Last Week In Code 007

I recently took time off to build an app I have had in my projects folder for ages. Its a cli-first finance app. Its like, you don't open and click and wait to get things done. You type short commands, press enter to get it all done.. More on that later.

Anyways, I created an empty folder for the project and a main.py file. Since I knew what it was going to become, I wrote a skeleton for exactly that. A lot of empty functions all over. Most if not all of which weren't central to the basic task that was keeping track of my expenses.

The initial excitement started to wear off. And I started to feel impatient. How long till I use this thing. It seems I have to connect this to that then that to this then learn about that to work on this.

So I closed the project. And I was like, oh that famous unix line. Keep things simple. So I returned but this time built a simple expense calculator. It felt good I could already use it. Then I added a simple cli. Then one command, then the next. Then the next. And before I knew it, I had everything I was really using out of firefly (well that's obviously me using hyperbole but I guess you get my point).

Its a hard principle to follow. The mind seems to like to like to get the whole thing in one sitting. Oh, and about the choice of interface. The CLI.

Whenever I logged an expense with firefly, the workflow went something like:

• Open a browser
• Visit Firefly's url
• Wait for the network to return a page
• Use eyes to locate the transaction button.
• Click button
• If successfully went through all the above steps, fill in the form correctly then press 'Submit'.
• Wait for the page to re… ah..

To some that seems noraml. But if you have used the terminal for a while, you learn most of these steps is the UI making you dance around what you want to get done. Its all unneccesary. A simple command or two should get this done.

The said script's workflow:

• Open the terminal.
• Launch ledger cli with 'ledger'
• Type mv ac1 ac2 amt metadata then hit ENTER.

Boom! That it. Money moved from account one to account two of amount amt with notes metadata. Super instant. Less than one percent of the time it took me for Firefly.

Combine that with a bit of scripting and workflows that take 100 minutes are brought down to 4 or 5.

I'm sure the same gains would apply to almost everywhere where as a result of swapping GUI for CLI. A lot could be said on this topic but I guess I'll leave that for another article.

What does Git do? Version Control. But is it that obvious? It seems it is. Until it isn't.

So this site is a static site. Uploads are via pushing changes to the site's repo. It is hosted on Github's servers. I quite a while back had a problem I thought I did not know how to solve. I wanted to serve older versions of certain articles or certain versions of the site as I worked locally on others.

But then, I don't know how it happened .. kinda like a Eureka Moment it was. I was like../Wait, Version Control? Isn't that the stuff of Git./

Its crazy that its only after a long time thinking that I arrived at an aswer so obvious. Its like , "Oh, so I can use a Version Control System for Version Controlling Articles?" level crazy. I found this real interesting and had to look up the underlying phenomenon. I was sure it was happening in a whole lot of other domains, for others as well. Its the problem of transfer.

My brain seemed to have associated Git Version Control to Programming Environments. Outside a programming environment, any version control problem seemed novel. Needing novel solutions. Here's another scenario.

Recently I was watching a candidate do a mock interview. She had to solve some gambling related problem. She only had a few minutes to solve it. And she did. But its interesting that the math concept she used, a math concept I already knew, seemed so obviously applicable to the problem.. yet somehow I just couldn't make the connection. Just like I missed the connection between Version Controlling articles and Git Version Control.

A question I try to ask myself these days is how I can apply a new skill to fields outside the primary operations environment. What is the conceptual problem this tool is solving.. type of questions. Pr if faced with a problem, asking what the conceptual common problem is that can be solved with already available solutions. This happens quite a lot when writing a program. You run into some challenge that looks so familiar to some question from algorithm class.

I find I always have a lot of ideas. Some unreal. Some real. Some more elaborate than others. Some about areas which ,even if they are IT related, are way over my current proficiency level. I used to keep a note book for them. Until I stopped. I recently oppend a section I labelled rthots for (random thoughts) just for this.

I have found after some time collecting ideas and not working on them that I don't really have to turn every idea I come up with into a project.

Its like, sometimes, the ideation itself is the gain. Say I think of a project to solve some problem. Then I look up info on what would be needed to make it work and to write down my thoughts on a high level solution. I then leave it without implementing a thing and move on to another one. Any conceptual problem similar to those from previous ideas won't take up much of my time.

Eventually I'll come to a project whose core componets I have already studied about. And I still don't have to build it. But if I do, all the previous ideations will have paid off. Sounds all made up. And I can't point you to any sources to corroborate this. I usually use LLMs for that and am trying to refrain.

But I guess you have to try it to believe it.