Proposal: Automatic Resource Management (original) (raw)

Stefan Schulz schulz at e-spirit.de
Sat Mar 7 09:51:21 PST 2009

Stephen Colebourne wrote:

I think this is understood, however the reality is that it will be. As soon as the ARM code is in the compiler, an open source project will appear with code to wrap other classes to make them work with the ARM block.

Of course that doesn't mean ARM is good or bad, just that we should be realistic about how widely used it will be.

I completely agree. Whatever tool a programmer is given, he or she will apply it for good and bad.

However, with closures officially defined as 'out' I have to choose between solving real issues now (in the 80% case) or waiting another 3 years+. I choose pragmatism - solve the 80% now with ARM, and reconsider BGGA/JCA later.

To me, ARM does not come even near 80% of the cases I would like to see being solved and what BGGA/JCA could provide. It's a somewhat the least common denominator of the desired solution. Together with for-each, it will introduce yet another language level control for solving a niche of problems. (That does not mean, it's a bad idea.)

In the discussion about whether to use a language level (finally) or API level (interface) solution, the siginficant problem I see is the reuse of an existing interface/signature. As some pointed out, existing classes that have or comply to such an interface may not be ready for this kind of usage within a try() statement, although they can be applied. My first reaction was to prefer the language level solution, as no existing class can be used without re-considering its use in that new statement. But a newly introduced interface having a method most likely not matching existing ones for resources would nearly have the same effect. I would also oppose to allow multiple interfaces because of the above reasons.

Stefan

More information about the coin-dev mailing list