andrewgross

// Thinking token levels
const THINK_LEVEL = {
  HIGHEST: 31999,
  MIDDLE: 10000,
  BASIC: 4000,
  NONE: 0,
};

// Language cues for thinking intensity
const THINK_CUES = {

andrewgross

import base64
import json
import requests
from pathlib import Path
from typing import Union, Optional

class RemarkableError(Exception):
    pass

def upload_to_remarkable(

andrewgross

import os

from urllib.parse import urlparse

from pyspark.sql.functions import desc, asc
from pyspark.sql.types import (
    StructType,
    StructField,
    StringType,
    LongType,

andrewgross

import re
import pyspark.sql.types as T
from math import ceil


def repartition_for_writing(df):
    count = df.count()
    sampled_df = get_sampled_df(df, count=count)
    string_column_sizes = get_string_column_sizes(sampled_df)
    num_files = get_num_files(count, df.schema, string_column_sizes)

andrewgross

import datetime
import json

BUCKET_NAME = "<s3_bucket_name>"
INVENTORY_PREFIX = "<prefix_given_to_s3_inventory>" # Should have data/, hive/, and some dated folders inside of it
ACCOUNT_CUID = "<your_canonical_user_id_for_cross_account>"  # Account which is not the owner of S3 bucket, but trying to access it. Controls ROLE_ARN
ROLE_ARN = "<role_in_cross_account_that_can_assume_to_main_account>"


def role_arn_to_session(role_arn):

andrewgross

def get_files_per_partition(df, partition_key, file_type="parquet", compression="snappy", byte_array_size=256):
    rows = df.count()
    print "Dataset has {} rows".format(rows)
    schema = df.schema
    num_partitions = 1
    if partition_key is not None:
        num_partitions = df.select([partition_key]).distinct().count()
        print "Dataset has {} distinct partition keys".format(num_partitions)
        _df = df.drop(partition_key)
        schema = _df.schema

andrewgross

Hey,

This is going to be a bit of an esoteric ticket.  I noticed some strange behavior recently when comparing Spectrum and Redshift results on the same dataset.

Redshift Data: fare.txt
Parquet Data: fare.parquet

The parquet data was generated from fare.txt with PySpark using convert.py on Spark 2.2.0

Redshift Table Schema:

andrewgross

Keybase proof

I hereby claim:

• I am andrewgross on github.
• I am andrewwgross (https://keybase.io/andrewwgross) on keybase.
• I have a public key ASDsj8ie3y_QBUpm4aBzm-ty7Hr9w_Y5PtWIcLZfQlt9JQo

To claim this, I am signing this object:

andrewgross

import cmd
from urlparse import urlparse
import sys
import re
import subprocess

def write_to_clipboard(output):
    process = subprocess.Popen(
        'pbcopy', env={'LANG': 'en_US.UTF-8'}, stdin=subprocess.PIPE)
    process.communicate(output.encode('utf-8'))

andrewgross

# -*- coding: utf-8 -*-
from __future__ import unicode_literals

import base64
import json
import logging

from urlparse import parse_qs

import requests