tianhuil

From https://hackmd.io/@nirvana/deploy_python_m1

Build image with linux platform for heroku servers. Replace {NAME_OF_HEROKU_APP} with your own tag:

docker buildx build --platform linux/amd64 -t {NAME_OF_HEROKU_APP} .

Tag your app with the url for your apps registry. make sure to use the name of your Heroku app in the url and tag name:

docker tag {NAME_OF_HEROKU_APP} registry.heroku.com/{NAME_OF_HEROKU_APP}/web

tianhuil

// Type assertions, before I discovered https://www.npmjs.com/package/type-assertions

type AssertExtends<A, B> = A extends B ? true : false
export const assertExtends = <A, B>(_: AssertExtends<A, B>): void => {}
export const assertInstance = <B>(
  a: any,
  _: AssertExtends<typeof a, B>
): void => {}

assertExtends<'abc', string>(true)

tianhuil

# Poetry needs to be installed globally

poetry init

poetry add \
  matplotlib \
  pandas \
  wordcloud \
  tqdm \
  python-dotenv \

tianhuil

Switching to DNS over TLS on macOS

How to configure your Mac to use DNS over TLS in five easy steps:

1. Install Stubby with Homebrew (https://dnsprivacy.org/wiki/display/DP/DNS+Privacy+Daemon+-+Stubby):

    brew install stubby
   

2. Edit the configuration file:

tianhuil

```
$ git push server main
+ export DOKKU_HOST_ROOT=/home/dokku
+ DOKKU_HOST_ROOT=/home/dokku
+ export DOKKU_DISTRO
++ . /etc/os-release
++ echo ubuntu
+ DOKKU_DISTRO=ubuntu
+ export DOCKER_BIN=docker
+ DOCKER_BIN=docker

tianhuil

# To run this command:
# curl https://gist.githubusercontent.com/tianhuil/0aa9b265f55413dc7198/raw > setup_digitalocean.sh
# . setup_digitalocean.sh

# Update sudo apt-get
sudo apt-get update

# Installing scientific Python
sudo apt-get -y install --fix-missing build-essential python-dev python-numpy python-setuptools python-scipy libatlas-dev
sudo apt-get -y install --fix-missing build-essential python-sklearn

tianhuil

-149.8935557,61.21759217,Starbucks - AK - Anchorage  00001,"601 West Street_601 West 5th Avenue_Anchorage, Alaska 99501_907-277-2477"
-149.9054948,61.19533942,Starbucks - AK - Anchorage  00002,"Carrs-Anchorage #1805_1650 W Northern Lights Blvd_Anchorage, Alaska 99503_907-339-0500"
-149.7522,61.2297,Starbucks - AK - Anchorage  00003,"Elmendorf AFB_Bldg 5800 Westover Avenue_Anchorage, Alaska 99506"
-149.8643361,61.19525062,Starbucks - AK - Anchorage  00004,"Fred Meyer - Anchorage #11_1000 E Northern Lights Blvd_Anchorage, Alaska 995084283_907-264-9600"
-149.8379726,61.13751355,Starbucks - AK - Anchorage  00005,"Fred Meyer - Anchorage #656_2300 Abbott Road_Anchorage, Alaska 99507_907-365-2000"
-149.9092788,61.13994658,Starbucks - AK - Anchorage  00006,"Fred Meyer - Anchorage (Dimond) #71_2000 W Dimond Blvd_Anchorage, Alaska 995151400_907-267-6700"
-149.7364877,61.19533265,Starbucks - AK - Anchorage  00007,"Safeway-Anchorage #1817_7731 E Northern Lights Blvd_Anchorage, Alaska 99504_907-331-1700"
-149.8211,61.2156

tianhuil

Suppose we wish to understand if a given X/Y pair falls within a set of non-intersecting arbitrary polygons. For example, we are trying to see if a given lat,long falls within a Michigan minor civil division. There are 1500+ such divisions each one of which could be a polygon with hundreds or thousands of points. The following is a practical solution that is fast enough to return to a user.

1. Break up Michigan into $N$ box oriented tiles. We can easily calculate if x, y will only fall within a tile based on the value (x % WIDTH, y % HEIGHT). Select $N$ such that most (e.g. ~60%?) tiles are fully contained within a polygon, and the vast majority (e.g. 95%) are in no more than 4 polygons. I imagine this would happen easy at $N$ less than 100,000.
2. Preprocess the polygon shape files using Shapely to determine a map

tianhuil

import gzip
import os
from itertools import islice
import argparse

# from https://stackoverflow.com/a/41333436/8930600
def grouper(iterable, n):
   iterator = iter(iterable)
   while True:
     group = tuple(islice(iterator, n))

tianhuil

from sklearn.base import BaseEstimator, TransformerMixin

class ClassifierTransform(BaseEstimator, TransformerMixin):
    def __init__(self, clf):
        self.clf = clf
        
    def fit(self, X, y=None):
        self.clf.fit(X, y)
        return self