This tracks: years ago I saw a CS report on the prevalence of programming bugs in Excel spreadsheets that found about the same number of bugs per LoC-equivalent as in Visual BASIC written by low-proficiency programmers. Spreadsheets encourage spaghetti code and opaque variable naming: meanwhile spreadsheet debugging tools are stuck in the stone age (i.e. 1970s). The defect rate goes down in organizations where spreadsheets are developed by the IT department, not MBAs …
https://paquita.masto.host/@rinze/112959459848001627
@cstross Something very sad happened in the '90s.
Lotus had two spreadsheet products. The most popular, 1-2-3, was a clone of VisiCalc. It used a flat table metaphor because they thought that's what accountants (who were used to tables of data) wanted.
In 1991, they released Improv, which had a far superior model. Improv had a clean separation of tables and formulae. Rather than copying a formula over a column with magic sigils to indicate which cell references would update, you wrote a formula that took rows and columns and created new rows and columns. Accountants loved it because it freed them from the limitations of paper, rather than replicating them in computers.
Sadly, the market dominance of 1-2-3 gave way to Microsoft Excel, which was marketed agressively. Improv had no corresponding marketing / monopoly power behind it and so died. Everyone since then was forced to learn a model for data that was based on a paper abstraction that no one even remembers.
Quantrix is the only survivor of the Improv lineage (there's an open-source Improv clone, but it's not been actively developed for 10-15 years).
The error rates in Improv-style spreadsheets are vastly lower than in VisiCalc-style spreadsheets.
@david_chisnall @cstross The approach we were taking at Arithmix was a step up from Improv and much more flexible and easier to use than Quantrix. Sadly we couldn’t get funding to finish it. This linkedin video is all that’s left of it and only shows a fraction of what the thing could do. https://www.linkedin.com/posts/arithmix_arithmix-demo-activity-7033849515711287296-pwVC?utm_source=share&utm_medium=member_desktop
@bjn @david_chisnall @cstross From 2'12 onwards, your demo looks a lot like Grist. Do you know Grist? https://www.getgrist.com
@david_chisnall @enroweb @cstross Our thing was multi-dimensional throughout, with a few twists on top. Grist just seems to be a friendly layer on top of an SQL database. I’d have to look at the open source code to get a better idea. Dimensional modelling seems to be a hard thing for people to get their heads around to start with, but once they see how vlookups or joins all get automated away, there’s an ‘ah-ha’ moment. Needs more upfront thinking than a spreadsheet, but less than a database.
@david_chisnall @bjn @cstross Grist seems to be built along these lines, but I am not techie enough to be sure. Also, it accepts Python *and* Excel formula, making it low threshold and high ceiling.
