Re: [j-devel] format and gray streams

[XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX]

On Thu, Jun 11, 2009 at 8:58 AM, Alessio Stalla<ale...@gm...> wrote:
> On Wed, Jun 10, 2009 at 10:46 PM, XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX wrote:
>> The solution however, seems to be available: eliminate the use of
>> temp-files for COMPILE. That way, the use of CLOSE should not be
>> required: we'll store the content of the temp file in-memory and load
>> the class from there. Alessio has already done some prep work for it.
>> I / we just need to hook that up.
>>
>> So, the next priority would be to eliminate temp files in the
>> compiler. Who wants to take a stab at it?
>
> I'm still interested in eliminating the use of temp-files in the
> compiler; I've just been busy with other things lately.

No problem! I have been too: when you make a toolkit, you need 'to eat
your own dogfood' to understand the strength and limitations of what
you're doing. IOW: I was creating a program *using* abcl.

> I have a couple of ideas which I'd like to try soon.
> However, is eliminating temp files sufficient for not using close
> anymore?

Well, ofcourse does elimination of temp files in COMPILE not eliminate
the use of CLOSE at all, but it eliminates the circular dependency
where compilation of CLOSE requires a call to CLOSE: upon compilation
to memory no CLOSE is required: the bytes will be moved into the
target buffer.

> What about compile-file?

It still uses CLOSE, but the point is not to eliminate the use of
CLOSE entirely: the point is to eliminate the circular dependency.
When using CLOSE the first time, an effective method is compiled. This
compilation uses temp files, currently. Hence the circular dependency.
Now, if we were to compile this CLOSE method to memory, no further
issues would occur there.


> Bye,
> Alessio


Bye,


Erik - who hopes to make sense today.