Python on AW — notes & thoughts

A (reasonably) complex radar chart

Fri, 11 Mar 2022 21:28:43 -0500

A radar (or spider) chart is a convenient method of displaying multivariate data in a form of 2-dimensional chart.

When I say convenient, I mean it is particularly well-accepted by people and gives a good intuition when comparing different items/subjects.

Think about comparing 2 (or more) football players in terms of dribbling, running speed, shots and pressure. Even with only four variables bar plot would become cumbersome to deal with and not immediate to perceive. That’s why they are particularly popular in sports analytics.

Google 20 for 1 stock split

Wed, 02 Feb 2022 21:28:43 -0500

Google announced Tuesday its results for Q4 2021 and that they plan to split shares at 20 for 1 this Tuesday. Basically, if you are a shareholder, for each 1 share of Google at (about) 3000$ dollars you will have (at the ex-date) 20 shares at 150$.

Nothing should really change, but there might be some financial effects, for instance, due to the inclusion of the stock in new indices or increased market liquidity.

Short notes on namedtuples, NamedTuples and Data classes

Mon, 11 Jan 2021 21:28:43 -0500

Both dataclasses and tuples are based on the attrs project, the one Python Library everybody needs and are fast object types (factory methods) designed to simplify and reduce code.

namedtuple and NamedTuple

Both immutable
Tuple-based (hence fast) but far better than tuples
NamedTuple is the typed version of namedtuple
immutable, iterable, hashable, unpackable
backward-compatible with tuple (e.g you can access a namedtuple by index)
default arguments supported from Python 3.7+
fast! C-based
example ¹

Point = namedtuple('Point', 'x y')
pt1 = Point(1.0, 5.0)
pt2 = Point(2.5, 1.5)

from math import sqrt
# use index referencing
line_length = sqrt((pt1[0]-pt2[0])**2 + (pt1[1]-pt2[1])**2)
 # use tuple unpacking
x1, y1 = pt1

Dataclasses

Are mutable
post-init processing can be used to create fields depending on other fields or even to perform input validation²
all implementation is written in Python, so slower wrt tuple-based data types
Validation of types at runtime not supported natively (or cumbersome) but easy with decorator @enforce_typing
from Python 3.7
are just regular Classes (e.g. inheritance) withot writing boilerplate code
inappropriate when API compatibility with tuples or dicts is requested
bonus PyCon talk, if you have time
example³

from dataclasses import dataclass

@dataclass(unsafe_hash=True)
class InventoryItem:
 '''Class for keeping track of an item in inventory.'''
 name: str
 unit_price: float
 quantity_on_hand: int = 0

 def total_cost(self) -> float:
 return self.unit_price * self.quantity_on_hand

Other fancy data types (not from the stdlib)

pydantic: enforces type hints at runtime providing user-readable errors when data is invalid. Seems relatively popular.

# sample input val via Regexp
from dataclasses import dataclass
import re
@dataclass
class Widget:
 id: int
 def __post_init__(self):
 id_condition = re.match(r"[0-9]{4}", str(self.id))
 if not id_condition:
 print(f"{self.id} doesn't follow pattern [0-9]{{4}}")
 raise CustomException

brief explanation on stackoverlow ↩︎

Add a row to pandas' DataFrame

Sat, 31 Oct 2020 21:28:43 -0500

Quite often, it is needed to fill or modify dataframes with data that gets computed at runtime. This post draws heavily from the stackoverlow question “Add one row to pandas DataFrame”. Here I show the 4 methods that were proposed to append the data and discuss them. A 5th “improper” method operating on columns is added just for sake of comparison with respect to the most efficient option.

If you want just the takeaways: