why is the term constructor optional in production `reifier`?

[pchampin]

The production rule for reifier is:

    [28] reifier ::=  '~' (iri | BlankNode)?

I'm curious about the rationale for the trailing question mark, i.e. why the term (IRI or bnode) after the ~ symbol is optional...

I find it more complicated to handle that in my parser, and it is not clearly stated in section 7.3.1 what to do when that term is omitted -- I would assume that a fresh bnode should be minted, but that is not explicit.

Also, I don't see much benefit for the user, except for allowing confusing notations such as << :s :p :o ~ >> which would be equivalent to << :s :p :o >>, or :s :p :o ~ {| ... |} which would be equivalent to :s :p :o {| ... |}.

I strongly suggest to change production 28 to:

    [28] reifier ::=  '~' (iri | BlankNode)

[afs]

For context:

:s :p :o ~ .

which is

:s :p :o .
_:new rdf:reifies <<( :s :p :o )>> .

It is a convenient way to allocated an id for later use (in time, not this file) so that downstream uses the same identifier (e.g. loading a database which made makes id choice permanent).

It could be

:s :p :o ~ [] .

though that is more verbose.

it is not clearly stated in section 7.3.1

curReifier is set by handle-reifier in 7.2 RDF Term Constructors.

suggest to change production 28

which would affect

:s :p :o ~ {| :q :z | }.

which can be chained for independent annotations:

:s :p :o ~ {| :q1 :z1 | } ~ {| :q2 :z2 | }

[pchampin]

Ok, I stand corrected about this being underspecified, but I still find this syntax odd and not really required.

As you point out, :s :p :o ~ []. does the same job. Slightly more verbose, but also more regular.

For annotations, this can be achieved by omitting the reifier production entirely, which is even less verbose:

:s :p :o {| :q1 :z1 | } {| :q2 :z2 | }.`

[TallTed]

As these documents are meant to educate, I strongly advise including all optional components in the initial presentation. It is fine to include secondary, tertiary, and more presentations which omit the optional components, preferably one-by-one, so that readers understand the "maximum verbosity" implied by each of the "less verbose" versions.

[afs]

A change is to Turtle, Trig and SPARQL and the respective tests.

:s :p :o ~ . has been used in discussions.

My larger concern is that reopening syntax for cases which value-judgements, rather than bugs, is that it signals any part of the grammar is up for discussion.

[TallTed]

Note that I'm not suggesting "reopening syntax". I'm suggesting using that syntax.

We have sugar for various things, and optional syntactical elements which are supposed to have specific meanings in the syntax as currently specified.

If we cannot write the syntax in full expansion, i.e., showing what is meant by all sugars and including all optional elements, as well as in full contraction, i.e., employing all sugars and omitting all optional elements, then we ought not be certain that it will work as we think, nor that readers will be able to make use of any of that sugar and optionality.

[pchampin]

:s :p :o ~ . has been used in discussions.

I missed that completely, and so I half-assumed that this was an unintentional "bug" in the grammar. But if it was not...

My larger concern is that reopening syntax for cases which are value-judgements, rather than bugs, is that it signals any part of the grammar is up for discussion.

... then I sympathize entirely with this concern.

[ktk]

Discussed today, closing.