5 ms·

There’s something unbeatable about the purity of a spreadsheet. For certain kinds of problem, it’s one of the most understandable ways to communicate a program.

There’s something unbeatable about the purity of a spreadsheet. For certain kinds of problem, it’s one of the most understandable ways to communicate a program. But there are the obvious problems, as formulas get inscrutable and you really want some more powerful data types. So I’ve been playing with the new Excel features - lambdas and the new kind of array formulas. And they’re kind of great! I ported some non-trivial analysis algorithms from numpy to excel and it makes for a highly shareable and havkable programming environment for non-coders. There’s all sorts of crazy excel warts (I’m doing maths with complex numbers, and the handling of those is a true “WAT”) It’s kind of almost-great. I can’t put my finger on it but I feel like it’s close to a really winning programming environment for certain kinds of algorithms-transforming-data programs. I think Excel probably has too much baggage to get there, but these experiments are still really interesting.

nhatcher 3y ago
Are some of those Excel workbooks with non-trivial analysis and complex numbers something you could share? I'm truly curious.
- heisenzombie 3y ago
  It's annoying for me to share the spreadsheet, but I'm basically implementing the math in: https://lhc-div-mms.web.cern.ch/tests/MAG/FiDeL/Documentation/lhc-m-es-0001-30-00.pdf https://lhc-div-mms.web.cern.ch/tests/MAG/FiDeL/Documentatio... To do with calculating multipole coefficients from measured data and then doing coordinate transformations on them.
  - nhatcher 3y ago
    Oh wow! Many thanks! It's impressive what you can do on a spreadsheet these days.
bdavbdav 3y ago
Coming from finance where we've been trying to wean FO off excel, every time I see things like this, I'm reminded why we shouldn't get comfy relying so heavily on Excel: https://www.businessinsider.com/excel-partly-to-blame-for-trading-loss-2013-2?r=US&IR=T https://www.businessinsider.com/excel-partly-to-blame-for-tr...
- wslh 3y ago
  It is a grey area, spreadsheets are great but at some point you are building a mammoth with it. I also find Excel, and Google Sheets outdated on its own "form-factor". It would be great to hear from the HN community about ways to improve spreadsheets. For example, adding new types and easier integrations with other systems. Also having views/forms. Lotus Notes was revolutionary in this aspect. Excel is the measure because they have a good engine. Probably we need to create a great DAG engine and put the spreadsheet over it.
  - bradrn 3y ago
    > It would be great to hear from the HN community about ways to improve spreadsheets. I’m working on this! Very early days so far, and I have nothing much to show for it yet, but I do have some ideas on integrating static type inference and proper data structures into spreadsheets. My initial prototype is at [0]: it’s horrifically buggy in just about every way (don’t even try to get it working!), but the screenshot should give some idea of what I’m thinking should be possible. [0] https://github.com/bradrn/nexo https://github.com/bradrn/nexo
  - operator-name 3y ago
    There's https://pyspread.gitlab.io/ https://pyspread.gitlab.io/, which uses python. On the more programmable side you've got Jupyter with https://github.com/ipyflow/ipyflow https://github.com/ipyflow/ipyflow and https://github.com/fonsp/Pluto.jl https://github.com/fonsp/Pluto.jl.
    - pcthrowaway 3y ago
      Grist looks similar, and also interesting: https://www.getgrist.com/ https://www.getgrist.com/ The tables are sqlite tables, so the schema has to be rigid, while I think they also offer views that allow you to make more user-friendly views. They also let you run Python inside the formula

nhatcher 3y ago

Are some of those Excel workbooks with non-trivial analysis and complex numbers something you could share? I'm truly curious.

heisenzombie 3y ago
It's annoying for me to share the spreadsheet, but I'm basically implementing the math in: https://lhc-div-mms.web.cern.ch/tests/MAG/FiDeL/Documentation/lhc-m-es-0001-30-00.pdf https://lhc-div-mms.web.cern.ch/tests/MAG/FiDeL/Documentatio... To do with calculating multipole coefficients from measured data and then doing coordinate transformations on them.
- nhatcher 3y ago
  Oh wow! Many thanks! It's impressive what you can do on a spreadsheet these days.

bdavbdav 3y ago

Coming from finance where we've been trying to wean FO off excel, every time I see things like this, I'm reminded why we shouldn't get comfy relying so heavily on Excel: https://www.businessinsider.com/excel-partly-to-blame-for-trading-loss-2013-2?r=US&IR=T https://www.businessinsider.com/excel-partly-to-blame-for-tr...

wslh 3y ago
It is a grey area, spreadsheets are great but at some point you are building a mammoth with it. I also find Excel, and Google Sheets outdated on its own "form-factor". It would be great to hear from the HN community about ways to improve spreadsheets. For example, adding new types and easier integrations with other systems. Also having views/forms. Lotus Notes was revolutionary in this aspect. Excel is the measure because they have a good engine. Probably we need to create a great DAG engine and put the spreadsheet over it.
- bradrn 3y ago
  > It would be great to hear from the HN community about ways to improve spreadsheets. I’m working on this! Very early days so far, and I have nothing much to show for it yet, but I do have some ideas on integrating static type inference and proper data structures into spreadsheets. My initial prototype is at [0]: it’s horrifically buggy in just about every way (don’t even try to get it working!), but the screenshot should give some idea of what I’m thinking should be possible. [0] https://github.com/bradrn/nexo https://github.com/bradrn/nexo
- operator-name 3y ago
  There's https://pyspread.gitlab.io/ https://pyspread.gitlab.io/, which uses python. On the more programmable side you've got Jupyter with https://github.com/ipyflow/ipyflow https://github.com/ipyflow/ipyflow and https://github.com/fonsp/Pluto.jl https://github.com/fonsp/Pluto.jl.
  - pcthrowaway 3y ago
    Grist looks similar, and also interesting: https://www.getgrist.com/ https://www.getgrist.com/ The tables are sqlite tables, so the schema has to be rigid, while I think they also offer views that allow you to make more user-friendly views. They also let you run Python inside the formula