On 26 April 2013 10:15, Matthew Adams <matt...@matthewadams.me> wrote:
> On Fri, Apr 26, 2013 at 11:54 AM, Archie Cobbs <arc...@dellroad.org>wrote: > >> On Fri, Apr 26, 2013 at 11:31 AM, Matthew Adams >> <matt...@matthewadams.me>wrote: >> >>> While the Java familiarity afforded by (a) is appealing, the verbosity >>> is a turn-off for me. AspectJ pointcuts already are very terse and a bit >>> cryptic for the uninitiated, so I'm not afraid of just continuing down that >>> road with (b). >>> >> >> Your suggestions in the (b) case are very reasonable. >> > > Glad *someone* agrees... :) > > >> Regarding the larger (a) vs. (b) question, the primary contributors will >> have to make that call since they are the ones doing the work. >> >> Agreed. > > >> Of course that won't stop me from throwing in a few personal opinions :) >> >> The phrase "terse and a bit cryptic" describes a bug, not a feature. >> > Perhaps I should've said "succinct and extremely powerful"? :) > > >> Verbosity is a small price to pay for the ability to re-use the existing >> Java syntax knowledge that 100% of AspectJ users will already have. >> >> Regarding implementing option (a).... since we're talking about a piece >> of software that already knows how to compile Java expressions, >> implementing option (a) is "just" a matter of hooking up that existing >> stuff and applying it. I see no reason why the compiler couldn't compile >> these expressions into bytecode/transient classes as it compiles, then >> attempt to load and evaluate them during compilation. If this fails because >> the expression is not compile-time evaluable, then just weave the >> expression into the aspect and fall back to evaluation at runtime like it >> already does with the existing annotation support. >> >> I'm sure I'm making sound easier than it is though. >> >> I have the same feeling that it does sound easier than it actually is. > > Yes, I have that feeling too. Unfortunately with many feature requests for AspectJ right now you can have the cheap and easy solution now-ish or you can wait on the amazing solution which may-or-may-not ever get implemented... If it can be done in a few days I try to get to these things, if it takes weeks it is very hard to schedule. Archie is perfectly right, as a Java parser we can parse expressions like those you would encounter in (a) and attempt static evaluation possibly pushing any non statically resolvable information for evaluation at runtime. (We do that already for some constructs). If someone wants to work on this with me and give me a pull request, awesome, but I'm not sure when our schedules here will be clear enough for me to take a crack at it. cheers, Andy
_______________________________________________ aspectj-users mailing list aspectj-users@eclipse.org https://dev.eclipse.org/mailman/listinfo/aspectj-users
