Skip to content

Latest commit

 

History

History
180 lines (136 loc) · 7.93 KB

README.md

File metadata and controls

180 lines (136 loc) · 7.93 KB
Raw

geoviz-notebooks

image

image

Python tool for analyzing geospatial data in cities.

This repo consists of three components that can be run separately or together:

  • https server and UI that allow running the whole analysis process automatically and browsing the result maps from the comfort of your favorite web browser
  • import scripts that import a variety of open geospatial datasets for your favorite city
  • notebook or export script that
    • combine the datasets to form an index value per H3 hex, and
    • create a Kepler.gl H3 map of the index

It is recommended to use the UI, if you want to just get a fancy map, but you may run the scripts and notebook manually if you want more control in what is included in your analysis.

Requirements

  • Docker
  • Docker-compose

OR

  • Python >= 3.8
  • Osm2pgsql
  • PostGIS accepting connections at localhost:5432

How to get started with docker

If you wish to run the UI to run analyses with a few clicks, you may start the database and development server locally by typing

docker-compose up dev

Set up the needed username and password as in configuration.

If you wish to run the analysis notebook manually, you will get the database and notebook up by typing

docker-compose up notebook

The https production server can be started by

cp server/flask.subdomain.conf server/swag/nginx/proxy-confs/
docker-compose up serve

Do note that the server requires a domain you must register to get a https certificate. Look at configuration for details.

How to get started without docker

If you're not running docker, we recommend creating your own conda env, pyenv, or pyenv which contains conda wheels. The last option should make installing all dependencies easier:

pyenv install miniconda3-latest
pyenv local miniconda3-latest
pip install -r requirements.txt

You also need to have PostGIS running. To run the Flask dev server and UI locally, after installing the basic requirements,

cd server
pip install -r requirements-serve.txt
flask run

Or, if you want to use the notebook instead of the UI, start the notebook server by

jupyter notebook

Configuration

Configuration is by the .env file or environment variables. You may copy .env.example to a file called .env and fill in your secrets.

To use the UI, set the username and password that allows access to the UI by setting the desired username and password hash in .env file or the corresponding environment variable.

If you wish to import data from the Flickr API, fill in your Flickr api key and secret in a .env file or the corresponding environment variable. The API key may also be set in the UI for each import run, but it will not persist to the server environment.

To get https certificates on AWS EC2, you need to add your own domain and subdomain in .env and your AWS access credentials in server/swag/dns-conf/route53.ini. If you use MFA, you have to create a separate non-MFA-role specific to your EC2 instance and instead add role_arn and credential_source=Ec2InstanceMetadata in route53.ini just like in AWS Config file. This will allow Swag to automatically retrieve and update your certificate. If you are running on a provider other than AWS, read Swag instructions and change DNSPLUGIN value at [docker-compose.yml#L20].

How to use the UI

Please, add your API keys and username/password hash in the configuration first. For development convenience you may set a clear-text password instead if running the dev container. Then, start your local dev server at http://localhost:5000 by

docker-compose up dev    # if you are running docker
cd server & flask run    # if you are not running docker

or your https server at https://yourdomain.com:443 by

cp server/flask.subdomain.conf server/swag/nginx/proxy-confs/
docker-compose up serve

image

You may import the city of your choice in the UI by

  1. selecting the datasets you desire,
  2. selecting your city from the autocomplete list,
  3. (optionally) adjusting the bounding box on the map, if the initial bbox doesn't look suitable for you,
  4. (optionally) adding a GTFS url for your city, if it is not known by the app already,
  5. clicking "Import Datasets".

The process will take a while depending on how many and which datasets you are importing. Importing small datasets such as Ookla and Kontur data is very fast, while using the Flickr API will be particularly slow and will keep you waiting for a long time.

image

You may look at how the request is processing by clicking the "View Log" button. Once the run is finished, click "View Results" to see all your datasets and the combination total index values on the hex map. All old import runs are listed under "Result maps".

How to use the command line

Importing

From the command line, you may do the import by running

docker-compose run notebook ./import.py Helsinki  # if you are running docker
./import.py Helsinki                               # if you are not running docker

or any other city. You may only import some datasets by --datasets parameter, e.g. ./import.py Helsinki --datasets "access gtfs ookla".

Do note that cities in bigger countries may be slow to import if the city is not available as a separate OSM extract. In that case, we will have to download the whole country. All other dataset sizes are determined by the size of the city.

If a city you want to import does not have a GTFS feed URL in scripts/import_gtfs.py GTFS_DATASETS variable, you may add the right URL manually there (please make a PR too), or alternatively run the import with the right URL as parameter, e.g.

./import.py Tallinn --gtfs http://www.peatus.ee/gtfs/gtfs.zip

Running analysis notebook and creating result map

The UI calculates the result map automatically if the import is started by the UI. Once the run is finished, your map is available by clicking "View Results" next to your import run.

You may also run export automatically at the end of import by calling ./import.py with the --export parameter, or run the export script after import has finished by

docker-compose run notebook ./export.py  # if you are running docker
./export.py                               # if you are not running docker

On the other hand, you calculate the results manually using the export notebook. The notebook is available at localhost:8888/tree/notebooks/export.ipynb .

Open the notebook to adjust

  • the datasets you wish to include
  • the column/sum/mean value you wish to use from each dataset
  • the weight of each dataset in the result map

You may just run the notebook as-is to get the default index that contains representative columns/statistics from each dataset, with equal weights for each dataset. This is the same setup that the export script uses for calculating the results. The resulting map is displayed in the notebook and saved as a standalone HTML map in server/maps/city_name.html.