lwt_RemIsoEdge() bug - Nodes not restored so to be Isolated

[esseffe]

lwt_RemIsoEdge() does not correctly restore Nodes to an "Isolated" state after removing the Edge and this causes subsequent calls to lwt_RemIsoNode() to unexpectedly fail raising a "non-isolated node" exception. The following SQL snippet reproduces the issue:

BEGIN;
SELECT CreateTopology('topo', 4326, 0);
SELECT ST_AddIsoNode('topo', NULL, ST_SetSRID(ST_MakePoint(1, 1), 4326));
SELECT ST_AddIsoNode('topo', NULL, ST_SetSRID(ST_MakePoint(2, 2), 4326));
SELECT ST_AddIsoEdge('topo', 1, 2, ST_GeomFromText('LINESTRING(1 1, 2 2)',4326));
SELECT ST_RemoveIsoEdge('topo', 1);
SELECT ST_RemoveIsoNode('topo', 1);
ROLLBACK;

[strk]

I've tested this with 2.1 and it also does not update nodes on RemoveIsoEdge, so that part is not a regression (but still a bug). I guess it also happens in 2.0.

What's different in 2.1 is that it does let you remove the node, which is due to RemoveIsoNode in 2.1 checking for isolation by querying the edge table rather than by looking at the "containing_face" field.

So this would be a regression in RemoveIsoNode and a long-standing bug in RemoveIsoEdge.

[strk]

esseffe do you confirm that removing the isolated edge with ST_RemEdgeModFace makes the test succeed ? (there's really no reason to use ST_RemIsoEdge…)

[strk]

(In [14335]) Set edge endpoints as isolated on ST_RemoveIsoEdge

See #3351

[strk]

(In [14336]) Set edge endpoints as isolated on ST_RemoveIsoEdge

See #3351

[strk]

(In [14337]) Set endpoints isolation in lwt_RemIsoEdge (ST_RemoveIsoEdge)

See #3351

[strk]

(In [14338]) Set endpoints isolation in lwt_RemIsoEdge (ST_RemoveIsoEdge)

See #3351

[strk]

r14335 in 2.0 (2.0.8), r14336 in 2.1 (2.1.9), r14337 in 2.2 (2.2.1) and r14338 in trunk (2.3.0)