From mboxrd@z Thu Jan 1 00:00:00 1970 X-Spam-Checker-Version: SpamAssassin 3.4.4 (2020-01-24) on polar.synack.me X-Spam-Level: X-Spam-Status: No, score=-1.3 required=5.0 tests=BAYES_00,INVALID_MSGID autolearn=no autolearn_force=no version=3.4.4 X-Google-Language: ENGLISH,ASCII-7-bit X-Google-Thread: fac41,9a0ff0bffdf63657 X-Google-Attributes: gidfac41,public X-Google-Thread: 103376,4b06f8f15f01a568 X-Google-Attributes: gid103376,public X-Google-Thread: 1108a1,9a0ff0bffdf63657 X-Google-Attributes: gid1108a1,public X-Google-Thread: f43e6,9a0ff0bffdf63657 X-Google-Attributes: gidf43e6,public From: Matthew Heaney Subject: Re: Software landmines (loops) Date: 1998/09/02 Message-ID: #1/1 X-Deja-AN: 387196123 Sender: matt@mheaney.ni.net References: <6renh8$ga7$1@nnrp1.dejanews.com> <6rf59b$2ud$1@nnrp1.dejanews.com> <6rfra4$rul$1@nnrp1.dejanews.com> <35DBDD24.D003404D@calfp.co.uk> <6sbuod$fra$1@hirame.wwa.com> <35f51e53.48044143@news.erols.com> <6sdiav$e0g$1@hirame.wwa.com> <6sfcft$70p$1@hirame.wwa.com> <6shp40$ec8$1@nnrp1.dejanews.com> <6sie46$eb7$1@hirame.wwa.com> <6siijm$h1m$1@hirame.wwa.com> <6sjl6v$5qh$1@hirame.wwa.com> NNTP-Posting-Date: Wed, 02 Sep 1998 10:48:26 PDT Newsgroups: comp.lang.eiffel,comp.object,comp.software-eng,comp.lang.ada Date: 1998-09-02T00:00:00+00:00 List-Id: "Robert Martin" writes: > >So no, you don't have to worry about bailing out early. There is no > >maintenance penalty for an early return from the loop. > > This is true if: > > 1. Your language has controlled types. Agreed. But C++ does have the equivalent of Ada95 controlled types. It may be that you have a philosophical problem with using deconstructors for dealing with resource management issues that can be handled via traditional techniques - like a single point of exit from a loop. Fair enough. But I argue that in a language with exception propagation, you're going to have to be thinking about resource management anyway, even if you're not writing a loop. So I think the resource management argument is a bit of a red herring. > 2. The cleanup that needs to be done can be done from the finalizer. For > example, in C++ the equivalent scheme is to use a destructor. However the > destructor is in a completely different scope. So extraordinary means may > need to be used to allow the destructor access to the variables that need > finalization. > > Consider the case of a complex data structure. You are in the midst of > changing a set of variables within it. Suddenly you realize that you must > exit early. Before you can exit, you have to put the data structure back > the way it was. In C++, getting a destructor to do this can be horrific. I don't see this as a hard problem. Let the object cache the new settings until a Commit operation is invoked. If you bail out early, then Commit doesn't get called, and the data structure retains its original value. > Btw, in essense a controlled type is simply a way to enforce that a scope > has a single exit. By using single-entry/single-exit style, *all* your > variables are controlled -- albeit manually. (excluding exceptions). But resource management in the presence of exceptions is precisely my point. Your parenthetical disclaimer is a little like saying, "All objects can levitate. (excluding gravitation)."