[XML-DEV Mailing List Archive Home] [By Thread] [By Date] [Recent Entries] [Reply To This Message]

Re: Which is more declarative? More XMLish?

• From: Dimitre Novatchev <dnovatchev@gmail.com>
• To: Michael Kay <mike@saxonica.com>, "xml-dev@l..." <xml-dev@l...>
• Date: Sun, 3 Dec 2017 16:23:37 -0800

>> Perhaps more interestingly, XPath 3.1 introduces maps, and there is no rule that says a map must be finite. It is possible to take a finite graph structure (such as schema component model) and represent each node in the graph as a map; the map will then be infinite (you can navigate around it for ever, going round and round in circles).
>
> Or perhaps more relevant to the topic, the result of serializing the map as JSON and then converting the JSON to XML will be an infinite XML document: one can envisage performing successful operations on such a document, for example asking how many level-2 elements it contains (the answer will be a finite number).
>

I know we are deviating from the initial topic, but it is interesting
to speak about infinite XML documents. An infinite document in general
cannot be stored, and one important case of infinite data is one that
is generated continuously and dynamically in time. What matters about
such data is the stream of its values in a specific finite time
interval, containing the current moment. Context about preceding
siblings generated before this interval is unavailable/lost, so is
context about following siblings in the far future.

Therefore, it is possible to design/imagine an XML *fragment* with
such properties. The meaning of such data is not the data as a whole,
but its context in any specific time interval. We still can perform a
number of meaningful operations on such infinite streams -- splitting,
joining, maintaining accumulators,
filtering/pattern-matching-of-the-event-stream, generating events and
notifications. Clients can subscribe to events and to
filtered/re-formatted streams.

The issue is that one cannot "enter an XML document in the middle",
any such data must have a set of "entry points" from which it is
meaningful to start processing the (current part of the) document.
Similarly, it would be meaningful to have a set of "exit points"
(after reaching which unavailable context is required) -- and it seems
obvious that the exit point corresponding to a entry point would be
its (of the entry point) closing tag.

Yes, there is work to be done, and this would be fascinating and
useful to work upon...


Cheers,
Dimitre




On Sat, Dec 2, 2017 at 1:13 PM, Michael Kay <mike@saxonica.com> wrote:
>
>> On 2 Dec 2017, at 18:15, Michael Kay <mike@saxonica.com> wrote:
>>
>>
>> I don't think there is any rule that says the number of elements in a document must be finite.
>>
>> ...
>
>> Perhaps more interestingly, XPath 3.1 introduces maps, and there is no rule that says a map must be finite. It is possible to take a finite graph structure (such as schema component model) and represent each node in the graph as a map; the map will then be infinite (you can navigate around it for ever, going round and round in circles).
>
> Or perhaps more relevant to the topic, the result of serializing the map as JSON and then converting the JSON to XML will be an infinite XML document: one can envisage performing successful operations on such a document, for example asking how many level-2 elements it contains (the answer will be a finite number).
>
> Michael Kay
> Saxonica

• References:
  • Which is more declarative? More XMLish?
    • From: "Costello, Roger L." <costello@mitre.org>
  • Re: Which is more declarative? More XMLish?
    • From: Dimitre Novatchev <dnovatchev@gmail.com>
  • Re: Which is more declarative? More XMLish?
    • From: Michael Kay <mike@saxonica.com>
  • Re: Which is more declarative? More XMLish?
    • From: Michael Kay <mike@saxonica.com>