RDF to Object conversion (closing #555) #619

anatoly-scherbakov · 2024-11-08T15:34:30Z

Why

Trying to solve #555.

What

Clarify what we do when can't serialize a number from RDF into JSON.

Preview | Diff

anatoly-scherbakov · 2024-11-08T16:03:34Z

I don't think I can invite reviewers to this PR 🤔 @pchampin @gkellogg

gkellogg · 2024-11-08T16:08:10Z

I don't think I can invite reviewers to this PR 🤔 @pchampin @gkellogg

I'll look at your permissions, you should have full permissions on this and other repos.

gkellogg

The changes are to the informative summary, but the issue actually calls out 2.4.3 (was 2.3.4) which is normative; that's the important area to address.

index.html

gkellogg · 2024-11-08T21:00:29Z

@anatoly-scherbakov, if you can create branches at json-ld-api rather than on your own fork, it is easier for others to collaborate.

gkellogg · 2024-11-08T21:03:40Z

As an alternative, we may treat these as error conditions, which is what JCS does.

3.2.2.3. Serialization of Numbers
...
Note: Since Not a Number (NaN) and Infinity are not permitted in JSON, occurrences of NaN or Infinity MUST cause a compliant JCS implementation to terminate with an appropriate error.

davidlehn · 2024-11-08T22:34:14Z

Can you please replace the emoji in the commit titles with words.

index.html

gkellogg

Minor change to table head: th elements need to be wrapped inside a tr.

index.html

Co-authored-by: Gregg Kellogg <[email protected]>

index.html

tests/fromRdf-manifest.jsonld

gkellogg

Looks good, subject to @TallTed's suggestions. We may want to revisit what happens if Infinity or NaN is attempted to be serialized.

Co-authored-by: Ted Thibodeau Jr <[email protected]>

anatoly-scherbakov · 2024-11-24T08:29:55Z

@TallTed created a separate issue about the version of the IEEE754 standard we refer to: #620

anatoly-scherbakov · 2024-11-24T08:30:27Z

@gkellogg do you mean we could convert them to JSON Strings instead of raising a JSON-LD processing error?

anatoly-scherbakov · 2024-11-24T08:31:57Z

GitHub tells me that

You're not authorized to push to this branch. Visit https://docs.github.com/repositories/configuring-branches-and-merges-in-your-repository/managing-protected-branches/about-protected-branches for more information.

Not sure why though, it might be related to the fact that I opened it in my fork.

pchampin

See my comments in the "Files changed" view.

index.html

pchampin · 2024-11-29T12:38:52Z

index.html

+          <tr>
+            <td>Any other value</td>
+            <td class="error">Throw <a data-link-for="JsonLdErrorCode">invalid JSON literal</a></td>
+          </tr>


I don't think that is a good idea, but I'll comment on that in the algorithm itself, below.

pchampin · 2024-11-29T12:39:23Z

index.html

+            <td class="error" rowspan="2">Throw <a data-link-for="JsonLdErrorCode">invalid JSON literal</a></td>
+          </tr>


Same as above; I don't think that is a good idea, see my comments in the algorithm.

pchampin · 2024-11-29T12:40:08Z

index.html

+          </tr>
+          <tr>
+            <td>Any other value</td>
+            <td class="error">Throw <a data-link-for="JsonLdErrorCode">invalid JSON literal</a></td>


This is not what the algorithm is saying!

index.html

pchampin · 2024-11-29T12:58:24Z

index.html

+                  according to [[XMLSCHEMA11-2]], and if it is a numeric value
+                  conformant to [[IEEE-754-2008]], then set <var>converted value</var>
                  to the result of converting the
                  <a>lexical form</a>
-                  to a JSON <a>number</a>.</li>
+                  to a JSON <a>number</a>.
+                </li>
+                <li>
+                  Otherwise, if <a>lexical form</a> expresses a numeric value
+                  outside of the ranges supported by [[IEEE-754-2008]], then set
+                  <var>converted value</var> to a JSON string literal containing
+                  said lexical form.
+                </li>
+                <li>
+                  Otherwise, if <a>lexical form</a> expresses a non-numeric value,
+                  such as <code>"+INF"^^xsd:double</code>, an
+                  <a data-link-for="JsonLdErrorCode">invalid JSON literal</a>
+                  error has been detected and processing is aborted.
+                </li>


I don't think this is the right way to change the algorithm, sorry @anatoly-scherbakov .

this is repurposing the invalid JSON literal error, which was initially about "literals of type @json", not about "JSON scalar values"...

we should definitely not raise error on well-formed literals (such as big integers or non-finite doubles).

I would not raise errors on ill-formed numbers either, but encode them as value objects -- this is what the spec currently says, and I don't think we should change that.

to improve round-tripping, I suggest that we use the same rules to differentiate integers and doubles as in the Object to RDF algorithm (currently, an very big xsd:integer would round trip to an xsd:double, which is not great).

So I suggest that step 2.4.3 of the original algorithm should be replaced with:

Otherwise, if the datatype IRI of value equals xsd:integer, its lexical form is a valid xsd:integer according to [XMKSCHEMA11-2], and its absolute value is strictly lower than 10^21, the set converted value to the result of converting its lexical form to a JSON number.

Otherwise, if the datatype IRI of value equals xsd:double, its lexical form is a valid xsd:double according to [XMLSCHEMA11-2] and is neither "NaN", "INF" not "-INF", then set converted value to the result of converting the lexical form to a JSON number.

Of course, the table and the tests should then be updated accordingly.

Any specific range examples should be careful to be informative, to not get ahead of XSD. Also, it will need to be consistent with JCS.

It might be that they way to do this is to just use JCS for serializing strings, and require that the result of serializing to JCS must result that is in the value space of xsd:integer or xsd:double, respectively.

Co-authored-by: Pierre-Antoine Champin <[email protected]>

gkellogg

I question if this is a class-2 editorial change; I think it's class-3 change, which will require Candidate Correction marking. We can discuss.

pchampin · 2024-12-03T09:57:24Z

I question if this is a class-2 editorial change; I think it's class-3 change, which will require Candidate Correction marking. We can discuss.

+1, this falls under "clears up an ambiguity or under-specified part of the specification" which is part of "class 3".
+1 also to discuss

gkellogg

I agree with @pchampin, but we need to be clear that it needs to be valid both as xsd:double and JCS number, and fallback to the expanded representation.

gkellogg · 2024-12-31T22:34:07Z

index.html

+
+          <tr>
+            <td>Whole number out of that range</td>
+            <td>JSON string literal</td>


Actually, this needs to be serialized as a value object with @type. The integer 1e54 (1000000000000000078291540404596243842305360299886116864) is outside of that range, but the value object needs to include `"@type": "xsd:integer".

gkellogg · 2024-12-31T22:46:34Z

index.html

+
+          <tr>
+            <th rowspan="3"><code>xsd:integer</code></th>
+            <td>Whole number in the <code>[-2<sup>53</sup> + 1, 2<sup>53</sup> + 1]</code> range, representable as per [[IEEE-754-2008]]</td>


There are more constraints in the lexical form of an xsd:integer than this.

Also, while it's likely consistent with JCS/ECMAScript, such numbers also need to be representable as integers in JCS/ECMAScript. See Appendix B.
We should be careful to not mis-describe this, and just refer to XSD Datatypes 1.1 and JCS.

gkellogg · 2024-12-31T22:48:26Z

index.html

+
+          <tr>
+            <td>Real number out of that range</td>
+            <td>JSON string literal</td>


See above comment; the ranges should be consistent with XSD and JCS and probably shouldn't re-define the criteria.

gkellogg · 2024-12-31T22:52:23Z

index.html

+                  according to [[XMLSCHEMA11-2]], and if it is a numeric value
+                  conformant to [[IEEE-754-2008]], then set <var>converted value</var>
                  to the result of converting the
                  <a>lexical form</a>
-                  to a JSON <a>number</a>.</li>
+                  to a JSON <a>number</a>.
+                </li>
+                <li>
+                  Otherwise, if <a>lexical form</a> expresses a numeric value
+                  outside of the ranges supported by [[IEEE-754-2008]], then set
+                  <var>converted value</var> to a JSON string literal containing
+                  said lexical form.
+                </li>
+                <li>
+                  Otherwise, if <a>lexical form</a> expresses a non-numeric value,
+                  such as <code>"+INF"^^xsd:double</code>, an
+                  <a data-link-for="JsonLdErrorCode">invalid JSON literal</a>
+                  error has been detected and processing is aborted.
+                </li>


Any specific range examples should be careful to be informative, to not get ahead of XSD. Also, it will need to be consistent with JCS.

It might be that they way to do this is to just use JCS for serializing strings, and require that the result of serializing to JCS must result that is in the value space of xsd:integer or xsd:double, respectively.

gkellogg · 2024-12-31T22:54:41Z

tests/fromRdf-manifest.jsonld

+        "useNativeTypes": true
+      },
+      "input": "fromRdf/0028-in.nq",
+      "expectErrorCode": "invalid JSON literal"


This is not invalid, it just can't be serialized as a native value.

anatoly-scherbakov · 2025-01-12T18:59:31Z

I believe I was misunderstanding a few things when writing this PR, sorry guys. Instead of reverting much of these changes, I just created another PR: #625

And I am closing this one; it won't disappear so as to maintain the history, just will be closed.

w3cbot · 2025-01-15T17:54:36Z

This was discussed during the #json-ld meeting on 15 January 2025.

View the transcript

w3c/json-ld-api#619

<gb> CLOSED Pull Request 619 RDF to Object conversion (closing #555) (by anatoly-scherbakov) [spec:editorial] [class-2]

anatoly-scherbakov: this is an earlier version that was closed after discussion and started over.

bigbluehat: Ok, I will take it out of the project

<anatoly-scherbakov> w3c/json-ld-api#625

<gb> Pull Request 625 (closes #555) Fall back to default logic in `useNativeTypes` mode for RDF numbers which are not JSON numbers (by anatoly-scherbakov)

anatoly-scherbakov force-pushed the 555-rdf-to-object-conversion branch from a8a9101 to 5c95152 Compare November 8, 2024 16:00

anatoly-scherbakov changed the title ~~555 rdf to object conversion~~ RDF to Object conversion (closing #555) Nov 8, 2024

anatoly-scherbakov marked this pull request as ready for review November 8, 2024 16:00

anatoly-scherbakov mentioned this pull request Nov 8, 2024

blind spot in RDF to object conversion #555

Open

gkellogg requested review from gkellogg and pchampin November 8, 2024 16:08

gkellogg requested changes Nov 8, 2024

View reviewed changes

index.html Outdated Show resolved Hide resolved

index.html Outdated Show resolved Hide resolved

gkellogg linked an issue Nov 8, 2024 that may be closed by this pull request

blind spot in RDF to object conversion #555

Open

davidlehn reviewed Nov 8, 2024

View reviewed changes

index.html Outdated Show resolved Hide resolved

gkellogg added spec:editorial class-2 Class-2 change labels Nov 11, 2024

Add .gitignore with JetBrains IDEs folder ignore

3d4391c

anatoly-scherbakov force-pushed the 555-rdf-to-object-conversion branch from 8f49ccd to 93bd915 Compare November 12, 2024 09:25

anatoly-scherbakov added 3 commits November 12, 2024 13:27

Add fromRdf-manifest.jsonld#t0027 to test "+INF"^^xsd:double

5fb21fa

Reformat Serialize RDF as JSON-LD Algorithm section a bit

52a47a4

Specify exceptions to useNativeTypes flag

17edb1b

anatoly-scherbakov force-pushed the 555-rdf-to-object-conversion branch from 93bd915 to 17edb1b Compare November 12, 2024 09:27

Format table

0b32f75

gkellogg requested changes Nov 12, 2024

View reviewed changes

index.html Outdated Show resolved Hide resolved

anatoly-scherbakov added 3 commits November 13, 2024 18:44

Add some CSS formatting for table.simple

2f34954

Add 2.4.4 & 2.4.5 sections in Algorithm chapter

fd224e1

Add links to errors in the informative table

fd89aed

anatoly-scherbakov force-pushed the 555-rdf-to-object-conversion branch from dc53fb3 to fd89aed Compare November 13, 2024 15:31

Split into 0027 and 0028 tests

2c51796

anatoly-scherbakov requested a review from davidlehn November 13, 2024 17:25

anatoly-scherbakov requested a review from gkellogg November 13, 2024 17:25

gkellogg requested changes Nov 18, 2024

View reviewed changes

index.html Outdated Show resolved Hide resolved

index.html Outdated Show resolved Hide resolved

Update index.html

05139b0

Co-authored-by: Gregg Kellogg <[email protected]>

TallTed reviewed Nov 22, 2024

View reviewed changes

index.html Outdated Show resolved Hide resolved

TallTed reviewed Nov 22, 2024

View reviewed changes

index.html Outdated Show resolved Hide resolved

TallTed reviewed Nov 22, 2024

View reviewed changes

tests/fromRdf-manifest.jsonld Outdated Show resolved Hide resolved

gkellogg approved these changes Nov 22, 2024

View reviewed changes

anatoly-scherbakov and others added 3 commits November 24, 2024 12:03

Update index.html

7e5167e

Co-authored-by: Ted Thibodeau Jr <[email protected]>

Replace range with ranges

35f105c

Co-authored-by: Ted Thibodeau Jr <[email protected]>

Insert a missing comma at tests/fromRdf-manifest.jsonld

d02a82b

Co-authored-by: Ted Thibodeau Jr <[email protected]>

anatoly-scherbakov mentioned this pull request Nov 24, 2024

We use an older 2008 edition of IEEE754 standard #620

Open

pchampin requested changes Nov 29, 2024

View reviewed changes

anatoly-scherbakov and others added 3 commits November 29, 2024 23:37

Update index.html

d4cd1f6

Co-authored-by: Pierre-Antoine Champin <[email protected]>

Update index.html

5161fc4

Co-authored-by: Pierre-Antoine Champin <[email protected]>

Update index.html

a6132a4

Co-authored-by: Pierre-Antoine Champin <[email protected]>

gkellogg requested changes Dec 3, 2024

View reviewed changes

gkellogg requested changes Dec 31, 2024

View reviewed changes

anatoly-scherbakov mentioned this pull request Jan 12, 2025

(closes #555) Fall back to default logic in useNativeTypes mode for RDF numbers which are not JSON numbers #625

Open

anatoly-scherbakov closed this Jan 12, 2025

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

RDF to Object conversion (closing #555) #619

RDF to Object conversion (closing #555) #619

anatoly-scherbakov commented Nov 8, 2024 •

edited by pr-preview bot

Loading

anatoly-scherbakov commented Nov 8, 2024

gkellogg commented Nov 8, 2024

gkellogg left a comment

gkellogg commented Nov 8, 2024

gkellogg commented Nov 8, 2024

davidlehn commented Nov 8, 2024

gkellogg left a comment

gkellogg left a comment

anatoly-scherbakov commented Nov 24, 2024

anatoly-scherbakov commented Nov 24, 2024

anatoly-scherbakov commented Nov 24, 2024

pchampin left a comment

pchampin Nov 29, 2024

pchampin Nov 29, 2024

pchampin Nov 29, 2024

pchampin Nov 29, 2024

gkellogg Dec 31, 2024

gkellogg left a comment

pchampin commented Dec 3, 2024

gkellogg left a comment

gkellogg Dec 31, 2024

gkellogg Dec 31, 2024

gkellogg Dec 31, 2024

gkellogg Dec 31, 2024

gkellogg Dec 31, 2024

anatoly-scherbakov commented Jan 12, 2025

w3cbot commented Jan 15, 2025

w3c/json-ld-api#619

		<td class="error" rowspan="2">Throw <a data-link-for="JsonLdErrorCode">invalid JSON literal</a></td>
		</tr>

RDF to Object conversion (closing #555) #619

RDF to Object conversion (closing #555) #619

Conversation

anatoly-scherbakov commented Nov 8, 2024 • edited by pr-preview bot Loading

Why

What

anatoly-scherbakov commented Nov 8, 2024

gkellogg commented Nov 8, 2024

gkellogg left a comment

Choose a reason for hiding this comment

gkellogg commented Nov 8, 2024

gkellogg commented Nov 8, 2024

davidlehn commented Nov 8, 2024

gkellogg left a comment

Choose a reason for hiding this comment

gkellogg left a comment

Choose a reason for hiding this comment

anatoly-scherbakov commented Nov 24, 2024

anatoly-scherbakov commented Nov 24, 2024

anatoly-scherbakov commented Nov 24, 2024

pchampin left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

gkellogg left a comment

Choose a reason for hiding this comment

pchampin commented Dec 3, 2024

gkellogg left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

anatoly-scherbakov commented Jan 12, 2025

w3cbot commented Jan 15, 2025

anatoly-scherbakov commented Nov 8, 2024 •

edited by pr-preview bot

Loading