Tijdens het gebruik van TransformerGeocodeToCoordinates zag ik dat df_with_xy.index = index fout ging omdat de df_with_xy groter was geworden dan de input df. Dit komt omdat er duplicaten in xy_information zitten (geen idee waarom), maar dit is dus gefixt met een drop_duplicates.

pyproj < 3 did not work for Apple devices (with M1 chip), and the package works on Windows without this requirement. Furthermore, the <3 version requirement in setup.py is inconsistent with the 3.3.* requirements in requirements.txt, so it makes sense to remove the <3 requirement in setup.py

Make the URL in SpoortakMapServices configurable to allow it to read multiple layers. Can be combined wihth FeatureServerOverview for added user simplicity

Added a mapping and inspection classes to work with spoortak model version 3 up to and including version 17

Note the spoortak models data has not yet been included. We are verifying if they can be distributed or not trough github. Thus the integration tests will fail due to missing data.

In particular:

• Base all 3 objects on a GeoDataFrame under self.data. PlottingPoints will require the most work here.
• Fix the popup argument in a consistent manner (none, str or tuple of strings)
• • Fix the data argument in a consistent manner (string or geopandas geodataframe)
• General support for self.color?

Additionally, if possible we can try to make PlottingPoints work for pandas dataframes too.

This started as an upgrade to the version number, but also upgraded the readme and added a requirement. The setup instructions were also verified on a Windows machine and are working now.

Making requests results in a lot of warnings:

InsecureRequestWarning: Unverified HTTPS request is being made to host 'mapservices.prorail.nl'. Adding certificate verification is strongly advised. See: https://urllib3.readthedocs.io/en/latest/advanced-usage.html#ssl-warnings InsecureRequestWarning,

Set up spoorkaarten outside of a context manager. Suggested approach would be to make the below workable.

map = SpoorKaart() map.add(plottable object) map.add(plottable object2) map.show()

map.save('save_location.html')

loguru 0.4.0 is not available on conda-forge only 0.4.1 old rtree will not properly install on windows, needed an updated version

simplified the readme to just conda install with conda-forge as extra channel. This will now work in windows as well as mac/unix

Switched to Python 3.8 in the readme (because I could, all the unit tests were still green)

Is dit iets wat we willen? Zelf vond ik dit een handige richting om ook andere features eruit te kunnen halen. Mochten we openspoor daar voor klaar willen maken dan moeten we nog regelen dat de URL's die je hier vindt gebruikt kunnen worden in latere queries: de meeste logica daarvoor bestaat al, het moet alleen een stukje generieker dan. Zomaar een idee, suggesties zijn welkom!

Openspoor seems to be incompatible with more recent versions of numpy (1.22 and above). This is further complicated by the fact that GDAL and Fiona are installed through conda instead of Pip. More futureproof versions (or installation procedures) would be helpful to keep up with recent developments.

The interpolation in the transform method in TransformerSpoortakToCoordinates.py is not valid. At the start of the geometry the lokale km is not (always) zero. So it should use the travelled distance on that geometry (difference in lokale_km from start and the point of interest). But that would still not be a correct way to do it because than it assumes that a geocode km is linear, which is not the case. We should use M-values for this.

There now exists a layer with both spoortakken and geocode information in it; this can be used as a replacement for the current xy_to_geocode api calls. Gathering this layer and interpolating based on the information in here would therefore be a more efficient way to get geocodes for a lot of coordinates. Some effort in this direction was already done by the beeldherkenning configuratie team. They pointed out that one fix would be is to use the geocode information (stored in the M (measurement) value in the featureserver can overwrite the Z-value in the data (which is not that useful), from where it can be used to interpolate along the linestrings.

https://mapservices.prorail.nl/arcgis/rest/services/Geleidingssysteem_010/FeatureServer/34

transformer PUICMapservices does not work with the most recent featureservers anymore. It was designed with the layers as described below with original_* in mind; the new_* versions are the latest versions of these layers.

new_wisselbeen = MapServicesQuery(puic_mapservices.wisselkruisingbeen_query.layer_url.values[0]).load_data() original_wisselbeen = MapServicesQuery("https://mapservices.prorail.nl/arcgis/rest/services/Geleidingssysteem_007/FeatureServer/12/").load_data() print(new_wisselbeen.shape, original_wisselbeen.shape)

new_spoortakdeel = MapServicesQuery(puic_mapservices.wisselkruisingbeen_query.layer_url.values[0]).load_data() original_spoortakdeel = MapServicesQuery("https://mapservices.prorail.nl/arcgis/rest/services/Geleidingssysteem_007/FeatureServer/13/").load_data() print(new_spoortakdeel.shape, original_spoortakdeel.shape)

Joost and Oscar have tried to get openspoor working with versions >= 3, but one acceptance test does not fully work then (test_acceptance_TransformerCoordinatesToSpoor). Most points are projected as expected, but 2 are not (of which 1 cannot be fixed with a larger buffer distance). This is strange, since a point just 4 meters next to it can be projected successfully.

Reason for fixing: pyproj < 3 does not work on M1 chip (macbook)