Jernej Tuljak <jern...@mg-soft.si> wrote:
> Martin Bjorklund je 25.9.2015 ob 15:22 napisal:
> > Jernej Tuljak <jern...@mg-soft.si> wrote:
> >> Ladislav Lhotka je 25.9.2015 ob 13:44 napisal:
> >>> Jernej Tuljak <jern...@mg-soft.si> writes:
> >>>
> >>>> Ladislav Lhotka je 25.9.2015 ob 9:58 napisal:
> >>>>> Jernej Tuljak <jern...@mg-soft.si> writes:
> >>>>>
> >>>>>> Martin Bjorklund je 24.9.2015 ob 13:54 napisal:
> >>>>>>> Jernej Tuljak <jern...@mg-soft.si> wrote:
> >>>>>>>> Ladislav Lhotka je 23.9.2015 ob 14:29 napisal:
> >>>>>>>>> Jernej Tuljak <jern...@mg-soft.si> writes:
> >>>>>>>>>
> >>>>>>>>>> Ladislav Lhotka je 23.9.2015 ob 12:52 napisal:
> >>>>>>>>>>>> On 23 Sep 2015, at 12:37, Jernej Tuljak <jern...@mg-soft.si>
> >>>>>>>>>>>> wrote:
> >>>>>>>>>>>>
> >>>>>>>>>>>> Ladislav Lhotka je 23.9.2015 ob 11:58 napisal:
> >>>>>>>>>>>>>> On 23 Sep 2015, at 11:45, Jernej Tuljak <jern...@mg-soft.si>
> >>>>>>>>>>>>>> wrote:
> >>>>>>>>>>>>>>
> >>>>>>>>>>>>>> Section 7.21.5., the first bullet describing the augmentation
> >>>>>>>>>>>>>> case
> >>>>>>>>>>>>>> does not consider that an "augment" may specify a top-level
> >>>>>>>>>>>>>> "choice"
> >>>>>>>>>>>>>> or a "choice"/"case" nested within a top-level "choice" as the
> >>>>>>>>>>>>>> target
> >>>>>>>>>>>>>> node. There will be no initial context node for the expression
> >>>>>>>>>>>>>> in
> >>>>>>>>>>>>>> such
> >>>>>>>>>>>>>> a case.
> >>>>>>>>>>>>>>
> >>>>>>>>>>>>>> leaf enable-foos {
> >>>>>>>>>>>>>> type boolean;
> >>>>>>>>>>>>>> }
> >>>>>>>>>>>>>> choice ch1 {
> >>>>>>>>>>>>>> case foo {
> >>>>>>>>>>>>>> choice foos {
> >>>>>>>>>>>>>> when "enable-foos = 'true'";
> >>>>>>>>>>>>>> leaf foo1 {
> >>>>>>>>>>>>>> type string;
> >>>>>>>>>>>>>> }
> >>>>>>>>>>>>>> leaf foo2 {
> >>>>>>>>>>>>>> type string;
> >>>>>>>>>>>>>> }
> >>>>>>>>>>>>>> }
> >>>>>>>>>>>>>> }
> >>>>>>>>>>>>>> container meh;
> >>>>>>>>>>>>>> }
> >>>>>>>>>>>>>>
> >>>>>>>>>>>>>> augment "/ch1/foo/foos" {
> >>>>>>>>>>>>>> when "enable-foos = 'true'";
> >>>>>>>>>>>>>> leaf foo3 {
> >>>>>>>>>>>>>> type string;
> >>>>>>>>>>>>>> }
> >>>>>>>>>>>>>> }
> >>>>>>>>>>>>>>
> >>>>>>>>>>>>>> OLD:
> >>>>>>>>>>>>>>
> >>>>>>>>>>>>>> o If the "when" statement is a child of an "augment"
> >>>>>>>>>>>>>> statement,
> >>>>>>>>>>>>>> then
> >>>>>>>>>>>>>> the context node is the augment's target node in the
> >>>>>>>>>>>>>> data tree,
> >>>>>>>>>>>>>> if
> >>>>>>>>>>>>>> the target node is a data node. Otherwise, the
> >>>>>>>>>>>>>> context
> >>>>>>>>>>>>>> node is
> >>>>>>>>>>>>>> the closest ancestor node to the target node that is
> >>>>>>>>>>>>>> also a data
> >>>>>>>>>>>>>> node.
> >>>>>>>>>>>>>>
> >>>>>>>>>>>>>> NEW:
> >>>>>>>>>>>>>>
> >>>>>>>>>>>>>> o If the "when" statement is a child of an "augment"
> >>>>>>>>>>>>>> statement,
> >>>>>>>>>>>>>> then
> >>>>>>>>>>>>>> the context node is the augment's target node in the
> >>>>>>>>>>>>>> data tree,
> >>>>>>>>>>>>>> if
> >>>>>>>>>>>>>> the target node is a data node. Otherwise, the
> >>>>>>>>>>>>>> context
> >>>>>>>>>>>>>> node is
> >>>>>>>>>>>>>> the closest ancestor node to the target node that is
> >>>>>>>>>>>>>> also a data
> >>>>>>>>>>>>>> node. If no such node exists, the context node is
> >>>>>>>>>>>>>> the
> >>>>>>>>>>>>>> root node.
> >>>>>>>>>>>>>>
> >>>>>>>>>>>>> Good catch, I support this change.
> >>>>>>>>>>>>>
> >>>>>>>>>>>>> Also regarding sec. 7.21.5, I think we agreed that a when
> >>>>>>>>>>>>> expression
> >>>>>>>>>>>>> must not reference nodes that are made conditional through that
> >>>>>>>>>>>>> when
> >>>>>>>>>>>>> statement. I can’t find this rule in the text.
> >>>>>>>>>>>> The "no referencing the initial context node and its descendants"
> >>>>>>>>>>>> text
> >>>>>>>>>>>> is currently in the guidelines draft only (6087bis).
> >>>>>>>>>>> IMO this should be a hard rule in 6020bis. Such expressions are
> >>>>>>>>>>> seriously broken and could lead to deadlocks.
> >>>>>>>>>> I think that was the purpose of the change in 3rd bullet. A "dummy"
> >>>>>>>>>> node
> >>>>>>>>>> with no value and child nodes has been introduced.
> >>>>>>>>> Not really, the dummy node solves the problem of a non-existent
> >>>>>>>>> context
> >>>>>>>>> node.
> >>>>>>>> Are you referring to "existence" in the instance document or the
> >>>>>>>> schema tree? Why is there no need for "no children, no value dummy"
> >>>>>>>> in
> >>>>>>>> 1st and 2nd bullet?
> >>>>>>> In these cases, the context node is well-defined anyway. But I think
> >>>>>>> that also these bullets need to have text that make sure that the
> >>>>>>> outcome is deterministic in all cases, such as:
> >>>>>>>
> >>>>>>> augement /foo {
> >>>>>>> when "bar > 1";
> >>>>>>> leaf bar { type int32; }
> >>>>>>> }
> >>>>>>>
> >>>>>>> or
> >>>>>>>
> >>>>>>> grouping foo {
> >>>>>>> leaf bar { type int32; }
> >>>>>>> }
> >>>>>>> container x {
> >>>>>>> uses foo {
> >>>>>>> when "bar > 1";
> >>>>>>> }
> >>>>>>> }
> >>>>>>>
> >>>>>>>
> >>>>>>> etc.
> >>>>>>>
> >>>>>>> So I suggest this NEW text:
> >>>>>>>
> >>>>>>>
> >>>>>>> o If the "when" statement is a child of an "augment" statement,
> >>>>>>> then
> >>>>>>> the context node is the augment's target node in the data
> >>>>>>> tree,
> >>>>>>> if
> >>>>>>> the target node is a data node. Otherwise, the context node
> >>>>>>> is
> >>>>>>> the closest ancestor node to the target node that is also a
> >>>>>>> data
> >>>>>>> node. If no such node exists, the context node is the root
> >>>>>>> node.
> >>>>>>> The accessible tree is tentatively altered during the
> >>>>>>> processing
> >>>>>>> of the XPath expression by removing all instances (if any)
> >>>>>>> of
> >>>>>>> the
> >>>>>>> nodes added by the "augment" statement.
> >>>>>>>
> >>>>>>> o If the "when" statement is a child of a "uses", "choice", or
> >>>>>>> "case" statement, then the context node is the closest
> >>>>>>> ancestor
> >>>>>>> node to the "uses", "choice", or "case" node that is also a
> >>>>>>> data
> >>>>>>> node. If no such node exists, the context node is the root
> >>>>>>> node.
> >>>>>>> The accessible tree is tentatively altered during the
> >>>>>>> processing
> >>>>>>> of the XPath expression by removing all instances (if any)
> >>>>>>> of
> >>>>>>> the
> >>>>>>> nodes added by the "uses", "choice", or "case" statement.
> >>>>>>>
> >>>>>>> o If the "when" statement is a child of any other data
> >>>>>>> definition
> >>>>>>> statement, the accessible tree is tentatively altered during
> >>>>>>> the
> >>>>>>> processing of the XPath expression by replacing all
> >>>>>>> instances
> >>>>>>> (if
> >>>>>>> any) of the data node for which the "when" statement is
> >>>>>>> defined
> >>>>>>> with a single dummy node with the same name, but with no
> >>>>>>> value
> >>>>>>> and
> >>>>>>> no children. The context node is this dummy node.
> >>>>>>>
> >>>>>> +1
> >>>>>>
> >>>>>> container foo {
> >>>>>> when "child::* or child::text() or not(self::*)"; // always
> >>>>>> false
> >>>>>> leaf bar {
> >>>>>> type string;
> >>>>>> }
> >>>>>> }
> >>>>> I think it would be better to clearly state that a "when" expression
> >>>>> MUST NOT refer to nodes that are made conditional by the same "when"
> >>>>> statement - it's similar to the rule that forbids mutual references.
> >>>>> This would also cover all uses of "when".
> >>>>>
> >>>>> The above example would be really confusing if the "bar" instance is
> >>>>> present.
> >>>> It is true that this introduces a caveat that is bound to confuse end
> >>>> YANG users.
> >>>>
> >>>>> I also don't like the fact that "when" and "must" would give
> >>>>> different results in this case.
> >>>> Just to be clear, you would like this, right?
> >>>>
> >>>> NEW:
> >>>>
> >>>> The XPath expression is conceptually evaluated in the following
> >>>> context, in addition to the definition inSection 6.4.1
> >>>> <https://tools.ietf.org/html/draft-ietf-netmod-rfc6020bis-07#section-6.4.1>:
> >>>>
> >>>> o If the "when" statement is a child of an "augment" statement,
> >>>> then
> >>>> the context node is the augment's target node in the data tree,
> >>>> if
> >>>> the target node is a data node. Otherwise, the context node is
> >>>> the closest ancestor node to the target node that is also a data
> >>>> node. If no such node exists, the context node is the root node.
> >>>>
> >>>> o If the "when" statement is a child of a "uses", "choice", or
> >>>> "case" statement, then the context node is the closest ancestor
> >>>> node to the "uses", "choice", or "case" node that is also a data
> >>>> node. If no such node exists, the context node is the root
> >>>> node.
> >>>>
> >>>> o If the "when" statement is a child of any other data definition
> >>>> statement, the context node is the node in the accessible tree
> >>>> for
> >>>> which the "when" statement is defined.
> >>>> An XPath expression MUST NOT reference nodes that are made
> >>>> conditional
> >>>> by the "when" statement that defines it.
> >>> Yes, plus the other two rules that are already in 6020bis:
> >>>
> >>> If the XPath expression references any node that also has associated
> >>> "when" statements, these "when" expressions MUST be evaluated first.
> >>>
> >>> There MUST NOT be any circular dependencies in these "when"
> >>> expressions.
> >>>> I don't see a case in which the context node is non-existent (in the
> >>>> schema tree) in any of the three bullets, even without "no value, no
> >>>> descendants" dummy node.
> >>> The third bullet is aimed at cases like this:
> >>>
> >>> leaf foo {
> >>> when "../bar > 0";
> >>> mandatory true;
> >>> ...
> >>> }
> >>>
> >>> If there is no instance of "foo", a validator has to find out whether
> >>> the "mandatory" statement applies (and the absence of "foo" is thus a
> >>> schema error), or whether it is overruled by the "when" expression
> >>> being
> >>> false. But in order to be able to evaluate that expression, XPath
> >>> requires a context node in the instance document, so we need to
> >>> tentatively add one.
> >> Then the dummy node text is broken all together. It instructs to
> >> modify the accessible tree (schema)
> > The accessible tree is not the schema, it is the instance data.
> >
> >> instead of instructing to
> >> instantiate a dummy if an instance of the context node does not
> >> exist.
> > No, the text says *replace* any instances with the dummy instance;
> > this ensures that the result is the same regardless of if the node
> > exists or not.
>
> Okay, the accessible tree is instantiated data nodes.
>
> container top {
> leaf foo {
> when "../bar > 0";
> mandatory true;
> ...
> }
> }
>
> "top": {}
>
> "top": { "foo": "" }
>
> How is a context node ensured for the two instances, if replacing
> something that may not exist? You would need to create a dummy if
> there is no "foo" instance, right? The text only says "replace".
Ok; the intention was "replace if it exists, and create otherwise"
(i.e., "replace" as in NETCONF edit operation). We can make this
explicit:
NEW:
- If the "when" statement is a child of any other data definition
statement, the accessible tree is tentatively altered during the
processing of the XPath expression by replacing all instances of the
data node for which the "when" statement is defined with a single
dummy node with the same name, but with no value and no children.
If no instances of the data node exists, the dummy node is created.
The context node is this dummy node.
> >> A context node always exists in the schema tree.
> > The context node is in the data tree.
> >
> >> IMO, this
> >> needs to be fixed in order to prevent the interpretation I came up
> >> with (and also Balazs, independently). This interpretation was clearly
> >> not the intention.
> >>
> >> NEW:
> >>
> >> The XPath expression is conceptually evaluated in the following
> >> context, in addition to the definition inSection 6.4.1
> >>
> >> <https://tools.ietf.org/html/draft-ietf-netmod-rfc6020bis-07#section-6.4.1>:
> >>
> >> o If the "when" statement is a child of an "augment" statement, then
> >> the context node is the augment's target node in the data tree, if
> >> the target node is a data node. Otherwise, the context node is
> >> the closest ancestor node to the target node that is also a data
> >> node. If no such node exists, the context node is the root node.
> >>
> >> o If the "when" statement is a child of a "uses", "choice", or
> >> "case" statement, then the context node is the closest ancestor
> >> node to the "uses", "choice", or "case" node that is also a data
> >> node. If no such node exists, the context node is the root node.
> >>
> >> o If the "when" statement is a child of any other data definition
> >> statement, the context node is the node in the accessible tree for
> >> which the "when" statement is defined.
> >> If an instance of the context node does not exist during
> >> processing
> >> of the XPath expression, the context node is tentatively instantiated
> >> as a single dummy node with the same name, but with no value and no
> >> children, and merged into the instance of the accessible tree.
> >>
> >> The result of the XPath expression is converted to a boolean value
> >> using the standard XPath rules.
> >>
> >> An XPath expression MUST NOT reference nodes that are made conditional
> >> by the "when" statement that defines it.
> >>
> >> If the XPath expression references any node that also has associated
> >> "when" statements, these "when" expressions MUST be evaluated first.
> >>
> >> There MUST NOT be any circular dependencies in these "when"
> >> expressions.
> >>
> >> Something like this also takes care of Martin's example ("* = 42").
> > How? It is not clear to me if the '*' is legal or not.
>
> It just occurred to me that it is very hard to not reference the
> context node in a "when" XPath expression, since the context node is
> always your initial XPath context (you always start with a node-set of
> size 1, containing the context node).
>
> "../bar" says, select the parent of the context node, then select
> "bar" child
> "*" says, select the children of the context node that are also
> element nodes
>
> So, yeah, "MUST NOT reference conditional nodes" doesn't work.
So do you think that my proposed text about "tentatively remove all
conditional nodes" works?
/martin
_______________________________________________
netmod mailing list
netmod@ietf.org
https://www.ietf.org/mailman/listinfo/netmod
