Re: [netmod] 'predicate' production

[Ladislav Lhotka <[e-mail]>]

> On 25 Feb 2016, at 14:51, Martin Bjorklund <[e-mail]> wrote:
> 
> Ladislav Lhotka <[e-mail]> wrote:
>> 
>>> On 25 Feb 2016, at 14:16, Martin Bjorklund <[e-mail]> wrote:
>>> 
>>> Ladislav Lhotka <[e-mail]> wrote:
>>>> Martin Bjorklund <[e-mail]> writes:
>>>> 
>>>>> Ladislav Lhotka <[e-mail]> wrote:
>>>>>> Hi,
>>>>>> 
>>>>>> another ABNF issue:
>>>>>> 
>>>>>>  predicate           = "[" *WSP (predicate-expr / pos) *WSP "]"
>>>>>> 
>>>>>>  predicate-expr      = (node-identifier / ".") *WSP "=" *WSP
>>>>>>                        ((DQUOTE string DQUOTE) /
>>>>>>                         (SQUOTE string SQUOTE))
>>>>>> 
>>>>>>  pos                 = non-negative-integer-value
>>>>>> 
>>>>>>  non-negative-integer-value = "0" / positive-integer-value
>>>>>> 
>>>>>> The value of 0 shouldn't be allowed for 'pos' because context position
>>>>>> in XPath is always positive, i.e. the first list entry is selected
>>>>>> with "[1]".
>>>>> 
>>>>> You are right - I have changed this to:
>>>>> 
>>>>>  pos = positive-integer-value
>>>> 
>>>> Actually, there is another problem: the ABNF allows, for example
>>>> 
>>>> /if:interfaces/if:interface[1][2]
>>>> 
>>>> which doesn't make sense.
>>> 
>>> Yes, you're right.
>>> 
>>>> So a correct version could be
>>>> 
>>>>  instance-identifier = 1*("/" (node-identifier (*predicate / pos))
>>> 
>>> this would make pos mandatory after every node-identifier.  It should
>>> be: 
>>> 
>>> instance-identifier = 1*("/" (node-identifier (*predicate / *1pos)))
>>> 
>>> or, in order to match the style in the rest of the grammar:
>>> 
>>> instance-identifier = 1*("/" (node-identifier [1*predicate / pos]))
>> 
>> If we want to be really precise, then the predicate for a leaf-list [
>> . = 'foo'] can also appear no more than once.
> 
> Correct.  But we can only do so much in the grammar... For example, we
> cannot verify that the name and number of keys are correct.

Yes, but these are already semantic constraints. I think the new productions below are in fact easier to comprehend. Perhaps a production like

quoted-string = (DQUOTE string DQUOTE) / (SQUOTE string SQUOTE)

would also be an improvement.

Lada

> 
> It would be:
> 
> instance-identifier = 1*("/" (node-identifier [1*key-predicate /
>                                               leaf-list-predicate /
>                                               pos]))
> 
> key-predicate       = "[" *WSP key-predicate-expr *WSP "]"
> 
> key-predicate-expr  = node-identifier *WSP "=" *WSP
>                      ((DQUOTE string DQUOTE) /
>                       (SQUOTE string SQUOTE))
> 
> leaf-list-predicate = "[" *WSP leaf-list-predicate-expr *WSP "]"
> 
> leaf-list-predicate-expr = "." *WSP "=" *WSP
>                      ((DQUOTE string DQUOTE) /
>                       (SQUOTE string SQUOTE))
> 
> pos                 = "[" *WSP positive-integer-value *WSP "]"
> 
> 
> 
> /martin

--
Ladislav Lhotka, CZ.NIC Labs
PGP Key ID: E74E8C0C