Subject Re: [IB-Architect] Embedded compiler?
Author adem
> At first, FPC isn't so very good compiler yet. Second, it is
> written in Pascal while FB is written in C (�++ here and there).
> I guess you are a programmer. Are you ready to do it (to embed FPC
> into FB)?

Yes I am a programmer, and no I am not ready to implement it yet;
that is whay I choose to ask the question here, in the ib-architect,
and not in the fb-devel. This was not a challenge question.

Plus, I can see your efforts on fb which I apreciate a lot. So my
perspective on the issue was after C++ conversion.

Thirdly, why do you say FPK is not a very good compiler yet?

> >-- We would not have to do mental switch everytime we need to write a
> >stored procedure or a trigger in FB. And, for those who are familiar
> >with Delphi/Kylix, learning curve would be a breeze too.
> If one's mental is so hard to switch, hire another person specially
> for writing SP/Triggers.

Again, instead of looking at it from short-sharp-answers geek angle,
dont you think there an element of thruth in what I said?

> >-- We would not need UDFs that need to be compiled for every new
> >platform, FB could compile them by itsef.
> Another step: let FB writes UDFs itself! :)

Of course you are a much better programmer than most, with more time
than most too; so you can port all your UDFs to whatever platform you
need to. Yet there is the mortals among us who would like this to be as
transparent as possible. Is it too much to consider?

> >Was the reason not to do so something like this a historic one
> >(i.e. nothing suitable vailable at the time) or something deeper than
> >that?
> Ordinary programming languages don't have operators to manage DB
> objects. So every language must be "adapted" to work as SP language.
> Usually DBMS developers don't want to spend time to implement
> compiler of "complex" language instead of concentrating on improving
> DBMS engine. And even if they do (in case of IBM and Oracle) results
> aren't impressive.

Improving the DBMS engine should not be mutually exclusive. The reason
why I put Pascal forward was because FB was very Delphi (Kylix) friendly.
And, that it might be a good idea to get them closer still.

And I agree that DBMS developers need not spend time to implement
complex xompilers; but that should not mean they have to object to the
idea on that premise alone.