[OGC] Polygon: fix rings order

[lbartoletti]

Description

I'm reaching out regarding an important initiative aimed at rectifying an anomaly observed in the digitization process within QGIS. Specifically, the adherence to OGC standards, upon which QGIS relies, dictates a distinct orientation requirement for polygon rings.

As per the OGC norm, the exterior boundary LinearRing must define the "top" of the surface, following a counterclockwise traversal along the boundary. Conversely, interior LinearRings should exhibit an opposite orientation, appearing clockwise when viewed from the "top."

Presently, there's a lack of verification within QGIS to ensure adherence to this prescribed behavior. To address this issue, I'll be initiating a series of PR; focusing only on the digitization process here. The primary goal is to align the behavior more closely with the established standards.

Nevertheless, I intend to maintain flexibility in the codebase, allowing for the preservation of the previous behavior programmatically. This approach aims to avoid directly enforcing orientation within the *Polygon classes. So, classes still accept "unordered" rings.

Before:

before-converted.mp4

After:

after-converted.mp4

PS: The motivation behind this is that certain tools perform stricter checks than the ones we currently conduct or those performed with GEOS.
PS2: I'm aware that ESRI Shapefile is CW unlike OGC, but should be a provider role to check this part?

TODO:

• fix tests
• add test

[qgis-bot]

@lbartoletti
This pull request has been tagged as requiring documentation.

A documentation ticket will be opened at https://github.com/qgis/QGIS-Documentation when this PR is merged.

Please update the description (not the comments) with helpful description and screenshot to help the work from documentors.
Also, any commit having [needs-doc] or [Needs Documentation] in will see its message pushed to the issue, so please be as verbose as you can.

Thank you!

[rouault]

Slight side tracking: determining which side is the top and which one is the bottom depends on the convention of the editing tool. Your change assumes that the editor looks at polygons from the top, which is a reasonable one. But for a 3D shape and 3D editor, determining the top from the inside cannot be just done by looking at the coordinates of the vertex of the ring (for vertical polygons,, the usual method would return a "random" value). You need to take into account the whole shape to determine the top from the bottom.

The addition of the constraint "exterior boundary = counter-clock-wise" is something "new" in SFA 1.2.1 - 06-103r4 (https://www.ogc.org/standard/sfa/). It is not present in SFA 1.1.0 - 05-126, which might perhaps explain why GEOS doesn't take that into account.

[lbartoletti]

Thanks @rouault for your insight.

Slight side tracking: determining which side is the top and which one is the bottom depends on the convention of the editing tool. Your change assumes that the editor looks at polygons from the top, which is a reasonable one.

Yes, what I should have written was that QGIS now allows you to mix the order of the interior rings; this is the most problematic aspect for me. Whether CW or CCW, all interior rings must have the same direction, which is not the case today.

But for a 3D shape and 3D editor, determining the top from the inside cannot be just done by looking at the coordinates of the vertex of the ring (for vertical polygons,, the usual method would return a "random" value). You need to take into account the whole shape to determine the top from the bottom.

Do you think, it's applicable for these tools? I suggest enforcing the digitalization of new geometry, but we can still apply CW or CCW with some algorithms/expressions in QGIS.

The addition of the constraint "exterior boundary = counter-clock-wise" is something "new" in SFA 1.2.1 - 06-103r4 (https://www.ogc.org/standard/sfa/). It is not present in SFA 1.1.0 - 05-126, which might perhaps explain why GEOS doesn't take that into account.

Yes, you're right. I was a baby geomatician when the new standard has landed. :) To be clear, I don't blame GEOS.
However, I don't see what's wrong with forcing digitization in a certain direction - that of the current standard - while being lax about what we receive/read. Does that sound right to you, or am I missing something?

[rouault]

Does that sound right to you, or am I missing something?

What you propose is probably OK and I don't strongly object, but I'd say the tools you're trying to be compatible with should only enforce the OGC rules when displaying/working with 3D volumes, not 2D surfaces where the concept of top/outside or bottom/inside doesn't really make sense. But if we are trying to be fully compatible with those tools (can be they named :-) ?), that's a lot of effort because they're plenty of places in QGIS, GDAL or GEOS where changes would have to be made.

[lbartoletti]

Does that sound right to you, or am I missing something?

What you propose is probably OK and I don't strongly object, but I'd say the tools you're trying to be compatible with should only enforce the OGC rules when displaying/working with 3D volumes, not 2D surfaces where the concept of top/outside or bottom/inside doesn't really make sense. But if we are trying to be fully compatible with those tools (can be they named :-) ?), that's a lot of effort because they're plenty of places in QGIS, GDAL or GEOS where changes would have to be made.

At the very least, the case that poses a real issue for me is the potential disorder when digitizing the rings and parts.
One of the tools, SFCGAL, is very strict about this, and this also applies to other proprietary tools whose names I might not necessarily know (mining activities in mind). There's an expectation, which seems logical to me, of having the exterior rings in one direction and the interior rings in the opposite direction.
Even if the new standard mentions CCW, I could ignore it. However, I would like it so that if the exterior ring is in one direction, all interior rings are in the opposite direction, and the parts follow the direction of the exterior ring.

[rouault]

However, I would like it so that if the exterior ring is in one direction, all interior rings are in the opposite direction, and the parts follow the direction of the exterior ring.

that indeed makes sense

[lbartoletti]

However, I would like it so that if the exterior ring is in one direction, all interior rings are in the opposite direction, and the parts follow the direction of the exterior ring.

that indeed makes sense

I've made the changes so that the parts are in the same direction as the first exterior ring and the reverse for the interior rings, and I'm not forcing the direction of the main exterior ring.

[github-actions]

@lbartoletti
A documentation ticket has been opened at qgis/QGIS-Documentation#8658
It is your responsibility to visit this ticket and add as much detail as possible for the documentation team to correctly document this change.
Thank you!

[agiudiceandrea]

Hi @lbartoletti, please see #58333.