Hacker News new | past | comments | ask | show | jobs | submit login
Ask HN: Would Spreadsheets with Lisps be much more powerful?
68 points by sarvasvkulpati 6 days ago | hide | past | favorite | 64 comments
I'm a CS freshman who got obsessed with lisps after using Scheme in my intro class.

Tangentially, as a side project, I made a spreadsheet to learn how their update mechanism worked[0]. For the formula language, I implemented a basic lisp that does math expressions.

That got me wondering- would making a fully functional lisp the formula language add substantial power to spreadsheets? Of course, macros add power to anything, but I'm wondering if there's specifically new interactions that would be possible with sheets using a lisp as opposed to, say, JS.

An initial idea was being able to call (apply A1 B1) to apply a procedure from one cell to another, but that's not the most groundbreaking concept. Hence the question to those of you who understand lisps more than I do.

p.s: Not a pro. Go easy on me HN!

[0]: https://lispread.sarv.repl.co/






Dear sarvasvkulpati,

I appreciate your question! It demonstrates a curiosity about the nature of computing itself. Some will surely say that this idea is unreasonable or not practical. But, if this is an idea that speaks to you, I encourage you to try your hand at making a prototype of your idea! This is the hacker way. And being in college you have time, peers who are equally enthusiastic, energy, and the help of professors. Regardless of the practicality, all interesting progress comes from people who ask questions like this, and see what comes of trying to make something interesting.

In college I had a peer who convinced us to all use Gentoo Linux because it was the most hardcore Linux. I bricked my computer three times trying to use it and ultimately switched away from using Gentoo but learned a ton about Linux that I use to this day on a less hardcore distro. I still use Vim even though it's more complicated than other editors. I also went through a period where I tried learning every programming language I could find to see which one was "the best". I ultimately found a few I liked the most. None of these were efficient at first, but eventually led to highly cool results.

Ask for help if you get stuck. I bet something interesting will come of it, even if you just learn what doesn't work.

Something something "The reasonable man adapts himself to the world: the unreasonable one persists in trying to adapt the world to himself. Therefore all progress depends on the unreasonable man."

Your idea is really cool. Stay hungry and awesome. Cheers to you.


Couldn't agree more with you :) Staying Hungry and Passionate in terms of learning can take you really far

You may find SIAG (Scheme in a grid) interesting. It explores some of the same ideas: https://siag.nu/siag/

There's also been some work along these lines in Excel itself, although using the Excel formula language rather than Scheme. Excel now has both let and lambda, as well as a way to work with composite data structures in cells.

https://www.microsoft.com/en-us/research/blog/lambda-the-ult...

https://support.microsoft.com/en-us/office/let-function-3484...

https://exceljet.net/dynamic-array-formulas-in-excel

In general, I think your idea is sound. Almost since the beginning, Spreadsheets have had two major ways of operating on data - the formula language, which is automatic, declarative, and functional in nature, and commands (like on a menu), which are generally imperative and have to be explicitly invoked (or invoked through a separate macro language). Both have their places, but the functional language of formulas is nice for the same reason functional languages in general are appealing. There's also an easier ramp for users new to programming to progress from "=A1+B1" to more advanced sorts of operations (lookupus, etc.) without having to switch to a formal "programming language".

However, aside from additional standard functions, the formula language half (in Excel, at least) was largely static since the 1980s - limited essentially to scalar data types, and no internal way for creating abstractions. This severely limited the range of what you could do with formulas, when compared to more complete programming languages, and forced a lot of functionality to be done in (more brittle) macros.

So the direction you're going - broadening the power of what can be done with a spreadsheet language - is a useful one, and there's prior art to justify that the idea makes sense.


Excel has a new function called LAMBDA, which makes Excel Turing-Complete. (https://www.microsoft.com/en-us/research/blog/lambda-the-ult...) If you haven’t gotten there yet in your CS classes, Turing-Complete means you can compute anything that is computable, so provably as powerful as any other language.

Turing-Complete doesn’t say anything about convenience though, there are always reasons to play with embedding your favorite language into your other environments, so don’t let prior work stop you from playing and exploring! You will earn yourself many benefits by trying to do it yourself.

Excel, BTW, was fairly powerful as a programming environment even before the LAMBDA function, there have been multiple examples of crazy applications written in an Excel spreadsheet, video playback, games, ray tracing, etc. Here’s a ray tracing example: https://news.ycombinator.com/item?id=20764825


Excel’s lambda is still only avaiable in beta versions, I think? At least, I haven’t seen it yet, and I have been looking.

Is it theoretically possible to build a LISP-style language on top of Excel's LAMBDA function like Clojure did on top of Java/JavaScript and Hy on top of Python? Might prove to be an interesting venue...

> Turing-Complete doesn’t say anything about convenience though,

e.g. c++ template mechanism, also Turing complete.


CSS is also turing complete (given some assumptions).

There’s a nice list here if surprisingly Turing Complete stuff: https://www.gwern.net/Turing-complete

Good question! Smartass answer: No, because spreadsheets in programs like Excel are Turing-complete, which means any program you could write in a Lispy spreadsheet, you can write in Excel.

Useful answer: In practical terms, probably, for a very small subset of users. The kinds of users who are into programming and also Lisp, and also use spreadsheets a lot. For these four users, adding their favourite programming paradigm would help them to implement cooler things than they would otherwise do.


As someone who is into programming and lisps, I would prefer a declarative language (a la SQL and the relational model) which is just simply superior at manipulating data over some FP language (OOP langs are much worse of course). But then, many already use SQL instead of spreadsheets.

If the native language is Lisp, you could use it to build a declarative DSL pretty easily. If the native language is something like SQL, your options are much more limited.

Right on! I was thinking of suggesting something like minikanren. My likes can get a bit niche at this point so I’ll leave it at that.

> Excel's functions are turing-complete, and therefore equivalent to lisp

That is a smartass answer. Technically correct as long as you ignore the human factor.


> Smartass answer: No, because spreadsheets in programs like Excel are Turing-complete

There is a difference in expressivity between programming languages despite the fact that they may both be Turing-complete. Just look at how languages add more language features over time...


looks at C++

As far as I know, Excel functions have never had side effects, they’re pure functions. It's already lisp, just with the open paren in a different position, and commas that could be less meaningful.

About 40 years ago —

"An advanced object-oriented spreadsheet package for the Smalltalk-80 system

Asp™ is an analytic spreadsheet package that provides access to the full power of the Smalltalk-80™ programming environment. Developed by Xerox Special Information Systems (XSIS) as part of The Analyst™ Workstation System, ASP is now available as a separate package for the Smalltalk-80 environment.

Unlike other spreadsheet packages, ASP is designed for the analysis of more than just numbers and labels. Vectors, points, collections of data, text, charts, files, and images are just a few types of objects that ASP can manipulate and display.

While ASP provides a large array of predefined functions, it is also extensible. By adding new methods to the Smalltalk-80 environment, ASP can be instantly expanded to match user needs, without requiring extensive knowledge of either ASP or the Smalltalk-80 system."

http://www.bitsavers.org/pdf/xerox/xsis/XSIS_Smalltalk_Produ...

"Object-oriented spreadsheets: the analytic spreadsheet package"

https://dl.acm.org/doi/abs/10.1145/960112.28737


I haven’t used it myself, but I believe Simple Emacs Spreadsheets lets you write formulas in emacs lisp.

https://www.gnu.org/software/emacs/manual/html_node/ses/inde...


Org-mode allows it as well.

https://orgmode.org/worg/org-tutorials/org-spreadsheet-lisp-...

I've only dabbled with it, I am guessing for anything really substantial it would be quite slow compared to e.g. Excel.


It does. And while I like SES, it doesn't scale terribly well (in contrast to proper spreadsheet software). I did have a nice little stock dashboard using Yahoo!'s API about a decade ago paired with my purchases and such. I also made a nice little tax calculator and a few other things, I found it very useful (and used emacs lisp) for those kinds of programs. But the kind of scale that I've often seen Excel used for in businesses were a poor fit for it. Plus, good luck convincing non-technical managers (here I mean all kinds including financial managers and such) to use Emacs as a daily use tool. It's happened before, but it's not an easy sell.

Perhaps even more powerful would be spreadsheets + Prolog. Since Prolog's main strength is finding answers for you, I imagine putting that on grid would have very interesting results.

That sounds like a more powerful variation of Excel's Goal Seek function. The latter however uses an iterative approximation approach. Given the non-exact nature of floating point, it would be interesting to see if numeric approximation can be combined with a Prolog-like resolution algorithm.

Yes indeed! Successful (caveat that term appropriately ;) versions of this prolog+floats idea have used interval analysis in the past. Anyway, yep you can do it, and it is fun!

Spreadsheets + datalog has been a concept floating around for a while. You can find papers about NEXCEL, XcelLog, and probably some other attempts.

It’s a tricky problem.


Once you have a lisp, implementing microKanren is straight-forward. Just as the lisp and you get the Prolog for free!

See Kenny Tilton's Cells project, which more or less does the inverse.

https://github.com/kennytilton/cells


Cells is really neat. I last used it 10+ years ago but it was a delightful way to express ideas about market structure in a way that made sense to my excel-addled brain.

i've sometimes thought that APL could mesh really well with spreadsheets.

https://www.youtube.com/watch?v=v7Mt0GYHU9A https://www.youtube.com/watch?v=UltnvW83_CQ


The problem with regular programming is that you can see the code but not the data, and the problem with spreadsheets is that you can see the data but not the code.

You might be on to something with the “apply” idea. What if there was a split screen where you could see both code/operations and data and could mix them together with all the transparency of a spreadsheet? macros would just exploit code as data?

For a tangent that might spark interesting ideas, check out this structure editor project in Racket (a lisp) https://github.com/disconcision/fructure


One advantage of using lisp is it would make it easier to develop complex routines. You could even import external code, so you could write/debug some hairy algos externally and then use them in the spreadsheet. Imagine having the power of macsyma and doing pdes in your spreadsheet. Or being able to do advanced vector operations on columns.

Though I am not as fond of them a stack language like forth or postscript could also be the computing paradigm of your spreadsheet.

Or if you are feeling more perverse, there’s always TECO


I'm building a language (https://tablam.org) that, hopefully, could become the base for excel/access alternative.

lisp is not the better fir for excel, to see why, check this:

    "The memory models that underlie programming languages"
http://canonical.org/~kragen/memory-models/

Among them, array/relational could be a better fit (semantic and operational). Of course you also need to add some form of reactivity (or actors?) to mimic the immediacy of excel formulas.

---

To confirm this intuition, check how much pandas/numpy/R lang/apl/Julia/Sql people intersect with the usage of excel.

This means that languages/tools are closer in spirit than lisp.

Also, syntax wise, concatenative langs/array langs/forth langs/sql like is more nice for the terseness of excel usage.


Honestly, Pandas and R are actual Excel killers. At least in certain areas.

For a syntax in a spreadsheet you would want something that is short and familiar but you can certainly toy with idea.

I remember when the geeky engineers had their calculators in reverse Polish notation:

https://en.m.wikipedia.org/wiki/Reverse_Polish_notation


checks my top drawer What do you mean 'remember when'?! :P Now if only I could remember how to use the damn thing...

Numbers, enter, numbers, operation. That's about it, nice and simple. Programming in RPL was a lot of fun, sadly my HP-48G was lost to water damage (lost a bunch of things from that leaky roof). I still have my 32SII from college when a class required non-programmable calculators. Of course, it is programmable but they didn't know that (and I didn't use it) but it was easier to slip that out and stick with RPN than get used to a TI scientific calculator again.

I unironically have an RPN calculator app on my phone. I find it easier to use than a regular calculator.

Yes, just as "sharks with frikin lasers" are much more powerful.

But seriously, no. It is cult cargo thinking" to believe that merely juxtaposing two things will lead to something "more powerful. OTOH it is often the first step toward a new and useful idea. So all the magic is in the "how" of combining things.


Look, yes and no: I want to ignore Turing completeness and the tarpit, forget about lambda, and talk about intermediate results. I write excel so as to write out intermediates, so to calculate f(n) = f(n-1)+.. I’d write N lines. If you could just write a function to “skip to the end”, there’d be value in seeing all the values in between, in some sort of debugger view / virtual sheet. If that makes sense.

If I define (square n) and (twice f) to calculate a quartic function, what would I see in a single cell, when I’m not looking at the formula view? That’s what I wonder.


More powerful? Yeah, sure. But honestly, many companies have made more powerful versions of Excel over the years and most of them have gone to the wall.

If you think of power in a different way, that of empowering users who typically have little or no technology experience, things get a bit more interesting. Anything you come up with has to be instantly obviously more useful in at least one way. Google Docs and Zoho are still viable products despite from a CS perspective being significantly less powerful.

Like many answers you’ve received, it comes down to the difference between technology, product design and commercial concerns.

The only product to ever supplant a world-wide dominant spreadsheet is... Excel.


Given you have a finite amount of time, the relatively obvious answer is - this isn't a good idea.

Excel is made by Microsoft. Microsoft is strong in languages and tooling around languages. If there was something genuinely useful to add, they'd have done it.

Now, obviously you can't think like this about everything or you'd never do anything. But, you have finite time so pick something better, I'd say. Maybe pick an area where you have some deep insight. Or some area where the existing products stink. Excel is really freaking good given the large number of use cases they have to support.


I was at OPLSS[0] in 2018 and Enzo Alda gave a talk/demo on ZenSheets, his in-progress reactive spreadsheet thing where you could write lambdas in the cells, call them in formulas from other cells, etc. That talk wasn't recorded, but a similar demo can be found here.[1]

[0]: https://www.cs.uoregon.edu/research/summerschool/summer18

[1]: https://www.youtube.com/watch?v=mJa0_gKE6xo


Every time someone mentions extended spreadsheets I like to bring up Resolver One. In the mid 00s it was a spreadsheet interface in .Net that generated python modules under the hood running on IronPython. As a side effect you could import sections of one spreadsheet into another spreadsheet, go behind the scenes and edit the Python to connect to a DB or run arbitrary code, generate GUIs using the .Net facilities, etc. I thought it was neat. Unfortunately they never got traction and as far as I know they were aquihired by a financial company that used it internally.

If you're interested in this topic, you should look into emacs (a text editor that uses a Lisp-like language for customization) and it's accompanying org-mode (something that helps you take notes in emacs, and also has a built in spreadsheet), which allows for using emacs lisp for defining functions: https://orgmode.org/worg/org-tutorials/org-spreadsheet-lisp-...

Good luck!


Of course non-pure Lisps and JavaScript pretty much only differ in syntax, so there is no "new interaction possible".

They are more powerful than the basic formula languages in most spreadsheet software, which is why spreadsheet software usually offers the ability to write custom functions using traditional programming languages.


Yes. People who know how to program in matlab and python, but also have to manually open up excel / csv files need this. Loading up excel is slow, writing a python script where you have to search the docs to find the functions you want is slow too. A programmable and fast spreadsheet is needed.

Interesting idea. I like spreadsheets and a new way to program them would be interesting.

In your sample it seems like problematic entries cause the sheet to crash or become non-responsive. Probably a good idea to handle errors gracefully. If you can't evaluate, just give an error message.



You might find something useful here: https://people.eecs.berkeley.edu/~bh/pdf/ssch24.pdf

The answer is quite likely to be yes. But then when you consider who is the primary typical user of a spreadsheet and ask yourself "are they likely to learn Lisp?" then the answer is no.

Excel formulas are already functional. They’re just composed with Sexpressions. I’m not convinced semantics of one is any better than the semantics of another. Ultimately that boils down to personal preference.

Sarvasv,

You should explore this further! Once you are past the basic syntax, I think there are world of possibilities that can be explored.


It would be cool if each cell is an s expression on its own. - but are cells the best way to show s expressions?

from the other side, check out 'cells' for an interesting implementation of spreadsheet-like reactive updates within common lisp https://github.com/kennytilton/cells/wiki

Is lisp more powerful than JS? Not really; you can write lisp in JS. But that's a correct answer to a wrong question. Lisp is fun, but IMO it's wrapped in undeserved mythology. Maybe I'm just jaded, but if you move parens one word to the right and sprinkle commas between tokens, it's downright readable to a c programmer. That said, homoiconicity is powerful and spreadsheet-based metaprogramming could be super cool.

What I haven't seen is a maturation of spreadsheet-based programming. Code and data are mingled in cells, and excel (and clones I've used) hide the code and display the data in an opinionated manner. That's a huge UI problem, IMO. But, mingling code and data can be very fun! Perhaps lisp could be cool for that, but I'd actually prefer a concatenative array language to really take advantage of the spreadsheet UI. So, go read up on Forth and APL, and bake their unholy offspring into a spreadsheet. Nothing bad can come of this.


Well, I don't think "more powerful" is a good way to think about it. A spreadsheet is essentially a functional programming paradigm. One programming thingie is not going to be "more powerful" than another. That said, one could be cooler or more interesting than another one, at least to you.

My feeling is that it doesn't have to make sense, and you don't really need to have any justification. If you want to make a spreadsheet system with Lisp, I'd definitely do it. You learn the most when you do your own projects, and interest is the fuel.

Also, we're bad at determining beforehand what actually is important, so maybe you work on something as a toy, but later it (or some piece of knowledge developed while working on it) is crucial for something else later.


When people speak of "power" in a programming language, they generally mean expressive power. It's true that all (Turing-complete) programming languages can theoretically encode the same programs - but they do not all take the same time to implement. Lisp is more powerful than C; C is more powerful than x86 assembly; x86 assembly is more powerful than Brainfuck.

It's reasonable to ask if a given approach to language design results in more "power". It's not just a question of taste.


One interesting thing with lisp and spreadsheets would be that a you can transform data in the sheet into a program with a macro, so rather than just values one cell could turn other parts of the workbook into a dsl. Could be some interesting shenanigans you can get up to.

Thank you! This makes a lot of sense!

If I can run any arbitrary lisp program in a cell (assuming the env is global), I was wondering if this was just a 2d text editor for a lisp. But that's awfully reductionist.

But you're right I'm assuming I'll discover interesting things if I keep tinkering with it!


Mark your post as "Show HN". There's no way I will comment on your side project after I was tricked into reading about it.

From the HN guidelines: "Be kind. Don't be snarky. Have curious conversation; don't cross-examine. Please don't fulminate. Please don't sneer, including at the rest of the community."

Apologies if you felt tricked into reading it.

I only spoke about the project as context for the question in the title, but I see how I could have been clearer, and how I didn't necessarily need to include the title.

If you're experienced in Lisp, I'm still curious about the question: is adding a lisp to a sheet more powerful than adding, say JS, and if so, how?


If you're experienced in Lisp, I'm still curious about the question: is adding a lisp to a sheet more powerful than adding, say JS, and if so, how?

No, it won't be more powerful


Harsh



Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: