Technical Details:
Serverless Backend -> AWS Lambda function written in python that is accessed via API gateway. Also uses S3 for pre-stored geographic shape files for specific locations. Uses shapely package for geographical/spatial analysis.
- Gist of Python code:
* Receives inputs from frontend (commute method, commute time, commute location, budget most important)
* If coordinates of commute locations are not cached in s3, reaches out to Google Geocode API to get them.
* Based on commute inputs, calculates the intersectional commute area for all housemates. If this is 0, then
the program can return 0 properties to the user early as there are no houses there...
* The intersectional commute area is often a multipolygon (multiple shapes), and many APIs do not accept multiple shapes as a query input. Many APIs also do not like complex geometries (often only accepting x max points).
* To get around this, each polygon within the multipolygon is simplified with the Douglas-Peucker algorithm iteratively until the points in the resulting polygons within the multipolygon sum to a total of (x - 2(n-1)) where x is the maximum allowed shape points on an API and n is the number of individual polygons.
The remaining simplified multipolygon is then connected with nearly-infinitely-small bridges [which adds 2 points between each shape, hence x - 2(n-1)].
Once this geometry is calculated, valid properties can be found and returned to the user based on bedroom, bathroom, budget etc preferences.
Vanilla JS frontend - uses leaflet.js for the map functionality.
wow I’ve been looking for something like this for an upcoming move! Would be nice to also have options to do more granular filtering, like number of bathrooms.
You can do that! You can filter once the search results are returned - but thanks, I will try to make this more obvious within the UI and allow filters earlier on :)
As this is more of an MVP for now, I mainly rely on APIs such as Google Geocoding API, Mapbox, etc in the backend, but this is certainly limited. If there was a dedicated user base who found value in this, I'd look to increase granularity/accuracy and reduce costs by working with raw spatial data and doing GIS/network analysis type stuff in-house instead.
Although, I do use OSM on the site as a map view source!
Technical Details: Serverless Backend -> AWS Lambda function written in python that is accessed via API gateway. Also uses S3 for pre-stored geographic shape files for specific locations. Uses shapely package for geographical/spatial analysis. - Gist of Python code: * Receives inputs from frontend (commute method, commute time, commute location, budget most important)
* If coordinates of commute locations are not cached in s3, reaches out to Google Geocode API to get them.
* Based on commute inputs, calculates the intersectional commute area for all housemates. If this is 0, then the program can return 0 properties to the user early as there are no houses there...
* The intersectional commute area is often a multipolygon (multiple shapes), and many APIs do not accept multiple shapes as a query input. Many APIs also do not like complex geometries (often only accepting x max points).
* To get around this, each polygon within the multipolygon is simplified with the Douglas-Peucker algorithm iteratively until the points in the resulting polygons within the multipolygon sum to a total of (x - 2(n-1)) where x is the maximum allowed shape points on an API and n is the number of individual polygons.
The remaining simplified multipolygon is then connected with nearly-infinitely-small bridges [which adds 2 points between each shape, hence x - 2(n-1)].
Once this geometry is calculated, valid properties can be found and returned to the user based on bedroom, bathroom, budget etc preferences.
Vanilla JS frontend - uses leaflet.js for the map functionality.
wow I’ve been looking for something like this for an upcoming move! Would be nice to also have options to do more granular filtering, like number of bathrooms.
I hope it can be of good use!
You can do that! You can filter once the search results are returned - but thanks, I will try to make this more obvious within the UI and allow filters earlier on :)
Excellent! What are your data sources? OSM?
Thanks Trufi!
As this is more of an MVP for now, I mainly rely on APIs such as Google Geocoding API, Mapbox, etc in the backend, but this is certainly limited. If there was a dedicated user base who found value in this, I'd look to increase granularity/accuracy and reduce costs by working with raw spatial data and doing GIS/network analysis type stuff in-house instead.
Although, I do use OSM on the site as a map view source!