Adds jitter. Upgrades progress bar.

Uses a new package for better progress bar behaviour in parallel loops.

Adds a jitter parameter for injecting shortest-path algorithm jitter for use on rectilinear grids.

Minor release

Minor release which adds a dependency to requirements.txt.

Minor Release

Minor version bump which fixes a test

Minor release (docs)

Minor release with some small refinements to the documentation.

Added a new function for calculating the average walking distance for a given impedance. (Since this may be more intuitive than reasoning about maximum distances and spatial impedances.)

Enables fastmath flags throughout

Enables numba fastmath flags throughout.

(This has no affect on how the methods are called.)

Splits compute_aggregated into compute_landuses and compute_stats

Splits compute_aggregated method into compute_landuses and compute_stats.

Docs and tests are updated accordingly.

Landuse methods are now also able to make use of parallel threads.

Sorts out numba-mode progress bar

Minor bump resolving edge cases for numba mode progress reporting.

Parallelises centrality measures

Incorporates parallelisation via numba prange for centrality measures.

Tests are passing and the code is presently being tested against real-world usage.

Note that the numba progress bars can no longer be sequential due to parallelisation, so this is reworked to a non-sequential update.

(tqdm progress bars in non-jitted python code still work fine, though).

Removes zero-length self-looped edges if encountered in `nX_simple_geoms`

Removes zero-length self-looped edges if encountered in nX_simple_geoms

This resolves a situation where a converted OSM graph with a zero length edge causes downstream errors. This arises because inferring geometry for a looped OSM edge (same start and end node) implies a zero-length geom: this results in geometry.Point instead of geometry.LineString. These zero-length edges are now removed.

Updates setup.py

Updates setup.py.