Ask HN: CSV with Formulas I'm thinking about creating library to render and manipulte CSV. But it should have ability to use formulas, like =SUM(A:A) or =SUM(A1:A-CURRENT-1) And optionall formating maybe at the end of the cell {cell-css: {color: red, font-size: 28px}} or {row-css: {color: red, font-size: 28px}} (CSS attributes) Why? I want to create local personal databases. But a little more powerful than plain CSV. Text can be shared. Do you have proposition for syntax or architecture? |
What fascinates me is that you don't specify the order of the functions, instead the spreadsheet evaluates cells by forward chaining: it knows the graph of what depends on what. A similar system could work by backwards chaining, starting at a value it wants to compute. I'm fascinated by the computer science of that.
I made a simple model of a business in Excel that has 5 parameters and say another 5 variables, I wouldn't mind the spreadsheet evaluation over something that wasn't quite a grid but rather more like an HTML form.
See https://en.wikipedia.org/wiki/TK_Solver for a rather unique product that personifies a gridless spreadsheet.
It was a running gag back in the day that people used the wrong tool from Office for the job, Word when you should use Publisher, Visio when you should use Powerpoint, etc. Excel instead of Access was one of the most common.
An Excel-killer seems possible technically but impossible from a marketing perspective because "everybody" in the corporate office has Office 365 so Excel is free but professional who want to use this new product will need to get management to pay. Google Sheets is free, etc.
I see data scientists who think about "the April 2024 sales report" not the "X sales report" and I think a good product has to separate code and data so I can write up my analysis once and stick different data into it. There are a lot of tools that don't do that separation well, like Jupyter notebooks and Excel, that people really like.
Another take is that real business analysis is hyperdimensional. Management might care how much of SKU A got sold at store B on day C through the self-check register. See https://en.wikipedia.org/wiki/XBRL which is hated but highly adopted for regulatory reporting.
The question is about the use case and how much technical people are.
Excel format is bad because it’s not clear simple text in the raw text format. Also for simple local database with not that much data.
CSV with quirks could be edited with any text editor, put into LLM, shared, GIT history edits. That’s also the benefits of local notes in filesystem in MD format (oobsidian).
Problem with plain CSV is that I can’t do any formatting even the simplest or make simple formulas. I could make calculations using DuckDB or something but it have friction.