Re: [GD-Windows] Array foreach in scripting languages
Brought to you by:
vexxed72
Menu
▾
▴
Re: [GD-Windows] Array foreach in scripting languages
|
From: Mat N. \(BUNGIE\) <Mat...@mi...> - 2006-08-13 19:00:29
|
If it's a scripting language, don't allow direct manipulation of a = container while iterating through it. =20 If you want the side effects of such an operation, defer them until all = iterators have been destroyed. =20 MSN ________________________________ From: gam...@li... on behalf of = Carsten Orthbandt Sent: Sun 8/13/2006 4:21 AM To: Game Development for MS Windows Subject: Re: [GD-Windows] Array foreach in scripting languages I'm afraid I wasn't clear. The actual data structure of the list or array may be anything. In the case at hand, it is not a linked list. The linked list I was referring to would be a linked list of all iterators attached to a specific array instance specifically for adjusting the iterators on list changes. So yes, that's what I was about to do. Modifying arrays poses some serious challenges for multi-threaded operation anyway, I don't think it's getting much harder with this iterator scheme in place. I was planning to do heavy multi-threading through thread pools and task lists that operate in higher level chunks as to not spend to much time on the threading itself. Best regards, Carsten Orthbandt Founder + Development Director SEK SpieleEntwicklungsKombinat GmbH http://www.sek-ost.de <http://www.sek-ost.de/>=20 Wenn ich Visionen habe, gehe ich zum Arzt. - Helmut Schmidt =20 ________________________________ Von: gam...@li... im Auftrag von = Richard Fabian Gesendet: So 13.08.2006 12:13 An: Game Development for MS Windows Betreff: Re: [GD-Windows] Array foreach in scripting languages Dunno if it suits how you work, but you could always attach the = iterators to the linked list on iterator construction. this way you can tell all the currently attached iterators that element = X is gone, handle it. with thiis in place, we've never had any problems with our iterators, = but i suspect that we may have problems in the future with PS3 and such = because this technique doesn't lend itself to threaded access. On 8/13/06, Carsten Orthbandt <car...@se...> wrote: The reason I'd like to have this work is that it is a recurring pattern in lots of code I see (iterating a list and deleting/inserting entries). =20 If I go the route with tracked iterators, any insert or delete operation will work as expected since then I'm dealing with these cases per iterator: =20 - delete item before or at current position -> decrement position and size - delete item after current position -> decrement size - insert item before or at current position -> increment position and size - insert item after current position -> increment size =20 The case "delete item at current position" potentially has to deal with the actual element being deleted. In C++ this would be a problem but in my scripting language the list entry would go away but the actual value referenced will stay around due the iterator still holding onto it. =20 I'm quite surprised that the C# docs don't even mention this situation. I'm even more surprised to find that MS added a "for each" to C++. It's supposed to be only for managed code but it worked fine in my standard STL unmanaged test. =20 Simply banning this scenario would mean that users have to write such code: =20 int i,iC=3DList.GetSize(); for(i=3D0;i<iC;i++) { if(Condition(List[i])) { List.Delete(i); i--;iC--; } }; =20 manually. Which is goes against the whole idea of pro- viding a foreach and IMHO even more error-prone. I'm not sure if I more dislike the C#/STL behaviour of throwing an exception or the Python behaviour of simply ignoring the problem. =20 <rant> In my case, an exception is not an option since I'm not going to clutter my cute little scripting language with a concept (exceptions) that I consider totally broken in the first place. There's no place in a game for such stuff. It has to run, period. So you have to handle all error conditions anyway. </rant> =20 Best regards, =20 Carsten Orthbandt Founder + Development Director SEK SpieleEntwicklungsKombinat GmbH http://www.sek-ost.de <http://www.sek-ost.de/>=20 =20 Wenn ich Visionen habe, gehe ich zum Arzt. - Helmut Schmidt =20 =20 =20 ________________________________ =20 Von: gam...@li... im = Auftrag von Kent Quirk Gesendet: So 13.08.2006 04:39 An: Game Development for MS Windows Betreff: Re: [GD-Windows] Array foreach in scripting languages =20 =20 =20 In Python, I was shocked to find that this succeeds: =20 L =3D range(20) for i in L: print i if i=3D=3D4: del L[6] =20 Result: ------- 0 1 2 3 4 5 7 8 9 =20 I fully expected it to fail. I consider messing with the list = you're traversing with an iterator to be on a par with sawing off the = branch you're standing on, unless, as you imply, the list is known to = be implemented as a linked list. In python, you can surprise = yourself. Consider this: =20 for i in L: print i if i=3D=3D4: del L[3] # delete one of the items BEFORE the one = you're on =20 =20 0 1 2 3 4 6 7 8 9 =20 So what happened is that we ended up skipping element #5, even = though we deleted element 3. =20 This is really dependent on the implementation details of the = particular interpreter, and probably isn't guaranteed by the average = language spec. =20 If you feel strongly about being able to do this, you have to = deal with the case of deleting the item you're on right now, and also = about deleting the last element while you're at the last element. = Check this out: =20 >>> L ['h', 'e', 'l', 'l', 'o'] >>> for c in L: if c=3D=3D'l': L.remove(c) =20 >>> L ['h', 'e', 'l', 'o'] =20 In other words, even though I intended to remove BOTH 'l' = characters, only one of them was removed because deleting one caused me to = skip the next element. =20 I think this stuff is so fraught with potential errors by the = user, even if YOU get it right, that I'd prefer to just ban the practice = and make people use a "filter" concept like in python, or a remove_if as = in STL. =20 Kent =20 =20 Carsten Orthbandt wrote: > I'm currently building a small scripting language for = interactive stuff. Now I > know there's plenty out there, I'm just doing this for fun. > > A tricky question came up and I'd like to collect opinions and = experiences on > this. It's about foreach for dynamic arrays. Consider this = C++-like pseudo code: > > DynamicArray<int> list; > int i; > for(i=3D0;i<20;i++){list.Add(i);}; > foreach(i in list){ > print(i); > if(i=3D=3D8){list.Delete(12);}; > }; > > The obvious question is how should an active iterator react if = the traversed list > changes? To answer this, I simply tried it out in two = languages I know that do > provide both dynamic arrays and foreach: STL, Tcl and C#. > > In STL C++ code, I get an exception because of iterator = incompatibility. > In Tcl this problem isn't one because the iterator takes a = snapshot of the list > object on start and there's no language construct to actually = change this instance. > In C# with the List<int> generic, I get an exception because = the iterated object > changes (interestingly, this happens on ANY change to the = list, even appending). > > Now what I'd like to know: > - how do other languages cope with this? (Tcl doesn't need to, = C++/C# simply fails) > - what do YOU think would be sensible behaviour? > > Currently I'm considering having all iterators be elements of = a double-linked list > and let the array object be the dl-list head. So if the number = of array elements > changes, it can run through all attached iterators and fix = their indices or stop them. > > =20 =20 = -------------------------------------------------------------------------= Using Tomcat but need to do more? Need to support web services, = security? Get stuff done quickly with pre-integrated technology to make = your job easier Download IBM WebSphere Application Server v.1.0.1 based on = Apache Geronimo = http://sel.as-us.falkag.net/sel?cmd=3Dlnk&kid=3D120709&bid=3D263057&dat=3D= 121642 _______________________________________________ Gamedevlists-windows mailing list Gam...@li... = https://lists.sourceforge.net/lists/listinfo/gamedevlists-windows Archives: http://sourceforge.net/mailarchive/forum.php?forum_id=3D555 =20 =20 =20 = -------------------------------------------------------------------------= Using Tomcat but need to do more? Need to support web services, = security? Get stuff done quickly with pre-integrated technology to make = your job easier Download IBM WebSphere Application Server v.1.0.1 based on = Apache Geronimo = http://sel.as-us.falkag.net/sel?cmd=3Dlnk&kid=3D120709&bid=3D263057&dat=3D= 121642 _______________________________________________ Gamedevlists-windows mailing list Gam...@li... = https://lists.sourceforge.net/lists/listinfo/gamedevlists-windows Archives: http://sourceforge.net/mailarchive/forum.php?forum_id=3D555 =20 -- fabs(); -------------------------------------------------------------------------= Using Tomcat but need to do more? Need to support web services, = security? Get stuff done quickly with pre-integrated technology to make your job = easier Download IBM WebSphere Application Server v.1.0.1 based on Apache = Geronimo http://sel.as-us.falkag.net/sel?cmd=3Dlnk&kid=3D120709&bid=3D263057&dat=3D= 121642 _______________________________________________ Gamedevlists-windows mailing list Gam...@li... https://lists.sourceforge.net/lists/listinfo/gamedevlists-windows Archives: http://sourceforge.net/mailarchive/forum.php?forum_id=3D555 |
