[Laszlo-dev] org.openlaszlo.sc.CompilerError: class names only differ by upper/lower case: "Top" versus "top"

[P T Withington]

Got it.  Sounds good to me.

On 2008-08-06, at 16:35EDT, Donald Anderson wrote:

> The scheme is simple.
> The first 'bar' seen goes to the top level.
> Actually the first of any non-conflicting name goes to the top level.
> For any conflict, we put it into a subdirectory, the lowest numbered
> N such that N/name.as does not already exist as a conflict.
>
> So if we see these names:
>
>   bar, foo, BAr, FOO,  boink, BaR, Boink
>
> they are in:
>   ./bar.as
>   ./foo.as
>   ./0/BAr.as
>   ./0/FOO.as
>    ./boink.as
>   ./1/BaR.as
>    ./0/Boink.as
>
> I was originally thinking of putting each conflict into a unique  
> subdir, but each conflict adds
> a -compiler.source-path argument to the command line, and I don't  
> want that to get too long
> when compiling programs with conventions that push this.
>
> On Aug 6, 2008, at 4:22 PM, P T Withington wrote:
>
>> You'll need some more encoding, as Chuq points out -- where will  
>> `class BAr` and `class BaR` and &c. go?  Or do you just plan to  
>> create subdirectories as needed each time you get a conflict?
>>
>> On 2008-08-06, at 16:04EDT, Donald Anderson wrote:
>>
>>> Yeah, probably the same bonehead that mandated that backslash be  
>>> the DOS file separator.
>>>
>>> But remember - any subdirectory names we choose are only known to  
>>> us, and put in the $TEMPDIR/lzswf9/lzgenXXX dir.
>>> And if we wanted to be a little cute, we could allocate file names  
>>> on a first-seen first-serve in the top directory with 'overflow'  
>>> into numbered directories.
>>> class foo -  put into ./foo.as
>>> class bar -  put into ./bar.as
>>> var Foo - put into ./0/Foo.as
>>> class Bar - put into ./1/Bar.as
>>>
>>> That would normally have the effect of having everything in the  
>>> top directory and occasional entries in unique subdirectories.
>>>
>>> I just tried an experiment building an lfc by hand with variable  
>>> 'lime' and 'Lime' using this technique, and it appears to work.
>>>
>>>
>>> On Aug 6, 2008, at 3:36 PM, P T Withington wrote:
>>>
>>>> <flame>
>>>> Wow.
>>>>
>>>> Is the bonehead who decided your file name must match your class  
>>>> any relation to the bonehead who decided that line breaks had to  
>>>> be represented by CR _and_ LF?
>>>> </flame>
>>>>
>>>> <sigh />
>>>>
>>>> On 2008-08-06, at 15:10EDT, Donald Anderson wrote:
>>>>
>>>>> I don't think AS3 is the problem, the HFS+ file system is.
>>>>> And even if you take global vars out of the picture,
>>>>> you have the issue of someone that wants
>>>>> both class foo and class Foo in the same program.
>>>>> If you want to allow that (and I think AS3 probably does),
>>>>> we need a way to do that.  Putting the global vars into their
>>>>> own subdirectory just splits the namespace into two.  Admittedly
>>>>> a bandaid, but maybe good enough.
>>>>>
>>>>> I tried creating a java program creating names files with names  
>>>>> "\u0041" and "\u0061".
>>>>> The second failed, since this is exactly equivalent to creating  
>>>>> "A" and "a".
>>>>>
>>>>> http://forums.macosxhints.com/showthread.php?t=14607
>>>>>
>>>>> is one of many places that confirms that HFS+
>>>>> is case preserving, but it does not distinguish between upper  
>>>>> and lower case.
>>>>>
>>>>> My recollection is that Windows FAT32 has some similar  
>>>>> properties.  So I think we are stuck
>>>>> with subdirectories as a solution, and it's our choice how  
>>>>> sophisticated we want to be with that.
>>>>>
>>>>> - Don
>>>>>
>>>>> On Aug 6, 2008, at 1:39 PM, P T Withington wrote:
>>>>>
>>>>>> Surely in AS3 you can have a global and a class that only  
>>>>>> differ in case.  In which case, we need a different strategy  
>>>>>> for how we implement global vars.
>>>>>>
>>>>>> On 2008-08-06, at 13:15EDT, Henry Minsky wrote:
>>>>>>
>>>>>>> I think the flex compiler, like Java, requires that the file  
>>>>>>> name be
>>>>>>> the same as the top level definition which
>>>>>>> it publishes, so there's not any choice of file name for a given
>>>>>>> global declaration.
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> On Wed, Aug 6, 2008 at 12:13 PM, P T Withington  
>>>>>>> <ptw at pobox.com> wrote:
>>>>>>>> Maybe we just need to come up with an encoding for the names  
>>>>>>>> to indicate
>>>>>>>> case.  Is the filesystem case-insensitive if you use a utf8  
>>>>>>>> filename?
>>>>>>>>
>>>>>>>> On 2008-08-06, at 09:36EDT, Donald Anderson wrote:
>>>>>>>>
>>>>>>>>> Well, I think both 'var top' and 'class Top' are competing for
>>>>>>>>> file name top.as vs. Top.as.
>>>>>>>>>
>>>>>>>>> We can always use subdirectories to distinguish names (I  
>>>>>>>>> think!).
>>>>>>>>> Your idiom seems natural to me, and makes it seem like we
>>>>>>>>> should try it.  For example, all globals are pushed to their
>>>>>>>>> own directory:
>>>>>>>>> global/top.as
>>>>>>>>>
>>>>>>>>> You still wouldn't be able to have classes or globals that
>>>>>>>>> aren't distinguished within their own set.  If we wanted full
>>>>>>>>> distinguishing, we'd have to do something like:
>>>>>>>>>
>>>>>>>>> 0/ClassName.as
>>>>>>>>> 1/AnotherClassName.as
>>>>>>>>>
>>>>>>>>> yuck.
>>>>>>>>>
>>>>>>>>> - Don
>>>>>>>>>
>>>>>>>>> On Aug 6, 2008, at 8:51 AM, P T Withington wrote:
>>>>>>>>>
>>>>>>>>>> I should have included my example code:
>>>>>>>>>>
>>>>>>>>>>> <canvas debug="true" >
>>>>>>>>>>> <script when="immediate">
>>>>>>>>>>> var free = 'outer';
>>>>>>>>>>>
>>>>>>>>>>> class Top {
>>>>>>>>>>> function test () { return free; }
>>>>>>>>>>> }
>>>>>>>>>>>
>>>>>>>>>>> class Sub extends Top {
>>>>>>>>>>> var free = 'inner';
>>>>>>>>>>>
>>>>>>>>>>> function test () { return free; }
>>>>>>>>>>> }
>>>>>>>>>>>
>>>>>>>>>>> var top = new Top;
>>>>>>>>>>> var sub = new Sub;
>>>>>>>>>>> Debug.info("top.test() => %s", top.test());
>>>>>>>>>>> Debug.info("sub.test() => %s", sub.test());
>>>>>>>>>>> </script>
>>>>>>>>>>> </canvas>
>>>>>>>>>>
>>>>>>>>>> There is only one class Top.  Is the check overly  
>>>>>>>>>> conservative?
>>>>>>>>>>
>>>>>>>>>> On 2008-08-06, at 08:39EDT, Donald Anderson wrote:
>>>>>>>>>>
>>>>>>>>>>> That check is in because classes must be named to files.
>>>>>>>>>>> Class Glorp must be in file Glorp.as.  But some file systems
>>>>>>>>>>> including MacOS, do not have complete distinguishing  
>>>>>>>>>>> between case:
>>>>>>>>>>>
>>>>>>>>>>> $ echo foo >> glorp.as
>>>>>>>>>>> $ echo bar >> Glorp.as
>>>>>>>>>>> $ cat glorp.as
>>>>>>>>>>> foo
>>>>>>>>>>> bar
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> On Aug 6, 2008, at 8:06 AM, P T Withington wrote:
>>>>>>>>>>>
>>>>>>>>>>>> Why am I getting this error when compiling to swf9?
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> --
>>>>>>>>>>>
>>>>>>>>>>> Don Anderson
>>>>>>>>>>> Java/C/C++, Berkeley DB, systems consultant
>>>>>>>>>>>
>>>>>>>>>>> voice: 617-547-7881
>>>>>>>>>>> email: dda at ddanderson.com
>>>>>>>>>>> www: http://www.ddanderson.com
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> --
>>>>>>>>>
>>>>>>>>> Don Anderson
>>>>>>>>> Java/C/C++, Berkeley DB, systems consultant
>>>>>>>>>
>>>>>>>>> voice: 617-547-7881
>>>>>>>>> email: dda at ddanderson.com
>>>>>>>>> www: http://www.ddanderson.com
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> -- 
>>>>>>> Henry Minsky
>>>>>>> Software Architect
>>>>>>> hminsky at laszlosystems.com
>>>>>>
>>>>>
>>>>>
>>>>> --
>>>>>
>>>>> Don Anderson
>>>>> Java/C/C++, Berkeley DB, systems consultant
>>>>>
>>>>> voice: 617-547-7881
>>>>> email: dda at ddanderson.com
>>>>> www: http://www.ddanderson.com
>>>>>
>>>>>
>>>>>
>>>>
>>>
>>>
>>> --
>>>
>>> Don Anderson
>>> Java/C/C++, Berkeley DB, systems consultant
>>>
>>> voice: 617-547-7881
>>> email: dda at ddanderson.com
>>> www: http://www.ddanderson.com
>>>
>>>
>>>
>>
>
>
> --
>
> Don Anderson
> Java/C/C++, Berkeley DB, systems consultant
>
> voice: 617-547-7881
> email: dda at ddanderson.com
> www: http://www.ddanderson.com
>
>
>