Re: Pico/Ersatz diffs: chain, link, yoke
Hi Alex, Since you have inserted checks for ‘Env.make’, I suggest you fix this one also: When I do this in ersatz … (setq X 99) (val X) … then I get java.lang.NullPointerException, but in real PicoLisp I get 99 -- Variable expected”. /Jon (with a new email address, but old one will stay valid) On 25. Jun, 2014, at 18:17, Alexander Burger a...@software-lab.de wrote: Hi all, thanks for the feedback! You would be mad to use Erzats for performance anyway, so I'd wish for correctness or similarity. It would also serve as a documentation of what it takes to implement a full PicoLisp. Sort of a portable spec. OK, but it is also a bit mad to check for something we know it will be checked immediately after that again. And we can't tell the JVM don't check that, we've done it already. Samuel's proposal for 'Control' is perhaps not useful here, as it is for catch/throw only. We could catch the NullPointerException and then throw an error. But that's also rather ugly, and setting up the handler for that is probably more expensive than just checking the value. So I inserted checks for 'Env.make' and the corresponding error messages into ersatz/fun.src. ♪♫ Alex PԔ � j)mX��� ��zV�u�.n7�
Re: Pico/Ersatz diffs: chain, link, yoke
Hi Jon, Since you have inserted checks for ‘Env.make’, I suggest you fix this one also: When I do this in ersatz … (setq X 99) (val X) … then I get java.lang.NullPointerException, but in real PicoLisp I get 99 -- Variable expected”. See? This is exactly the reason why I didn't want to fix it! In fact I did already regret that I did it. There are probably hundreds of such cases. You need to clobber the whole interpreter with such checks, without any gain in functionality but just another error message. And I don't want to do that, for reasons I tried to explain also in the previous mail. - Alex -- UNSUBSCRIBE: mailto:picolisp@software-lab.de?subject=Unsubscribe
Re: Pico/Ersatz diffs: chain, link, yoke
Hi Jon, If you in normal PicoLisp do any of the following, without any use of ‘make' … (chain (1 2 3)) (link 1 2 3) (yoke 1 2 3) … then you’ll get a Not making” error message. However, if you do the same with Ersatz, you’ll get a java.lang.NullPointerException for ‘chain’ and ‘link’. You are right, this is a disputable issue. In fact, I deliberately omitted the runtime checks for a 'make' environment in ErsatzLisp, because the JVM already does such a check (a NullPointerException is also an error message, albeit a rather unprecise one). If you take a closer look, you see that most explicit error checks which exist in PicoLisp are omitted in ErsatzLisp. For simplicity, brevity and performance. For (yoke 1 2 3) you’ll get 3 This is indeed unaesthetic, yes. At least here it is needed. We could easily add proper error messages to ErsatzLisp too, checking for non-NULL 'Env.Make' and 'Env.Yoke'. How is the general opinion on that? ♪♫ Alex -- UNSUBSCRIBE: mailto:picolisp@software-lab.de?subject=Unsubscribe
Re: Pico/Ersatz diffs: chain, link, yoke
Hi all, thanks for the feedback! You would be mad to use Erzats for performance anyway, so I'd wish for correctness or similarity. It would also serve as a documentation of what it takes to implement a full PicoLisp. Sort of a portable spec. OK, but it is also a bit mad to check for something we know it will be checked immediately after that again. And we can't tell the JVM don't check that, we've done it already. Samuel's proposal for 'Control' is perhaps not useful here, as it is for catch/throw only. We could catch the NullPointerException and then throw an error. But that's also rather ugly, and setting up the handler for that is probably more expensive than just checking the value. So I inserted checks for 'Env.make' and the corresponding error messages into ersatz/fun.src. ♪♫ Alex -- UNSUBSCRIBE: mailto:picolisp@software-lab.de?subject=Unsubscribe
