[Cuis-dev] [Ann] Refinements to Exception handling

Andres Valloud ten at smallinteger.com
Mon Oct 14 21:02:39 PDT 2019


The below argument seems to focus on how easy it is to write new code. 
What I'm arguing is that the long term consequences of said code matter 
even more.  IME, non-local returns in exception handler blocks are not 
as beneficial as they initially seem to be.

I would object less if exceptions understood a message such as #return: 
that performed a non-local return.  At least then you can easily 
breakpoint that, because if instead of

	[:ex | ^5]

the code reads

	[:ex | ex nonLocalReturn: 5]

then you can reimplement nonLocalReturn: *in the exception of interest 
alone*, without breaking everything else.

This is different from ^5 in a regular method because there the context 
of ^5 is much more readily understood.  In an exception handler block, 
at first glance you have no idea of where the exception might be coming 
from, so what does ^5 mean in the context where the exception happened?

On 10/14/19 04:17, Hernan Wilkinson wrote:
> I checked and the exception handling mechanism is not affected by a 
> non-local return in an exception handler, it works fine.
> I think that having a non local return in an exc. handler is really 
> handy and I do not see the problem with that really. Of course that 
> people can make mistakes but that is related to not using non local 
> returns correctly, not to doing it in the exc. handler.
> The #return:, #pass, etc. do no have the same behavior as having a non 
> local return... with this change if I want to return from the method 
> when handling an exception a special object will have to be returned and 
> the a check with an if for it will be needed.
> 
> I think this change is a step backwards really, it makes things harder 
> for handler you want to return from the method and having a non local 
> return does not harm how exceptions work.
> 
> Hernan.
> 
> On Sun, Oct 13, 2019 at 5:12 PM Andres Valloud via Cuis-dev 
> <cuis-dev at lists.cuis.st <mailto:cuis-dev at lists.cuis.st>> wrote:
> 
>     The conversation that led to these changes went more or less like
>     this... first, I noted that exception handlers that look like this:
> 
>              [:ex | ^5]
> 
>     are asking for trouble, because:
> 
>     a) effectively, nobody is looking at how the exception handling
>     mechanism is implemented, so doing non-local return like that curtails
>     the private implementation of exception handling,
> 
>     b) that kind of behavior is what the messages #return, #resume, #pass,
>     #outer, etc are there for,
> 
>     c) when you need to debug why exceptions are behaving a certain way,
>     the
>     above exception handlers make debugging impossible because now there's
>     no place to put a breakpoint (e.g. in #return, #resume, #pass, #outer,
>     etc, which can be redefined for the purpose at hand in the single
>     exception of interest), and finally,
> 
>     d) it's possible to distort the exception handling mechanism behavior
>     with non-local returns.
> 
> 
>     The vast majority of the time, I see exception handlers with non-local
>     returns as very close to expressions such as these:
> 
>              #(1 2 3) asSortedCollection: [:x :y | ^x < y]
> 
>     where the behavior you get is not what you expected, even though the
>     code is doing exactly what was requested.  See attached for an actual
>     example of what I mean.  To run the code, do these and see what happens.
> 
>              ExceptionExample new breakingStuff
>              ExceptionExample new workingStuff
> 
>     To me, generally speaking, these considerations push the scales against
>     the apparent convenience of having exception handlers do a non-local
>     return.  I do not want to have to debug that kind of convenience under
>     time pressure.  I do not even want to have to change code that uses
>     such
>     convenience, because it's going to be more difficult than otherwise.
> 
> 
>     So Juan heard this and pointed out the reasoning is lovely but it's not
>     that helpful because the system is doing nothing to make users aware of
>     these facts.  He suggested the system should complain instead.  Thus we
>     (although he says mostly I) wrote code that detects non-local return
>     and
>     complains.  I believe that code was integrated recently (this bit, he
>     did on his own LOL).
> 
>     Andres.
> 
>     On 10/13/19 07:52, Hernan Wilkinson wrote:
>      > Could you provide the case were it generates problems?
>      >
>      > On Sun, Oct 13, 2019 at 11:40 AM Hernan Wilkinson
>      > <hernan.wilkinson at 10pines.com
>     <mailto:hernan.wilkinson at 10pines.com>
>     <mailto:hernan.wilkinson at 10pines.com
>     <mailto:hernan.wilkinson at 10pines.com>>> wrote:
>      >
>      >     Hi,
>      >       I think that forbidden a non local return in an exception
>     handler
>      >     is not really the solution... Although I understand the
>     motivation I
>      >     think that changing the exception handling mechanism would be
>     better.
>      >       I'm going to take a look at it to see if it is possible to
>     do it.
>      >
>      >     Hernan.
>      >
>      >     On Sun, Oct 13, 2019 at 10:48 AM Juan Vuletich via Cuis-dev
>      >     <cuis-dev at lists.cuis.st <mailto:cuis-dev at lists.cuis.st>
>     <mailto:cuis-dev at lists.cuis.st <mailto:cuis-dev at lists.cuis.st>>> wrote:
>      >
>      >         Hi Folks,
>      >
>      >         Some time ago, Andrés (with just a tad of help from me)
>     fixed a
>      >         problem
>      >         in Exception handling. If exception handler blocks do
>     non-local
>      >         return
>      >         (^stuff), they will skip execution of part of the Exception
>      >         handling
>      >         system code, breaking exception return values and
>     possibly other
>      >         'bad
>      >         things'.
>      >
>      >         I just pushed to GitHub a few updates with this work.
>     Now, if an
>      >         exception handler does a non-local return, an Error will be
>      >         raised. We
>      >         also fixed a couple of places in the image where this was
>     being
>      >         done.
>      >         BaseImageTests pass.
>      >
>      >         This updates have some risk of breaking your code. If you you
>      >         get this
>      >         error: 'Exception handler blocks must not do non local
>     returns',
>      >         then
>      >         you need to adjust your code. See updates #3917 to #3922
>     for the
>      >         changes
>      >         done to the base image.
>      >
>      >         Thanks,
>      >
>      >         --
>      >         Juan Vuletich
>      > www.cuis-smalltalk.org <http://www.cuis-smalltalk.org>
>     <http://www.cuis-smalltalk.org>
>      > https://github.com/Cuis-Smalltalk/Cuis-Smalltalk-Dev
>      > https://github.com/jvuletich
>      > https://www.linkedin.com/in/juan-vuletich-75611b3
>      >         @JuanVuletich
>      >
>      >         --
>      >         Cuis-dev mailing list
>      > Cuis-dev at lists.cuis.st <mailto:Cuis-dev at lists.cuis.st>
>     <mailto:Cuis-dev at lists.cuis.st <mailto:Cuis-dev at lists.cuis.st>>
>      > https://lists.cuis.st/mailman/listinfo/cuis-dev
>      >
>      >
>      >
>      >     --
>      >     *Hernán Wilkinson
>      >     Agile Software Development, Teaching & Coaching*
>      >     *Phone: +54-011*-4893-2057
>      >     *Twitter: @HernanWilkinson*
>      >     *site: http://www.10Pines.com <http://www.10pines.com/>*
>      >     Address: Alem 896, Floor 6, Buenos Aires, Argentina
>      >
>      >
>      >
>      > --
>      > *Hernán Wilkinson
>      > Agile Software Development, Teaching & Coaching*
>      > *Phone: +54-011*-4893-2057
>      > *Twitter: @HernanWilkinson*
>      > *site: http://www.10Pines.com <http://www.10pines.com/>*
>      > Address: Alem 896, Floor 6, Buenos Aires, Argentina
>     -- 
>     Cuis-dev mailing list
>     Cuis-dev at lists.cuis.st <mailto:Cuis-dev at lists.cuis.st>
>     https://lists.cuis.st/mailman/listinfo/cuis-dev
> 
> 
> 
> -- 
> *Hernán Wilkinson
> Agile Software Development, Teaching & Coaching*
> *Phone: +54-011*-4893-2057
> *Twitter: @HernanWilkinson*
> *site: http://www.10Pines.com <http://www.10pines.com/>*
> Address: Alem 896, Floor 6, Buenos Aires, Argentina


More information about the Cuis-dev mailing list