Geofileops

Several improvements and fixes. Especially if you process large files (> 5 GB) some of
them will be very useful.

Improvements

- Add options to choose the buffer style to use (37).
- Add option in makevalid to specify precision to use.
- Add support for dissolve with groupby_columns=None + explodecollections=True
- add batchsize parameter to all geo operations to be able to limit memory usage if
needed (38).
- Decrease memory usage for most operations.
- Run processing in low CPU priority so the computer stays responding.

Bugs fixed

- Fix error when processing very large files for buffer, convexhull, dissolve and
simplify (IO timeouts).
- Fix error in simplify when points on geometries should be kept.
- Don't give an error when result is empty for sql based operations.
- Explodecollections=True doesn't work on operations on 2 layers.

In this release, the main change is a new operation that has been added: nearest_join.

Improvements

- New operation nearest_join() (12).
- Improve performance for files where the rowid isn't consecutive (14).
- Improve documentation, mainly for the select operations.
- Add check in get_layerinfo on column names containing illegal characters (='"').
- Suppress fiona warning "Sequential read of iterator was interrupted."
- Dissolve: output is sorted by a geohash instead of random, which
improves performance when reading the file afterwards.

Changes that break backwards compatibility

- Drop support for spatialite < 5 and geopandas < 0.9.
- Bugfix: for the dissolve operation, when rows in the groupby columns
contained NULL/None/NaN values, those rows were dropped. This is the
default behaviour of (geo)pandas, but as the behaviour doesn't feel
very intuitive, it was decided to divert from it.

Improved performance for all operations that involve overlays between 2 layers
(intersect, union, identity/split,...).

Improvements

- Improved performance for all operations that involve overlays between 2
layers (intersect, union, identity/split,...). Especially if the input files are in
Geopackage format the improvement should be significant because in this case
the input data isn't copied to temp files anymore.
- Added the method geofile.execute_sql() to be able to execute a DML/DDL sql
statement to a geofile.
- Smaller code cleanups to improve readability on several places.

Bugs fixed

- In the identity/split and union operations, in some cases (mainly when input layers
had self-intersections) intersections in the output were unioned instead of
keeping them as seperate rows.
- When using an input file with multiple layers (eg. a geopackage), this
sometimes resulted in data errors.
- Fix an error in the lang simplification algorithm that occured in some edge
cases.

This release mainly brings improvements and bugfixes when using the Lang
simplification algorithm.

Additionally, this is the first version that gets a conda package.

Improvements

- Support for all geometry types for Lang simplification
- Some code restructuring + code cleanup in util (backwards compatible for now, but
isn't public interface anyway)
- Improve test coverage

Bugs fixed

- preserve_topology wsn't always respected in Lang simplification
- clip_on_tiles=False in dissolve resulted in some polygons not being
dissolved. Because there is no easy solution, parameter is deprecated
and always treated as True.

This release is mainly focused on improvements to the dissolve and simplify
operations.

It is important to note that for the dissolve operation the default parameters
behave slightly different, so the change is not backwards compatible!

Most important changes

- Dissolve
- Different handling of default parameters (not backwards compatible!)
- Better performance
- Several bugfixes
- Simplify: add support for different simplification algorythms:
ramer-douglas-peucker, visvalingam-whyatt and lang.
- Many improvements to documentation
- Improvements to type annotations
- Some new undocumented (so use at own risk!) functions were added to
geofileops.vector_util.
- simplify_ext: an "extended" simplify function:
- Supports multiple simplification algorythms: ramer-douglas-peucker,
visvalingam-whyatt and lang.
- Option to specify a "points-to-keep geometry": points/nodes of the
geometries to be simplified that intersect it won't be removed by
the simplify operation.
- count_number_points: count the number of coordinates that any shapely
geometry consists of

This version contains quite some fixes and small improvements. The most
important ones are listed here.

Improvements

- Stop using sqlite WAL output format as default: gives locking issues when
using files on a file share
- Improve handling of output geometry type
- Add scripts, dependencies,... to use pytest-cov to report unit test coverage
- Add first version of sphynx documentation
- Improve test coverage
- Improve benchmark tests

Bugs fixed

- Negative buffer shouldn't output rows with empty geometries
- Dissolve gave an error in some specific cases
- Join_by_location operation gave wrong results