michalc

from pprint import pprint
import boto3

repository_name = 'REPLACE_ME'

client = boto3.client('ecr', region_name='eu-west-2')

image_details = (
    image_detail
    for page in client.get_paginator('describe_images').paginate(

michalc

{
    init: function(elevators, floors) {

        const intSort = function(a, b) {
            if (a < b) return -1;
            if (a > b) return 1;
            return 0;
        }

        const floorWantsUp = function(floorNum) {

michalc

# Restores all objects under a prefix in an S3 bucket to the version they were before being deleted.
# Works by deleting the latest delete marker for the objects.
import boto3

bucket = 'the-bucket'
prefix = 'the-prefix/'
region = 'eu-west-2'

client = boto3.client('s3', region_name=region)
paginator = client.get_paginator('list_object_versions')

michalc

# docker run --rm -it -p 5432:5432 -e POSTGRES_PASSWORD=password postgres:14

import pprint
import threading
import time
from concurrent.futures import ThreadPoolExecutor
from contextlib import contextmanager

import psycopg2
import psycopg2.extras

michalc

import json
import os
from urllib.parse import urlparse, parse_qsl

import requests
from rich import box
from rich.console import Console
from rich.table import Table

with open(f'{os.environ["HOME"]}/.cf/config.json') as f:

michalc

# Deletes objects in bulk using boto3's delete_objects, but using multiple threads to achieve some
# parallelism - in spite of the GIL multiple HTTP requests to S3 should happen at the same time.
# Instead of looping over all keys under the root prefix, it walks the tree of keys of delimiter-
# defined "folders" in a depth-first way, which allows each page to be processed by a separate
# thread as its discovered. Depth-first is done because the depth is limited by the maximum key
# size of 1024 in S3, and so means that there is a limit to the memory used by the algorithm to
# store the next requests to make. This would not be the case with breadth-first because there is
# no limit to how many keys are in any folder.
#
# To do the search in parallel, each bit of work (i.e. an HTTP request to fetch a page of keys

michalc

import base64
import json

import boto3
from cryptography.hazmat.primitives.ciphers.aead import AESGCM

s3 = boto3.resource('s3')
bucket = s3.Bucket('my-bucket')
kms_client = boto3.client('kms')

michalc

# Often it's claimed that a Zip 2.0 file cannot be bigger than 4GiB
# Here's how to make one that's just under 8GiB
from datetime import datetime
from stream_zip import stream_zip, ZIP_32

now = datetime.now()
perms = 0o600

def files():
    for i in range(0, 0xffff):

michalc

class MyPipeline(_PipelineV2):
    # Everything is a method so nothing happens on import time for flexibility (although possibly
    # does a bit of discovery magic... need to think about that...)
    # Everything is a _static_ method: nothing on self since things are run on different bits of hardware,
    # and gets any run-time dependencies injected in
    #
    # _PipelineV2 won't actually have any code: other parts of the system will interrogate its
    # subclasses as needed. For example
    # - Code in Data Flow would construct a DAG
    # - The test harness would the run this and upstream pipelines synchronously

michalc

from contextlib import contextmanager
from ctypes import POINTER, cdll, c_char_p, c_void_p, c_int, create_string_buffer, byref
from sys import platform

# Uses a _little_ endian CTR counter, which OpenSSL doesn't directly support.
# Could be used to decrypt AES-encrypted ZIP files
def decrypt_aes_256_ctr_little_endian(
    key, ciphertext_chunks,
    get_libcrypto=lambda: cdll.LoadLibrary({'linux': 'libcrypto.so', 'darwin': 'libcrypto.dylib'}[platform])
):