Allow ST_Boundary() to return start/end point of closed LineString

[strk]

At the moment ST_Boundary() returns an EMPTY geometry when called against closed linestrings. This makes sense from a topological point of view, but in some cases you may want to get an official startpoint and it's currently annoyingly hard to get that, especially for multilines.

I think we do support some kind of BoundaryNodeRule for other operations, and then I guess it would be nice to have for ST_Boundary as well.

What do you think DrJTS ?

[strk]

The BoundaryNodeRule is supported by ST_Relate: ​https://postgis.net/docs/ST_Relate.html

[komzpa]

I believe ST_StartPoint should be allowed to run on any geometry instead.

[strk]

What would your ST_StartPoint return against a MULTILINESTRING((0 0, 20 0, 10 5, 0 0),(100 0, 120 0)) ?

An ST_Boundary() with BoundaryNodeRule 2 would return MULTIPOINT((0 0), (100 0), (120 0)) . Note how BOTH start and end points are returned for the second geometry…

[mdavis]

Replying to strk:

What would your ST_StartPoint return against a MULTILINESTRING((0 0, 20 0, 10 5, 0 0),(100 0, 120 0)) ?

Perhaps ST_StartPoint could be extended to support MultiLineStrings as input (returning a MultiPoint containing the start points of each element)?

[mdavis]

This is the one of the reasons that the BoundaryNodeRule concept was introduced in JTS: it's handy to be able to use it to get the node point of closed LineStrings. (And it also allows getting the endpoints of degree 1 in a MultiLineString, rather than being hostage to the less-useful Mod-2 rule of the standard.

So +1 for adding BoundaryNodeRule support to ST_Boundary, especially since it is already provided for ST_Relate.