Add a minimal action I/O implementation to improve SearchAction support #237
Conversation
| @@ -235,6 +235,17 @@ def _extract_property_value(self, node, items_seen, base_url, itemids, force=Fal | |||
| elif node.get("content"): | |||
| return node.get("content") | |||
|
|
|||
| # https://schema.org/docs/actions.html#part-4 | |||
There was a problem hiding this comment.
It seems most example in this link use an alternative representation, with a string ("url-output": "required").
- How is it handled in this PR, and how was it handled before? Could it be that we start to lose some data which was easily available before? I.e. previously we just returned
"url-output": "required", but now it's"url-output": {}, something like this? - Do you know which approach is more common? Is it common on real websites to use the "Textual representations of Input and Output", or is it a rare edge case?
There was a problem hiding this comment.
How is it handled in this PR, and how was it handled before?
As far as I know, before it was not handled at all, at least for the required and name fields of query-input. Now it is handled but only those 2 fields.
I could extend the scope to all fields from those docs, but I am not sure how to map defaultValue given its documented data type (Thing, DataType), which was the main reason I stuck to the fields that were actually of interest to me.
Do you know which approach is more common?
The string representation you mention is the only one I have actually seen, in JSON-LD structures of websites (e.g. required name=search_terms). But it was my impression that the output of extruct’s microdata extractor tends to be more structured (e.g. it has property fields that are not used in JSON-LD), so I assumed this output was preferred here. I am OK with going the string approach, though.
No description provided.