4sskick

The Architect’s Guide to Huffman Coding: From Information Theory to Raw Bits

In 1951, David Huffman was an MIT graduate student facing a brutal ultimatum from his professor, Robert Fano: take a grueling comprehensive final exam, or solve a problem that had stumped the founding fathers of information theory—finding the absolute most efficient binary coding scheme possible.

After months of dead ends, Huffman was on the verge of giving up to study for the exam. He gathered his scrap papers to throw them into the wastebasket when the core insight struck him: don't build the code from the top down; build it from the bottom up.

By merging the least probable symbols first, he proved that a simple greedy algorithm could achieve optimal prefix-free compression. His method bypassed the flawed top-down heuristics of Claude Shannon and Robert Fano, giving birth to a foundational pillar of modern computing.

---

4sskick

The Change Was Small. The Regression Test Wasn't.

A small backend change created a much bigger testing problem than I expected.

I added a new attribute to identify requests coming from the mobile app. The attribute allowed the intended mobile request path, but other call paths that did not include it could fail with an error.

The problem was the scope. Our endpoints were spread across multiple microservice modules, including Personnel, Claims, Attendance, Leave etc., with more than 400 endpoints across the different features. A change that looked small could affect flows far beyond the code I had touched.

My manager asked me to test every feature through its full cycle. That made sense, but manually repeating the same steps across so many features would take a long time. Each flow required opening screens, entering data, checking results, navigating back, and repeating the process for the next feature. Doing all that again for regression testing would be tedious and difficult to keep consistent.

4sskick

Stop Installing Jupyter Globally.

If you use Python + VS Code for ML, data science, or experiments, you probably have this problem:

Which Python environment is my notebook actually using?

Global Jupyter. Random pip install. Multiple virtualenvs. Broken kernels.

4sskick

React Query Showed Me the Wrong Popup — Then the Correct One

Ever had a React app do something that makes you think:

“Why the hell did this code run twice?”

That happened to me with React Query.

I had a screen that fetches a server-side rule and decides which popup to show:

4sskick

We ship node_modules by hand. So I built this.

I work on a React Native app that's old enough to remember when class components were the default. Some packages in node_modules/ have been edited by hand — not patched, not forked, literally opened in a text editor and saved. The "fixes" have been there for years. Nobody kept track of which files. And whenever we onboard a new dev, we hand them a 250 MB .rar of the entire node_modules/ folder, because running yarn install would wipe everything out.

Our setup runs like an old lady. Not in a mean way — it works, it gets to the store and back, it just shouldn't be doing what it's doing in 2026. We have a postinstall hook that mostly works on prayer. We have a Confluence page from 2021 that lists "which packages are safe to reinstall." Half the packages on that list no longer exist.

Somewhere in the last few years, 12+ packages accumulated hand-edits. Some are one-liners. Some are entire sections of code commented out or rewritten. The senior dev who

4sskick

Cherry-Pick Check Report

I ran into a situation where I needed to verify whether a feature branch (J27910_fix) had already been included in a release branch (2616).

At first glance, the history looked messy — multiple merges, unrelated commits, and overlapping changes from other tickets. This made it unclear whether my work (tagged with J27910) had already been merged or if I needed to cherry-pick it manually.

Instead of guessing or risking duplicate changes, I decided to validate it properly using Git.

---

4sskick

We have all been there: someone from management needs a list of commit IDs and change summaries for an audit, release notes, or a compliance check. Rather than manually copying 40-character hashes from a web UI into a spreadsheet, you can use native Git formatting flags to generate structured, presentation-ready reports instantly.

---

Exporting Clean CSVs Instantly

To generate a clean CSV containing the short commit hash, author, date, and commit message, run this command in your repository:

git log --pretty=format:'"%h","%an","%ad","%s"' --date=short > commit_report.csv

4sskick

🧠 Context

While maintaining a long-lived production repository, I noticed an unusually large .git/objects/pack file consuming hundreds of MBs.
This was affecting cloning speed, CI performance, and overall repository health.

---

🧪 Investigation

4sskick

Install Python 3.12 on Ubuntu 20.04 (WSL) using pyenv

This guide provides a clean and safe method to install Python 3.12 on Ubuntu 20.04 running under WSL2, without touching the system Python.

Linux distros normally have a Python version tied to the distro and used for various admin scripts. You should not expect to follow the latest releases of Python. And don't try to force change it because you could break your OS. If you need newer Python versions for your work, install it in user space.

Linux 20.04 come with python 3.8 which not supported natively to latest version of python (ex. 3.12.x), so by using pyenv will help you install other version of python without touching OS system, and it won't break your system.

---

4sskick

Detecting Changes in node_modules for Patch Package

When you edit files inside node_modules, Git does not track them --- so git status won't help detect changes.
This guide explains how to use sha1sum (or md5sum) to detect modified packages.

---

🧰 Check Compatibility