On Jan 28, 2011, at 9:21 AM, DenesL wrote:
>
>
>
> On Jan 28, 10:59 am, Jonathan Lundell <jlund...@pobox.com> wrote:
>>
>> It's not so obviously a bug.
>
> Agreed.
>
>> URL('f',args=['arg1','arg2']) is /app/ctl/f/arg1/arg2 or
>> /app/ctl/f/'arg1'/'arg2' (quotes added for clarity)
>>
>> so suppose arg2 is ''; the consistent transformation is:
>>
>> URL('f',args=['arg1','']) -> /app/ctl/f/'arg1'/'' or /app/ctl/f/arg1/
>> without the quotes
>>
>> I see three defensible choices. In all cases, .../arg1 -> ['arg1']
>>
>> 1. .../arg1/ -> ['arg1', ''] .../arg1// -> ['arg1', '', '']
>>
>> 2. .../arg1/ -> ['arg1'] .../arg1// -> ['arg1', '']
>>
>> 3. .../arg1/ -> ['arg1'] .../arg1// -> ['arg1']
>>
>> #1 follows Python split/join. Each trailing slash is another empty-string
>> arg.
>>
>> #2 adds a trailing slash on output iff the last arg is '', and always
>> deletes one trailing slash on input if it's present.
>>
>> #3 ignores all trailing slashes (that's the current code)
>>
>> #1 is the most simple, consistent and elegant, it seems to me. Its drawback
>> is that we sort of expect that .../arg1 and .../arg1/ are the same URL. But
>> that's not always the case, and I don't see a big problem with saying that a
>> visible difference has consequences. Also, we're mostly generating our own
>> URLs, so we control what they look like; they're not often being typed.
>
> #1 also seems to go along with the spirit of the RFC where the slash
> is a segment separator.
> The only worry is that I think some browsers add the trailing slash if
> it is not there (don't quote me on this though), but does it matter?.
If a browser did that, we'd see an extra '' arg, I guess, but I wonder if that
actually happens.
FWIW, we already see an extra '' arg when there's a trailing slash in the
current web2py. It's not intentional; the URL regex parser has a little bug,
and (again by accident) the logic that normally treats empty args as errors
doesn't catch this one.
So it must be at least tolerable. Mostly. But that's how this subject came up.
Kenneth Lundström wrote:
> I´m a little bit surprised that I havn´t noticed this before as I have used
> len(request.args) quite much to determine the state of an operation.
Though here's a question. Kenneth also wrote:
Has it allways been like this or?
> http://...../application/controller/function/args1 len(request.args) = 1
> http://...../application/controller/function/args1/args2 len(request.args)
> = 2
>
> but
> http://...../application/controller/function/args1/ len(request.args) = 2
>
> I would have guessed that the last case would have returned length as 1
Kenneth: how did you happen to notice this? How did the extra trailing slash
get there in the first place?
>
>> #2 follows the expectation that .../arg1 and .../arg1/ are the same, but it
>> does so at the expense of treating '' differently from all other arg values.
>>
>> #3 takes a different approach, flatly declaring that trailing slashes are
>> never significant. The downside is that trailing empty args ('') cannot be
>> made explicit in the URL (though in fairness, in current web2py they're
>> illegal, so...).
>>
>> We should get this resolved before Massimo releases the next stable version.
>> Massimo, do you have a preference? I'm leaning toward #1 at the moment, but
>> not very strongly.
>>
