Wang–Müller line generalization algorithm in PostGIS

This is Wang–Müller line generalization algorithm implementation in PostGIS. Following "Line generalization based on analysis of shape characteristics" by the same author, 1998.

Status

It mostly works. Read mj-msc-full.pdf for visual examples and possible gotchas.

Structure

There are be 2 deliverables:

• wm.sql, the implementation.
• paper mj-msc-full.pdf, a MSc thesis, explaining it.

It contains a few supporting files, notably:

• tests.sql synthetic unit tests.
• test-rivers.sql tests with real rivers.
• Makefile glues everything together.
• layer2img.py converts a PostGIS layer to an embeddable image.
• aggregate-rivers.sql combines multiple river objects (linestrings or multilinestrings) to a single one.
• init.sql initializes PostGIS database for running the tests.
• rivers-*.sql are national dataset snapshots of rivers (Makefile contains code to update them).
• ... and a few more files necessary to build the paper.

Running

make help lists the select commands for humans. As of writing:

# make help
mj-msc-full.pdf    Thesis for publishing
test               Unit tests (fast)
test-rivers        Rivers tests (slow)
refresh-rivers     Refresh river data from national datasets
clean              Clean the current working directory
clean-tables       Remove tables created during unit or rivers tests
help               Print this help message

To execute the algorithm, run:

• make test for tests with synthetic data.
• make test-rivers for tests with real rivers. You may adjust the rivers and data source (e.g. use a different country instead of Lithuania) by changing the Makefile and the test files. Left as an exercise for the reader.

N.B. the make test-rivers fails (see test-rivers.sql), because with higher dhalfcircle values, the unionized river (salvis) is going on top of itself, making the resulting geometry invalid.

Building the paper (pdf)

# make -j$(nproc) mj-msc-full.pdf

mj-msc.tex results in mj-msc-full.pdf, which will be at some point published to this repo. It needs quite a few dependencies, including a functioning Docker environment, postgresql client, geopandas, pygments, osm2pgsql, poppler, and a "quite extensive" LaTeX installation. Tested on Debian 11.

in-container script may be helpful if the above sounds like too much.

Contributing

This repository will soon be frozen and does not accept contributions. Please fork it. If fork has improved the algorithm substantially, feel free to ping me, I will link to it in this README.

Credit

Nacionalinė Žemės Tarnyba for the river data sets.

License

GPLv2 or later.