move jsonfg from plugin to default

[pvgenuchten]

Is your feature request related to a problem? Please describe.

I noticed a plugin jsonfg, by now is it an idea to move it to default status, since in may it was adopted by OGC as formal standard

Additional context

which is the formal way to request jsonfg from a server?

accept: draft: application/vnd.ogc.fg+json / final: application/fg+json (although not registered at IANA yet)

do we support the shortcut ?f=fg or f=jsonfg

[webb-ben]

The pygeoapi formatter implementation uses Formatter.f as the f= query arg match and Formatter.mimetype for Accept header.

As implemented in downstream plugin, https://reference.geoconnex.us/collections/hu02/items?f=jsonfg resolves to JSON-FG

[pvgenuchten]

which accept header is that? application/fg+json or application/vnd.ogc.fg+json

or is it this one? doesn't that conflict with default geojson?

https://github.com/cgs-earth/pygeoapi-plugins/blob/c8fd875b4f7b2775968cb22e3dc38ff5d2e7a666/pygeoapi_plugins/formatter/jsonfg.py#L70

[webb-ben]

Yep. Will fix in a sec

[pvgenuchten]

hi, sorry, just noticed this discussion opengeospatial/ogc-feat-geo-json#129

seems the group decided to not register a dedicated accept header, but use an additional profile parameter (wonder what the geojson community thinks about that)

• to request using accept header: Accept: application/geo+json; profile="http://www.opengis.net/def/profile/OGC/0/jsonfg"
• for the f=xx shortcut, they suggest ?f=json&profile=jsonfg

curl \
  -H 'Accept: application/geo+json; profile="http://www.opengis.net/def/profile/OGC/0/jsonfg"' \
  https://example.org/collections/buildings/items

wow, this stuff is complex

[webb-ben]

pygeoapi does not have that functionality at the moment. As valuable as it would be to have JSON-FG I do have questions about creating a new content negotiation profile

Will revert to application/geo+json seeing as pygeoapi looks for application/json

[pvgenuchten]

I will discuss with tom and those present in timisuara next week

[doublebyte1]

I guess that this would superceed this Draft PR? #2136

I think the PR was left in draft because of a dependency on GDAL, but this plugin seems to take a different approach in creating the JSON-FG structure

[doublebyte1]

hi, sorry, just noticed this discussion opengeospatial/ogc-feat-geo-json#129

seems the group decided to not register a dedicated accept header, but use an additional profile parameter (wonder what the geojson community thinks about that)

• to request using accept header: Accept: application/geo+json; profile="http://www.opengis.net/spec/json-fg-1/1.0/conf/core"
• for the f=xx shortcut, they suggest ?f=json&profile=fg

curl \
  -H 'Accept: application/geo+json; profile="http://www.opengis.net/spec/json-fg-1/1.0/conf/core"' \
  https://example.org/collections/buildings/items

wow, this stuff is complex

Please note that the name of the profile is jsonfg. We should refer to the published version of the Standard on this: https://docs.ogc.org/is/21-045r1/21-045r1.html#_using_a_json_fg_client

<http://www.opengis.net/def/profile/OGC/0/jsonfg>; rel="profile"

[tomkralidis]

Thanks @doublebyte1. For pygeoapi, this means a "format" and a "profile" would resonate to a formatter which we would resolve internally. I can put forth a PR to support the profile concept (noting this would be valuable in the sense of OGC API - Records as well), if that works? Perhaps a new issue noting support for "profile" is helpful, then.

[webb-ben]

I think the PR was left in draft because of a dependency on GDAL, but this plugin seems to take a different approach in creating the JSON-FG structure

Indeed. I found it rather inefficient to serialize the GeoJSON-dict to a string to then parse it with GDAL to produce JSONFG-dict when the two were already quite similar to one another. Will also note that using GDAL does not provide a straightforward path for adding the time or place attributes from simple features (i.e. not making any usable changes at the feature level). I use the crs aware formatter to convert geometry back to CRS84 and the time_field in provider def.

[pvgenuchten]

Please note that the name of the profile is jsonfg

@doublebyte1, Where did you notice incorrect usage of the standard name?

[webb-ben]

Please note that the name of the profile is jsonfg

@doublebyte1, Where did you notice incorrect usage of the standard name?

@pvgenuchten I believe the request pattern from the spec would be:

curl \
  -H 'Accept: application/geo+json; profile=jsonfg' \
  https://example.org/collections/buildings/items

With the URI for JSON-FG in the response headers:

Link: <http://www.opengis.net/def/profile/OGC/0/jsonfg>; rel="profile"
Content-Type: application/geo+json

Would this require formatters to have the ability to write headers?