view schema/updates/1466/01.fix_ISRSrange_axis.sql @ 5513:68358e4603c8

Use current axis only for calculating bottleneck areas This is a fixup of rev. cf25b23e3eec, which introduced historic data for the waterway axis but missed to take this into account in the calculation of bottleneck areas, leading to sometimes excessive runtime and bad results due to multiple (almost) equal axis geometries being considered as candidates in the bottleneck stretch. The database migration tries to recalculate all bottleneck areas, while some might fail that did not fail on import. A warning message is emitted for these and the area is left untouched.
author Tom Gottfried <tom@intevation.de>
date Tue, 19 Oct 2021 13:12:39 +0200
parents
children
line wrap: on
line source

CREATE OR REPLACE FUNCTION ISRSrange_axis(
    stretch isrsrange,
    tolerance float
    -- in m, up to which linestrings will be connected at their boundary
) RETURNS geometry
AS $$
DECLARE z int;
DECLARE result_geom geometry;
BEGIN
    -- Find best matchting UTM zone
    z = best_utm(stretch);

    CREATE TEMP TABLE axis AS
        SELECT row_number() OVER () AS id,
                geom AS wtwaxs,
                ST_Boundary(geom) AS bdr
            FROM waterway.waterway_axis,
                ST_Dump(ST_Transform(wtwaxs::geometry, z))
            WHERE validity @> current_timestamp;
    CREATE INDEX axs_bdr ON axis USING GiST (bdr);
    ANALYZE axis;

    WITH RECURSIVE
        -- In order to guarantee the following ST_Covers to work,
        -- snap distance mark coordinates to axis
        points0 AS (
            SELECT ST_ClosestPoint(
                    wtwaxs,
                    ST_Transform(geom, z)) AS geom
                FROM ST_Dump(ISRSrange_points(stretch)), (
                    SELECT ST_Collect(wtwaxs) AS wtwaxs
                        FROM axis) AS ax),
        -- Ensure two distinct points on axis have been found
        points AS (
            SELECT geom
                FROM points0
                WHERE 2 = (SELECT count(DISTINCT geom) FROM points0)),
        axis_snapped AS (
            -- Iteratively connect non-contiguous axis chunks
            -- to find the contiguous axis on which given distance marks lie
            (SELECT ARRAY[id] AS ids, wtwaxs
                FROM axis, points
                WHERE ST_Intersects(
                    ST_Buffer(axis.wtwaxs, 0.0001), points.geom)
                FETCH FIRST ROW ONLY)
            UNION
            -- Connect endpoint of next linestring with closest
            -- endpoint of merged linestring until a contiguous
            -- linestring connecting both distance marks is build up
            (SELECT refids || id,
                    ST_LineMerge(ST_Collect(ARRAY(
                        -- Linestring build up so far
                        SELECT refgeom
                        UNION
                        -- Fill eventual gap
                        SELECT ST_ShortestLine(
                                ST_Boundary(refgeom), bdr)
                        UNION
                        -- Linestring to be added
                        SELECT geom)))
                FROM axis_snapped AS axis_snapped (refids, refgeom),
                    axis AS axis (id, geom, bdr),
                    (SELECT ST_Collect(points.geom) AS pts
                        FROM points) AS points
                WHERE id <> ALL(refids)
                    AND ST_DWithin(
                        ST_Boundary(refgeom), bdr, tolerance)
                    AND NOT ST_Covers(ST_Buffer(refgeom, 0.0001), points.pts)
                ORDER BY ST_Boundary(refgeom) <-> bdr
                FETCH FIRST ROW ONLY)),
        axis_segment AS (
            -- Fetch end result from snapping
            SELECT wtwaxs AS line
                FROM axis_snapped,
                    (SELECT ST_Collect(points.geom) AS pts
                        FROM points) AS points
                -- Return end result only if both distance marks were connected
                WHERE ST_Covers(ST_Buffer(wtwaxs, 0.0001), points.pts))
        -- Use linear referencing to clip axis between distance marks.
        -- Simplification is used to work-around the problem, that
        -- ST_LineSubstring might generate very small line segments at an
        -- end of the resulting linestring, that significantly differ from
        -- the direction of the input linestring due to finite precision
        -- of the calculation. The generated small segment of the
        -- resulting line would lead e.g. to unexpected results in an area
        -- generated by ISRSrange_area().
        SELECT ST_SimplifyPreserveTopology(ST_LineSubstring(
                    axis_segment.line, min(fractions.f), max(fractions.f)),
                0.0001) AS line
        INTO STRICT result_geom
        FROM axis_segment, LATERAL (
            SELECT ST_LineLocatePoint(axis_segment.line, points.geom) AS f
                FROM points) AS fractions
        GROUP BY axis_segment.line;

    -- Drop temporary table to avoid side effects on PostgreSQL's MVCC,
    -- because otherwise subsequent invocations of the function will not see
    -- changes on the underlying waterway.waterway_axis that might have
    -- occured.
    DROP TABLE axis;

    RETURN result_geom;
END;
    $$
    LANGUAGE plpgsql
    PARALLEL RESTRICTED;