geosolutions-it/nexus-geonode

Issue with dateline

Closed this issue · 8 comments

simboss commented
Freshdesk Ticket ID: 49348
Freshdesk Ticket Agent: Support
Freshdesk Ticket Agent Email: support@geo-solutions.it
Ticket Priority: Medium
Freshdesk Ticket Description:
Below is the link to the datasets that I loaded to the platform from the ShareDrive that does not fully show up as we viewed it because of the dateline of Fiji.


The corrections demo that you show during our meeting is really helpful and I request that those corrections of test issues that we highlight during our meeting last night could be included during the training sessions so we can perform those corrections to issues as we progress further towards the final upload.
simboss commented
Agent Support linked Freshdesk ticket 49348 for this issue.

afabiani commented

PR available here: GeoNode/geonode#9071

simboss commented
Agent Simone Giannecchini created comment.

Dear Tupou,
ticket 49348  has been assigned today in our support experts backlog.
You will hear from us shortly.

Best Regards,
GeoSolutions Support.
==
Enterprise Support Services from the experts for GeoNode, GeoServer, MapStore and GeoNetwork.
Visit http://bit.ly/gs-services for more information.
==

GeoSolutions Group

http://www.geosolutionsgroup.com
http://twitter.com/geosolutions_it

-------------------------------------------------------
giohappy commented

PR available here: GeoNode/geonode#9071

@afabiani

  1. beyond submitting a PR, could you explain what it solves, how it solves it, and how it is related to the issue? Did you check the specific case? I know you're proposing an improvement to handle bboxes spanning the whole world, but in this specific case there's something with the data itself and/or their representation
  2. a quick view to your PR reveals that you're proposing to calculate the split of the bbox when a filtering is requested. This means that the split is performed on each (spatially filtered) search. It doesn't look very efficient, and it can be a significant work in case of complex polygons with many vertices. Did you consider splitting the bbox when it's calculated?
afabiani commented

@giohappy only the search bbox is splitted, the query are going to be merged in one queryset. They are lazy query therefore in the end it will be an or condition with the intersection.

The solution is to split the search polygon crossing the dateline and search for datasets intersecting the two parts of the polygon. Tipically the search polygon is a rectangle, therefore I don't think it will be expensive to split it.

giohappy commented

ok. @afabiani I see now that we're splitting the search bbox.
In any case I would open an issue for this and connect the PR to that one, since it doesn't seem to solve the thing raised by SPC about the specific dataset.

giohappy commented

@afabiani I guess you meant to connect #584

simboss commented
Agent Support created comment.

Hi Tupou,

as expected the issue with the "Exposed Reeds" dataset (beyond the not well defined CRS as the other cases) are the geometries crossing the dateline.

As you can see in the following two images, QGIS also fails at rendering it correctly, because lines crossing the dateline are interpreted as connected the other way.



As you can see on http://epsg.io/3460:

  • the WGS84 longitude boundaries go from -176.81 to 176.81. We know that it means across the datelines, but a computer cannot assume it
  • as a result the visualization of the box on the map in that page also has the issue. It spans the whole world.

The only solution here is to clip the polygons / lines crossing the dateline. in QGIS this can be accomplished with the "Split with lines" tool.
I have attached a "split line" shapefile, containing a single line along the dateline that can be used to execute the split.
After running the split tool you can see that the original polylines have been cut (see image below)
​

​
​
 
 
==
Our support, Your Success! Visit https://geosolutionsgroup.com/enterprise-support-services for more information.
==

Support @ GeoSolutions 


Ticket attachments :