only 6 megs are unique strings.
>
> Is there a way (as JAVA) to tell Smalltalk to "re-use" string objects?
> I do not think so but I would like to be sure.
>
> Is there something else I should do other less that I may have some
> internal issues with the application code.
>
> Regards,
> Michel
>
>

>
>
>
> Hi,
>
>
>
> sorry for breaking the Threadview of your mailclients with this mail,
>
> but I unfortunately already got the ongoing thread deleted and this just
> came to mind. (Oh and sorry for clogging up your space in case you weren't
> interested ;-) )
>
>
>
> If I recall correctly there was a question in the direction whether this can
> be achieved in Smalltalk:
>
> | a b |
>
> a := 'String'.
>
> b := 'String'.
>
> (a = b) = (a == b). "This should answer true but doesn't (just print it)"
>
> Another question came to my mind, would this behaviour sensible? In my
> opinion there are two reasons why we wouldn't want Smalltalk to do this.
>
>
>
> The first one would be, there is already a suitable class for Texts which
> only exist once, as you all pointed out, Symbol is readily available. The
> other reason however is even more important in my opinion, which is
> consistency.
>
>
>
> Why you may ask? Let me illustrate. I wrote all the code in Smalltalk
> Syntax, as I am not sure how many of you know Java and know it well enough.
> For the Java literate, = should be the same as String's equal() instance
> method.
>
> I wrote a small Java App and this is the comparison against Smalltalk:
>
> (Sorry for the ugly code..)
>
> | c1 c2 c all |
>
> c1 := 'te'. c2 := 'st'.
>
> c := c1 , c2.
>
> all := Array
>
> with: 'test' "String literal"
>
> with: 'test' "Same String literal, different object in ST)"
>
> with: 'te' , 'st' "Java: ""te"" + ""st"", just added two literals"
>
> with: c "Create two Strings by literals and
>
> add them like done before this."
>
> with: (String fromString: 'test'). "JAVA: new String(""test" ")"
>
>
>
> all do: [:elem1 | all do:
>
> [:elem2 |
>
> Transcript
>
> show: elem1 == elem2;
>
> space].
>
> Transcript cr]
>
>
>
> Give it a thought, what would you expect for this to yield? Hint: The output
> is formatted like a table which compares each Element against each other
> (actually only half of these are needed as #== is implemented reflexive, but
> what the heck) So all of us Smalltalkers would expect this:
>
> true false false false false
>
> false true false false false
>
> false false true false false
>
> false false false true false
>
> false false false false true
>
> Which is exactly the output generated by smalltalk, very lovely world :-)
>
> If you want, the code can be copied and run directly as typed above.
>
> However if you type exactly (well not exactly) this code in Java you get an
> interestening result being:
>
> true true true false false
>
> true true true false false
>
> true true true false false
>
> false false false true false
>
> false false false false true
>
> Now apparently in Java with this: a = "te"; b = "st";
>
> the results of a + b are not the same object as "te" + "st" which is very
> strange in terms of consistency. Strange as well is that "Test" saved in a
> variable twice is the same object, a new String from the String "Test"
> however is not. I would call that very weird...
>
>
>
> What I had to learn in both languages is, that comparing using the ==
> Operator is bad (most of the time) but in Java, in many cases I can actually
> do that without getting hurt. I like the ST solution better, which is using
> Symbols for unique Strings, the JAVA optimizations may save space and/or
> time, but from the point of OOP I find it very strange and questionable.
>
>
>
> I thank everyone for following me so far,
>
> and btw I would be interested to know what leads to your multiple String
> objects for the same String, if you can say something about the details,
> maybe we can tell you better whether it's feasible to use Symbols instead?
>
>
>
> Greetings,
>
>
> Malte Zacharias
>
>
>
>
>
>
>
>
> On Wednesday 23 August 2006 20:36, you wrote:
>
> > You welcome and thanks to let me know if you found something else.
>
> >
>
> > I know JAVA a little and I am always interested to learn it since
>
> > unfortunately I never had the chance working with it.
>
> >
>
> > But anyway I prefer Smalltalk :)
>
> >
>
> > Michel
>
> >
>
> > On 8/23/06, Malte Zacharias <[hidden email]> wrote:
>
> > > I just conducted a test, and it seems you're right, just FYI..
>
> > > I didn't know that yet, thanks for telling :-)
>
> > >
>
> > > > I am not a JAVA expert but by default :
>
> > > > A:= 'hi'
>
> > > > B:= 'hi' means A and B hold the same object.
>
> > > >
>
> > > > It is when you explicitly use the class String that the string object
>
> > > > are not unique.
>
> > > >
>
> > > > Michel
>
> > > >
>
> > > > On 8/23/06, Malte Zacharias <[hidden email]> wrote:
>
> > > > > > Hi,
>
> > > > > > I have some memory issues with a VW 5i image.
>
> > > > > >
>
> > > > > > By instance, I found out that for the class ByteString I have 12
>
> > > > > > megs of memory used and for that 12 megs only 6 megs are unique
>
> > > > > > strings. Is there a way (as JAVA) to tell Smalltalk to "re-use"
>
> > > > > > string objects? I do not think so but I would like to be sure.
>
> > > > >
>
> > > > > How would one tell Java that? Just curious, you know.
>
> > > > >
>
> > > > > Greetings,
>
> > > > > Malte Zacharias
>
>