Part 19 — Reproducible Workflow

Fundamental Series — Part 19 of 20

Analisis yang baik harus bisa direproduksi. Part ini membahas praktik-praktik yang membuat workflow Python reproducible dan terstruktur.

---

Struktur Folder Project

my_project/
├── .venv/                 # Virtual environment
├── src/                   # Kode utama
│   ├── import_data.py
│   ├── clean.py
│   └── analisis.py
├── data/
│   ├── raw/               # Data mentah (JANGAN diubah)
│   └── processed/
├── output/
│   ├── figures/
│   └── tables/
├── notebooks/             # Jupyter/Quarto notebooks
├── requirements.txt       # Package versions
├── pyproject.toml         # Project metadata
└── README.md

---

pathlib — Path yang Portable

from pathlib import Path

# Definisikan root project
PROJECT_ROOT = Path(__file__).parent
DATA_DIR = PROJECT_ROOT / "data" / "raw"
OUTPUT_DIR = PROJECT_ROOT / "output"

# Baca file
import polars as pl
df = pl.read_csv(DATA_DIR / "input.csv")

# Simpan file
df.write_csv(OUTPUT_DIR / "result.csv")

PentingJangan Pakai Path Absolut

# SALAH
df = pl.read_csv("C:/Users/Budi/project/data.csv")

# BENAR
from pathlib import Path
df = pl.read_csv(Path("data") / "input.csv")

---

Virtual Environment

# Buat virtual environment
python -m venv .venv

# Aktivasi
.venv\Scripts\activate      # Windows
source .venv/bin/activate   # Mac/Linux

# Install packages
pip install pandas polars numpy

# Lock versions
pip freeze > requirements.txt

# Orang lain:
pip install -r requirements.txt

---

Random Seed

import numpy as np
import random

# Set seed untuk reproducibility
np.random.seed(42)
random.seed(42)

np.random.choice(range(100), 5)    # [51, 92, 14, 71, 60]

np.random.seed(42)
np.random.choice(range(100), 5)    # [51, 92, 14, 71, 60] (sama!)

---

Quarto — Literate Programming

---
title: "Analisis Data"
format: html
---

## Import Data

```{python}
import polars as pl
from pathlib import Path

df = pl.read_csv(Path("data") / "input.csv")
df.head()
```

## Hasil
...

---

__name__ == "__main__" Pattern

# src/analisis.py

def proses_data(filepath):
    import polars as pl
    df = pl.read_csv(filepath)
    return df.filter(pl.col("x") > 0)

if __name__ == "__main__":
    # Hanya dijalankan jika file ini dieksekusi langsung
    result = proses_data("data/input.csv")

---

Best Practices

1. Virtual environment per project
2. Path relative (via pathlib)
3. Data mentah read-only
4. Script harus runnable dari atas ke bawah
5. Set seed untuk operasi random
6. requirements.txt atau pyproject.toml
7. Version control — gunakan Git
8. Pisahkan kode — import, cleaning, analisis, visualisasi

---

Latihan

BahayaLatihan 19.1

# 1. Buat struktur folder project:
#    data/raw/, data/processed/, output/figures/, src/
# 2. Buat src/config.py dengan:
#    PROJECT_ROOT, DATA_DIR, OUTPUT_DIR pakai pathlib
# 3. Buat src/analisis.py yang import config dan baca data

---

Ringkasan

Praktik	Tool
Portable paths	pathlib
Package management	venv + requirements.txt
Reproducible random	np.random.seed()
Literate programming	Quarto / Jupyter
Script entrypoint	if __name__ == "__main__"
Version control	Git

---

Sebelumnya: Part 18 — Debugging & Error Handling Selanjutnya: Part 20 — Mini Project