I think it would be great to do the reverse - take existing programs and map out the calls in a graph structure that can be visually displayed.

Hardest part of looking at existing code base is to "mentally map" the call stacks and overtime, atleast for the way my brain operates, I tend to build a 2d map in my mind. It doesn't have concrete positions for blocks (functions), but there is a sense of structure in the brain that persists. I can come back to the code base and it is evoked automatically.

We need visual tools for examining codebases just like we have tools for examining database schemas. Perhaps its not that simple though.

Having worked in actual visual programming environments I'd argue that even the most trivial applications are too complicated to mapped out visually. There just isn't enough visual resolution in a screen, or for that matter, in your eye to render it sufficiently. Zoom out and lose detail. Zoom in and use lose context. I've seen programs printed out into huge graphs covering entire boardroom tables and I can assure you nobody benefited from the display.

Ironically the convoluted programs that people think this would be a solution for make for the most disgusting and hard to follow diagrams.

I found the reverse. I had to do a tech debt cleanup of the code and found that doing an abstract diagram made it much easier to map out the interrelation between all the classes which was hard to keep in my head all at once. Also made it easier to communicate to others visually why it was spaghetti & easy to create a diagram about how I'd like to clean it up. I used PlantUML because what a lot of text -> visual things miss is that they don't provide the necessary knobs to tune the level of detail so doing the diagram by hand was a necessity.

There may be multiple ways to look at something - sequence graphs, data flows - and additionally you don't want to capture all the details in the visualization, you want to be able to cut-off the diagrams at choke points where they're not helpful (e.g. I don't care about package X in the code) which is what all visualization packages try to do. This is directly analogous to geographic maps.

Imagine if you viewed the map for a state/province whatever & had the local road network detailed. That map would be unusable & discarded quickly. Diagramming tools need to be able to take text & generate a visualization that you can zoom into/out of & lose/gain the appropriate level of detail necessary, be able to switch between different aspects (like you can with street view vs regular vs satellite, vs topography). I think the reason we don't have equivalent tools for programming is three-fold:

1. There's a lot of popular languages. Maintaining support for parsing & mapping that to the visualization abstraction is going to be a lot of work. Probably manageable though.

2. As engineers we tend to think in either solving the problem or not. No visualization is going to be good as a hand-drawn diagram in the moment nor is it probably good enough (at least the first version) as a vehicle for making the modification.

3. There's lot of type erasure/dynamism that's impossible to capture in code analysis (e.g. type-erased callbacks like `std::function`, `std::any`). That makes the visualization incorrect or cause it to fail prematurely.

The whole thing maybe.

But often I'm in a function and I'd like to know "who calls this", or "what does this call". So being able to get a graph limited to the current function (node) and a couple of levels out on either end could be very helpful. Or maybe being able to mark two functions and see the ways they are connected (if at all).

I guess there might be some tools that do this though, haven't looked.

The model of a program as a linear file of 80-column text is a historical accident based on IBM mainframes in the 1950s, and the decks of punch cards they read. It seems like there must be a way to treat text in a more flexible way without going to a full visual programming language.

For example, consider all the things that have a physical order and position in a file that is actually irrelevant. Why can't methods (for instance) get displayed as convenient, without worrying about where they are in the file?

Limiting programs to fixed-width, single-font text also seems like a historical artifact. (The Xerox Alto allowed full WYSIWYG text for program code.) Having to draw diagrams with ASCII graphics just seems wrong.

Admittedly, these examples are somewhat trivial, but my point is that program structure is probably a local maximum, not the best possible solution.

Why can't methods (for instance) get displayed as convenient, without worrying about where they are in the file?

Get displayed as convenient what?

Limiting programs to fixed-width, single-font text also seems like a historical artifact. (The Xerox Alto allowed full WYSIWYG text for program code.) Having to draw diagrams with ASCII graphics just seems wrong.

I can imagine benefits of better diagrams, for documentation, but how would more fonts in the code help?

It seems like there must be a way to treat text in a more flexible way without going to a full visual programming language.

I've mentioned this a few times on past articles on HN, but I think the issue is we're marrying two different concepts with text based programming languages, namely presentation and representation. Consider - in an idealized context - HTML and CSS. One being the representation, the other being the presentation or more accurately how to present the representation. A user can change how the HTML is presented to them without changing the HTML directly. I think something like this would be useful for programming source code.

So as a thought experiment for this in terms of programming, the source code would be some backing representation that you edit which can be rendered in any way a user desires, including back to regular text as we have now. Tooling (compilers/interpreters, version control, editors, etc) would deal primarily with the representation rather than the presentation.

I think this would also solve a lot of problems that tooling ends up dealing with. Consider version control not having to deal with formatting changes because that's just an artifact of how a programmer chooses to render the representation rather than the representation itself. Autoformatters (gofmt, clang-format, etc) would be less about putting source code in a canonical format for the sake of tooling and more about rendering source code as text in a particular chosen way for individual programmers. Similar for tooling such as IDE code folding or refactoring, document generation, etc.

This is actually something I have been pondering for a decade at this point. I have no idea how a system such as this could be implemented such that it is general and sound enough to be usable. But I still think it would be a neat way to program.

Why can't methods (for instance) get displayed as convenient, without worrying about where they are in the file?

This is something unison is playing with with their codebase manager.

https://www.youtube.com/watch?v=gCWtkvDQ2ZI

Lack of evidence of their worth. There are not many Visual Languages that would be generally agreed are ‘‘successful,’’ and there is little in the way of formal experiments or informal experience that shows that Visual Languages are good.

This is bullshit. A lot of musicians & visual artists are super successful with tools like Max/MSP, PureData, vvvv, or the ton of graph-based shader editors out there. Max/MSP is taught in a lot of place to music students, e.g. in conservatories. I promise that by the time you get them to understand for() and if() they've already made some music in Max, and the better ones have already started toying with sensors and controllers. Disclaimer: I worked 3.something years in a company where the owner routinely shipped software entirely built in Max.

People without any programming abilities are able to use Blender's node editor to make super nice visuals : https://www.youtube.com/watch?v=JhLVzcCl1ug and at the other end of the spectrum, code that runs in airplanes is routinely designed using SCADE (https://www.ansys.com/products/embedded-software/ansys-scade... - see e.g. these slides which talk about its use in the A380: http://www.artist-embedded.org/docs/Events/2008/Autrans/SLID...) and other tools based in the synchronous reactive model.

I can write programs in LabVIEW in a few hours that would take me weeks in another language. Not all of that is due to the graphical nature, most probably due to the domain specificity and IDE integration. LabVIEW makes it super easy to create GUIs, visualise data and debug whilst running.

Game dev frameworks Unity and Unreal also have good coding support for things like developing shaders, AI behavior trees, or just regular game logic.

There are benefits to visual coding such as being able to visually group or cluster related code, being able to use zoom in and zoom out to create 'region's of code visible from further out with large labels, and being able to follow the lines connecting your code and state visually.

But also there are big big big drawbacks, I don't need to mention the literal visual spaghetti code or the more difficult search / find and replace, regex find and replace, any text manipulation tool not available in visual coding. Hard to say whether visual coding in game frameworks is a boon or a curse. For large teams that include non-technical folks they do seem to be a net benefit.

being able to visually group or cluster related code

Text can be grouped as well.

being able to use zoom in and zoom out to create 'region's of code visible from further out with large labels

Can you elaborate on this?

In what circumstances, generally speaking, do you think visual programming is more desirable than textual programming? Amateur programmers? Media production? Something else?

When you're building a data or signal flow.

It's easier to reason about when you can see the branching, merging, and direction of flow.

Part of the problem with visual programming in the "flow" style is the very fact that "branching" and "merging" has two internally-consistent, equally worthwhile but totally incompatible interpretations. Namely, parallel flow of data (known from "dataflow" diagrams; with parallel juxtaposition representing a product, the empty product standing for a unit flow, and a natural branching operation standing for "copy" or "clone") vs. divergent choice (known from "flowchart" diagrams; with parallel juxtaposition representing an alternative, the empty case standing for a zero flow, and a natural merging operation standing for a processing step that can be reached in multiple ways).

This ambiguity clearly does not exist with a pipeline-like, single-source/single-sink "flow", but it does arise as soon as one has multiple juxtaposed 'flows' in a single diagram or section of a diagram. I'm not aware of visual programming methodologies that try to handle both semantics with equal clarity, and without introducing significant confusion between them.

When you're building a data or signal flow.

All code, or at the very least functional code, can be directly mapped to a DFG. Should all functional programming be done visually?

If a visual interface makes it easier, yes. If it makes it more difficult, no.

That depends on what is being built, and who is building it.

Like most things, it isn't one size fits all.

And you're going to have to actually try it out to discover what works best.

Of course not, because if you write say a sorting algorithm the standard tool for that is structured programming primitives. It would be explained as text in a textbook, with ifs, whiles and variables assignations, not CFG / DFG prilitives.

In contrast for instance in music, patchs were already the main interaction and creation mean even before computers existed - see e.g. music studio patchbays, modular synths or even guitarist pedalboards. So people in that field already have the mental model of a signal flow and bricks connected together. Both models can mix - a lot of people use small snippets of code in Max/MSP for instance. In the software I develop, https://ossia.io it's super common that people will rely on a JS script embedded in a timeline for one small part of their score. Etc etc..

When you're working on a touch screen device without a keyboard.

routinely shipped software entirely built in Max. care to elaborate? music-related software?

Visual programming is great for some application areas.

I recently wrote software that allows you to transform data step-by-step visually (https://www.easydatatransform.com). I think the visual, flow-based approach here is a real win, especially for non-programmers.

However I wrote the software in a primarily text-based programming environment (C++ and QtCreator). I think it would have been a nightmare to do in a purely visual programming environment.

Horses for courses.

"Flow-based" is only one kind of visual programming. UML alone has over a dozen different kinds of standardized diagrams. I can write a state machine with a big switch statement, for example, but I still find them easier to read when presented visually.

Maybe we just haven't figured out good visualization types yet for some kinds of programs we write. Interestingly, the opposite is not true. I have yet to see a case where we had a half-decent visualization, and people still said "This would be easier to understand if it were 1000 lines of C++".

During times when the world isn't falling apart, I teach Scratch to kids on weekends: https://scratch.mit.edu/

I don't pretend that Scratch's drag-and-drop block paradigm is a viable replacement for text-based programming, but I do think it demonstrates that there might be a text-like system we could create that would free us from keyboards, without reducing our source code to piles of spaghetti.

Also, is declarative programming typically mappable to a visual paradigm without severe compromise? Think of all the applications that produce non-executable documents; don't they factor into the visual programming discussion?

These issues are old.

- Large programs / graphs? I can browse maps and photos of the entire planet from my web browser. ZUIs are well understood now. Also, computers got really big and fast. My graphics card has 8 GB (!) of RAM. Find a VPL that was designed for the era where graphics are cheap.

- Automatic layout is hard? Yes, an optimal solution to graph layout is NP-complete, but so is register allocation, and my compiler still works (and that isn't even its bottleneck). There's plenty of cheap approximations that are 99% as good.

- No formal spec? They sound desperate to come up with excuses why something can't work. How many (textual) programming languages had a formal spec before they got popular? I tried to look up when Ruby got a formal BNF grammar, for comparison, and it seems the answer in 2020 is still "nope, whatever parse.y does (13,000+ lines of C), that's Ruby".

- Difficulty to create editors? This is a legitimate concern, as it's true "there are no general purpose Visual Language editors", but that's an economic issue, not a technical one. Plain text is one of the very few formats where we have general-purpose editors. In most areas of computing, profitability has taken over, and companies refuse to cooperate. There's no one word processing or spreadsheet or database format, either.

- Lack of evidence? Again, they're holding VPLs to a higher standard. When a language like Ruby appeared, nobody asked for proof. Simply creating something that some programmers liked was sufficient.

- Poor representation? That's an issue with all kinds of programs! If I had a nickel for every horribly written program I've had to deal with in my life...

- Lack of portability? This is another symptom of the editor problem, above. Make a standard file format, and actually figure out how to get people to use it, and there's no problem.

These issues are from an article written in the 1980's. The same day as its publication, Apple released the Macintosh IIci, with a 40 or 80 MB hard disk, 1 or 4 MB of RAM, and a 640x480 pixel display. (For all that computing power, the starting price was $6269 -- only slightly less than the $6615 starting price for a Honda Civic that year.) Clearly, visual programming was a non-starter in that day and age. So were 90% of the other things I do on my computer every day. That doesn't mean they're bad ideas.

If you'd have asked me to write down a list of issues with C++ in 1989, I'd have a list that was similarly useless in 2020.

One more general-purpose thing I miss very much is filesystem. Really what is a good option for a usb disk filesystem which you would like to use on Mac, Win and Linux? Seems to me that without using Paragon/Tuxera/etc. you're screwed.

Agreed. For sharing a filesystem, I assume the best you can do today is probably FAT32 (or whatever the latest of that family is), which is not great.

Why can't we get people from major tech companies together and have them develop a good on-disk format which supports the features we all need, and then everybody can go back and write their own implementations? Like we do with Unicode, or IP. And like we should do for many other features.

For all the talk about the importance of separation of interface and implementation, operating system designers really suck at it.

The modern answer is exFAT. August last year Microsoft opened up the specifications, and pledged that it won't be patent encumbered[].

How well an operating systems has a separation of concerns between interface and implementation really has very little (unfortunately) to do with business incentives. Even if the Linux kernel had perfect separation of concerns, there's no way to force Microsoft or Apple to support it in their products.

Not to mention the problem with standards: https://xkcd.com/927/

[1] https://cloudblogs.microsoft.com/opensource/2019/08/28/exfat...

Visual programming languages are commonly used for the creation of chatbots in frameworks such as IBM Watson Assistant. They end up being an absolute complete mess. The reliance on global variables and go-to statements for control flow makes them extremely bug prone and unintelligible.

But who’s using them? From the symptoms you describe I would guess it’s people very far removed from software development.

The Sims uses a visual programming language called "SimAntics" for scripting behaviors and "AI" of people and objects.

The Sims, Pie Menus, Edith Editing, and SimAntics Visual Programming Demo. This is a demonstration of the pie menus, architectural editing tools, and Edith visual programming tools that I developed for The Sims with Will Wright at Maxis and Electronic Arts.

https://www.youtube.com/watch?v=-exdu4ETscs

https://medium.com/@donhopkins/the-sims-pie-menus-49ca02a74d...

The Sims Steering Committee - June 4 1998. A demo of an early pre-release version of The Sims for The Sims Steering Committee at EA, developed June 4 1998.

https://www.youtube.com/watch?v=zC52jE60KjY

https://medium.com/@donhopkins/the-sims-pie-menus-49ca02a74d...

Mod The Sims Wiki: SimAntics:

https://modthesims.info/wiki.php?title=SimAntics

Simantics Wiki:

http://simantics.wikidot.com/

SimsTek Wiki: SimAntics

https://simstek.fandom.com/wiki/SimAntics

From Ken Forbus's game design course:

Under the hood of The Sims:

http://www.cs.northwestern.edu/~forbus/c95-gd/lectures/

Programming Objects in The Sims:

http://www.qrg.northwestern.edu/papers/Files/Programming_Obj...

Lack of evidence? Again, they're holding VPLs to a higher standard. When a language like Ruby appeared, nobody asked for proof. Simply creating something that some programmers liked was sufficient.

If there was a visual programming language with anywhere near the popularity of Ruby, I'd be willing to consider that maybe the idea has some merit.

Excel.

I could turn your argument around: If Ruby were anywhere near as popular, widely used, and successful as Excel, I'd be willing to consider that maybe the idea that Ruby is a viable programming language has some merit.

But I won't, because whether or not something is a visual programming language isn't up to a popularity contest.

Can you come up with a plausible definition of visual programming languages that excludes Excel, without being so hopelessly contrived and gerrymandered that it also arbitrarily excludes other visual programming languages?

I could turn your argument around: If Ruby were anywhere near as popular, widely used, and successful as Excel, I'd be willing to consider that maybe the idea that Ruby is a viable programming language has some merit.

How is Excel, as widely used, a programming language? Excel sheets in their widely used form are not instructions or behaviour; you can't run them, only manually edit them. Perhaps there is a way to turn Excel sheets into programs, but I highly doubt such a way of using Excel would be more popular than Ruby.

Can you come up with a plausible definition of visual programming languages that excludes Excel, without being so hopelessly contrived and gerrymandered that it also arbitrarily excludes other visual programming languages?

I'd say a visual programming language is a programming language for which the primary way of making changes to the program is visual. Even if we were to consider Excel a programming language, the primary way in which people make changes to Excel sheets - especially advanced users who are doing more programming-like things - is by editing textual formulae, IME.

Are we talking about the same "Excel"? I mean Microsoft Excel, in case there's something else called Excel that I haven't heard of.

Microsoft Excel spreadsheets certainly do contain instructions and behavior. That's the whole point of spreadsheets, that distinguish them from plain CSV files! They continuously run every time you make any change to them. And Excel is several orders of magnitude more popular and successful than Ruby: there is no comparison.

Excel even supports "Programming by Demonstration" through its macro recorder. Programming by Demonstration happens to be one of Brad Myer's favorite areas of research!

Brad wrote the 1989 paper about visual programming that we're discussing, and Visual Programming, Programming by Demonstration, and Programming by Example are closely interrelated topics that he has done a lot of work with and written a lot about.

I think it's important to consider some of his and other people's more recent work, before extrapolating on the technological limitations that he wrote about in 1989, without taking into account 31 years of progress, research, and development.

https://en.wikipedia.org/wiki/Brad_A._Myers

"Myers is a leading researcher in the field of programming by demonstration and created the Garnet and Amulet toolkits."

https://en.wikipedia.org/wiki/Programming_by_demonstration

"In computer science, programming by demonstration (PbD) is an end-user development technique for teaching a computer or a robot new behaviors by demonstrating the task to transfer directly instead of programming it through machine commands."

Watch What I Do: Programming by Demonstration. Edited by Allen Cypher, Daniel Conrad Halbert, David Kurlander, Henry Lieberman, David Maulsby, Brad A. Myers, and Alan Turransky.

https://archive.org/details/watchwhatido00alle

http://acypher.com/wwid

Watch What I Do: Foreword by Alan Kay

http://acypher.com/wwid/FrontMatter/

I don't know who first made the parallel between programming a computer and using a tool, but it was certainly implicit in Jack Licklider's thoughts about "man-machine symbiosis" as he set up the ARPA IPTO research projects in the early sixties. In 1962, Ivan Sutherland's Sketchpad became the exemplar to this day for what interactive computing should be like--including having the end-user be able to reshape the tool. [...]

Microsoft Excel Macro Programming History

https://en.wikipedia.org/wiki/Microsoft_Excel#Macro_programm...

"History: From its first version Excel supported end-user programming of macros (automation of repetitive tasks) and user-defined functions (extension of Excel's built-in function library). In early versions of Excel, these programs were written in a macro language whose statements had formula syntax and resided in the cells of special-purpose macro sheets (stored with file extension .XLM in Windows.) XLM was the default macro language for Excel through Excel 4.0. Beginning with version 5.0 Excel recorded macros in VBA by default but with version 5.0 XLM recording was still allowed as an option. After version 5.0 that option was discontinued. All versions of Excel, including Excel 2010 are capable of running an XLM macro, though Microsoft discourages their use."

I'd say a visual programming language is a programming language for which the primary way of making changes to the program is visual.

Your arbitrarily gerrymandered definition turns C++ into a visual programming language, which it clearly is not.

So how is making changes to a C++ program in VI or Emacs not visual, then? People use visual editors to make changes to textual C++ programs all the time.

The "V" in "VI" STANDS FOR "Visual", because it is the "visual mode" of the line editor called ex.

But the actual structure and syntax of a C++ program that you edit in VI is simply a one-dimensional stream of characters, not a two-dimensional grid of interconnected objects, values, graphical attributes, and formulas, with relative and absolute two-dimensional references, like a spreadsheet.

https://en.wikipedia.org/wiki/Vi

The original code for vi was written by Bill Joy in 1976, as the visual mode for a line editor called ex that Joy had written with Chuck Haley.[3] Bill Joy's ex 1.1 was released as part of the first Berkeley Software Distribution (BSD) Unix release in March 1978. It was not until version 2.0 of ex, released as part of Second BSD in May 1979 that the editor was installed under the name "vi" (which took users straight into ex's visual mode),[4] and the name by which it is known today. Some current implementations of vi can trace their source code ancestry to Bill Joy; others are completely new, largely compatible reimplementations.

The name "vi" is derived from the shortest unambiguous abbreviation for the ex command visual, which switches the ex line editor to visual mode. The name is pronounced /ˈviːˈaɪ/ (the English letters v and i).[5][6]

They continuously run every time you make any change to them.

Which is very much unlike what a program does.

FYI, the "V" in "VI" STANDS FOR "Visual", because it is the "visual mode" of the line editor called ex. Your arbitrarily gerrymandered definition turns C++ into a visual programming language, which it clearly is not.

I appreciate your condescension, but I'm well aware of the history of vi and in any case quite capable of using wikipedia myself.

How is taking the plain meanings of the words in the definition "arbitrarily gerrymandered"? You tell me what visual means to you and how C++ is not a visual programming language, if that's somehow clear in a way that does not apply to Excel.

> They continuously run every time you make any change to them.

Which is very much unlike what a program does.

You're saying "Program that run continuously every time you make any change are very much unlike what a program does?" That doesn't make any sense to me at all, can you please try to rephrase it?

Speaking of program that run continuously, have you ever seen Bret Victor's talks "The Future of Programming" and "Inventing on Principle", or heard of Doug Engelbart's work?

The Future of Programming

https://www.youtube.com/watch?v=8pTEmbeENF4

Inventing on Principle

https://www.youtube.com/watch?v=8QiPFmIMxFc

HN discussion:

https://news.ycombinator.com/item?id=16315328

"I'm totally confident that in 40 years we won't be writing code in text files. We've been shown the way [by Doug Engelbart NLS, Grail, Smalltalk, and Plato]." -Bret Victor

Do you still maintain that "Excel sheets in their widely used form are not instructions or behaviour", despite the examples and citation I gave you? If so, I'm pretty sure we're not talking about the same Microsoft Excel, or even using the same Wikipedia.

Your definition is arbitrarily gerrymandered because you're trying to drag the editor into the definition of the language, while I'm talking about the representation and structure of the language itself, which defines the language, not the tools you use to edit it, which don't define the language.

I'll repeat what I already wrote, defining how you can distinguish a non-visual text programming language like C++ from a visual programming language like a spreadsheet or Max/MSP by the number of dimensions and structure of its syntax:

But the actual structure and syntax of a C++ program that you edit in VI is simply a one-dimensional stream of characters, not a two-dimensional grid of interconnected objects, values, graphical attributes, and formulas, with relative and absolute two-dimensional references, like a spreadsheet.

Text programming languages are one-dimensional streams of characters.

Visual programming languages are two-dimensional and graph structured instead of sequential (or possibly 3d, but that makes them much harder to use and visualize).

The fact that you can serialize the graph representation of a visual programming language into a one-dimensional array of bytes to save it to a file does not make it a text programming language.

The fact that you can edit the one-dimensional stream of characters that represents a textual programming language in a visual editor does not make it a visual programming language.

Microsoft Visual Studio doesn't magically transform C++ into a visual programming language.

PSIBER is an interactive visual user interface to a graphical PostScript programming environment that I wrote years after the textual PostScript language was designed at Adobe and defined in the Red Book, but it didn't magically retroactively transform PostScript into a visual language, it just implemented a visual graphical user interface to the textual PostScript programming language, much like Visual Studio implements a visual interface to C++, which remains a one-dimensional textual language. And the fact that PostScript is a graphical language that can draw on the screen or paper doesn't necessarily make it a visual programming language.

https://medium.com/@donhopkins/the-shape-of-psiber-space-oct...

It's all about the representation and syntax of the language itself, not what you use it for, or how you edit it.

Do you have a better definition, that doesn't misclassify C++ or PostScript or Excel or Max/MSP?

You're saying "Program that run continuously every time you make any change are very much unlike what a program does?" That doesn't make any sense to me at all, can you please try to rephrase it?

Running continuously every time you make any change is very much unlike what a program does. Programming is characteristically about controlling the sequencing of instructions/behaviour, and someone editing a spreadsheet in the conventional (non-macro) way is not doing that.

Do you still maintain that "Excel sheets in their widely used form are not instructions or behaviour", despite the examples and citation I gave you? If so, I'm pretty sure we're not talking about the same Microsoft Excel, or even using the same Wikipedia.

This is thoroughly dishonest of you. You edited those points and examples into your comment, there was no mention of macros or "programming by demonstration" at the point when I hit reply.

To respond to those added arguments now: I suspect those features are substantially less popular than Ruby. Your own source states that Microsoft themselves discourage the use of the things you're talking about. Excel is popular and it may be possible to write programs in it, but writing programs in it is not popular and the popular uses of Excel are not programs. Magic: The Gathering is extremely popular and famously Turing-complete, but it would be a mistake to see that as evidence for the viability of a card-based programming paradigm.

Your definition is arbitrarily gerrymandered because you're trying to drag the editor into the definition of the language, while I'm talking about the representation and structure of the language itself, which defines the language, not the tools you use to edit it, which don't define the language.

Anything "visual" is necessarily going to be about how the human interacts with the language, because vision is something that humans have and computers don't (unless you're talking about a language for implementing computer vision or something).

I'll repeat what I already wrote, defining how you can distinguish a non-visual text programming language like C++ from a visual programming language like a spreadsheet or Max/MSP by the number of dimensions and structure of its syntax:

But you can't objectively define whether a given syntactic construct is higher-dimensional or not. Plenty of languages have constructs that describe two- or more-dimensional spaces - e.g. object inheritance graphs, effect systems. Whether we consider these languages to be visual or not always comes down to how programmers typically interact with them.

PSIBER is an interactive visual user interface to a graphical PostScript programming environment that I wrote years after the textual PostScript language was designed at Adobe and defined in the Red Book, but it didn't magically retroactively transform PostScript into a visual language

There's nothing magical about new tools changing what kind of language a given language is. Lisp was a theoretical language for reasoning about computation until someone implemented an interpreter for it and turned it into a programming language.

Agreed! Many of those claims completely disregard the research in visual programming that has happened in the past 30 years. Specifically the idea that it is about visual vs. textual seems utterly outdated now that more and more modern development environments find ways to combine the best of both worlds.

If anyone finds the time to update these claims, please consider our vvvv in your research: - A visual programming environment for the .NET ecosystem - Compiles to C# using Roslyn - Can use code from any .NET library as visual code blocks - Its language VL augments dataflow with features from OOP and functional programming, supports generics and easy multithreading - Useful since 2002: https://vimeo.com/371511910

Free for non-commercial use without restrictions (Windows only): http://visualprogramming.net

vvvv definitely doesn't solve all known issues but we hope it raises the bar for visual programming and such discussions.

I'll preface by saying that vvvv looks impressive and certainly seems like one of the more polished options in this space. We have a type of visual programming as part of an internal configuration tool, and I will definitely look into vvvv for how things can be done differently.

That said, this screenshot from one of your tutorial videos highlights the type of thing I think people are worried about: https://i.xkqr.org/screenshot_2020-04-27T09:47:46.png

I'm not saying it's bad – it might just be a question of habit whether or not graph edges are better than named identifiers in text, but it is hard to build away, even with best in class tooling.

Whatever happened to Luna Lang? https://luna-lang.org/

This is a good and thoughtful essay, but in some respects reads like a rationalization of the author's preferences rather than an attempt to engage with the problems. He makes good points about the lack of formal specification, portability, and specialty software; this is indeed a problem, as all the best visual programming platforms are proprietary.

On the other hand, there's no fundamental reason that visual editors shouldn't be able to load and save source code; it's just a different way of thinking about program structure and user interface. There's already a lot of tools out there that generate dependency graphs within or alongside an IDE; the main shortcoming i see is a lack of nesting. There are few instances in which i want to see the entire structure of a program all at once, any more than I want to read all the source code in a single giant text buffer

One might equally ask if textual languages are so great, why are there so many of them? Couldn't everything be written in a nice capable high-level language like C? And of course it could, but lots of people don't like C's minimalism and lack of guardrails. It seems to me that most new general-purpose languages originate in a desire to have the computer do some of the tedious housekeeping (like casting variables) and frustration with awkward syntax.

I'm old, so when I was learning to write code it was just flat text files and every build necessitated a round of swearing over forgotten semicolons, mismatched brackets, or incorrectly spelled variables. Syntax highlighting, code folding, autocompletion, and bracket matchings which do so much to make development in a modern IDE pleasant and productive were once viewed as undesirable crutches that would lead to poor practices, lack of portability, and inaccurate second-guessing of developers' intention by the IDE, subjecting hapless coders to the tyrannical will of the interface illuminati.

I suspect that behind the aversion to visual programming lies a fear of standardization and automation that will result in widespread developer redundancy as it becomes easier for domain experts to build out their tools without code specialists, somewhat parallel to the development of electronics from hobby clubs where people helped each other with soldering irons, multimeters, and resistor identification charts to today's electronic factories populated only by pick-and-place machines assembling components too small for adult hands to manipulate at dizzying speeds.

The more automation and and ML that finds its way into IDEs, the more standardized the underlying libraries are likely to become. To the extent that programmers are freed from syntax production and housekeeping duties and processes can more easily be read (from visual graphs/charts) with domain knowledge alone, the more likely it is that the most common functions become wholly standardized, in the same way that chips are. 555 timer ICs have been around since 1971 and the most popular derivatives are dual and quad versions. Nobody is writing new and improved 555 timers and there's no real demand for them.

By the way there's a new version of ndepend, which is not quite the same thing but might be of interest to people here as it's conceptually related: https://www.ndepend.com/docs/visual-studio-dependency-graph

When a programmer says "I want visual programming" I think it's often one of those situations where 1) the asker has correctly identified a pain point, but 2) the thing they are asking for isn't necessarily the right way to do it.

Instead of a box-and-lines diagram, I think the stuff that people really want is:

- Structured editing. Valid change operations are obvious (maybe listed in a menu). No invalid states.

- Cross-cutting connections are explicitly displayed (with lines or arrows or etc) instead of being implicit.

- UI does a better job of highlighting just the relevant information and hiding everything else.

- Overall more simplicity. As in, I think some people imagine that their thousand-lines-of-code project could be fully represented as a few boxes and lines with labels. There's some unrealistic expectations tied up in this area. But simplicity is still a good goal.

Those are all good goals and there's different approaches to solve them. A lot of those asks can be solved by keeping the text document interface and adding enough cleverness on top of that (like EVE / LightTable). And there's other visual systems other than boxes-and-lines. People just need to think outside the box. (ha ha, sorry)

Visual programming doesn't just mean boxes and lines. It could mean anything other than plain text.

I do a fair bit of UE4 coding in both Blueprint & C++. Each has benefits and drawbacks, many of which have been mentioned already.

The biggest unmentioned one from a learning perspective though, is that with node-based editors, while you can still make logic errors, it’s difficult to actually make syntax errors. When you get used to it, you might not mind forgetting occasionally to match brackets, remember semicolons, use the right keywords in the right places etc, but as a beginner, all of these things can be quite frustrating and opaque, with often cryptic error messages.

With visual node-graphs though, many of these things are just non-issues, which I think helps beginners gain confidence way faster without bouncing off and giving up in frustration. You can’t even connect pins of the wrong type most of the time, so you get some pretty well communicated type safety to boot.

I’d actually love to see an evolution of a fairly advanced visual programming tool like Blueprint turn into a more general-use language, although preferably with files that are still stored as plain text (so at least play nicely with git etc) rather than the binary black boxes that are currently chewing up my LFS...

I would love to have a tool that would display code as some kind of annotated AST, rather than the layout imposed by most IDEs or tools such as gofmt.

Put a bag over Java’s ugly face, that kind of thing.

I don’t want to look at IntelliJ’s “staircase of doom” formatting, I want to see one parameter per line when there is more than one parameter.

I don’t want to trace down data “COME FROM”s (local variables that are only used once and have to be tracked down from whence they came).

Give me an alternate layout which directly tracks between the text and the diagram, similar to a structure layout, but showing details within functions/methods as a tree, as well. Show me clearly how values/expressions funnel or pipeline into each other, but filter out some of the noise.

And auto-display fuglyCaps names as kebab-case, while we’re at it :-). Cuz, science.

What you are describing is closely related to some of the ideas of Intentional Programming.

The core idea is that the code is just a model (AST, graph...) and your editor projects it to whatever format or syntax you like.

Development then is coding in these representations of the model or building new smart abstractions and model transformations (DSLs) for going from your screen representation to lower-level building blocks and eventually compiling to e.g. something that runs on the JVM.

It quickly becomes just as complicated as it sounds.

The best implementation in my experience is that of JetBrains’ free tool MPS (Meta Programming System). They start with the model (your AST) rather than the code.

The learning curve is steep and it is difficult to adopt it in teams where people are more focused on shipping features than building exotic tooling to help them do it.

I just want a modern spreadsheet-like interface that isn't limited to the concept of sheets. A visually-represented functional language that lets a user with no programming knowledge run a pipeline of data transformation and ultimately output the data in a management/client-ready format.

I want to build it myself, but I have neither the time nor the expertise.

What limitations are you looking to remove? 2d to multi-dimensional? Tables that float instead of living on sheets?

EDIT: I think users can only really easily operate with 2D data, but to borrow someone else's acronym, a ZUI with multiple layers / levels of data could be more ergonomic than a vanilla GUI, so in a way I am indeed arguing for 3D.

And yes to floating tables. Probably allowing users to also reference other sheets/tables semantically rather than just by row-column.

Also creating explicitly defined (visual) pipelines of how the data is being transformed so users can audit and understand complex "spreadsheets".

Auditing deeply linked calculations by following text formulas is too cumbersome for most users.

I've sat through a presentation done with a ZUI (Zoomable User Interface) and it was nauseating. I literally almost threw up due to the unexpected motion every few seconds.

Do you mean like Node-RED, or something different? https://nodered.org/

All programming is visual, think about it. Text is visual. I'm not just pointing out a semantic discrepancy here. Text is a specific method of encoding meaning and processes in a visual way.

In short, textual programming is just a subset of visual. People think of the dichotomy as a duality of text vs. visual, it is not.

Text just has features people tend to like: A Fixed line based structure. Isomorphism with the english vocabulary... Who says other visual programs can't have this either?

The issue with a lot of these "visual programs" is that they also represent a subset of Visual Programming similar to how text is just a tiny subset of visual programming. What you are in fact seeing with most "visual programs" is mostly programs represented as graphs.

The real dichotomy people are discussing is not text vs. visual but text vs. graph. Whenever I see something like labView or something labeled as a "visual program" it's always a program represented with a series of nodes and lines, aka a graph.

Hopefully by illustrating the reality of what's going on, people can see beyond representing programs as just text or just graphs.

There are other ways to represent modules and processes.

Well put! Text is graphics, but technology has historically limited how it looks and what you can do with it. The idea of text with control characters like "carriage return" and "newline", fixed width text, limited length lines, and text editors that wrap words when the cursor gets to the edge of the screen, and terminals with multiple lines that scroll, all incrementally evolved out of emulating punched card machines. And now we emulate VT100s emulating punched card machines. But it doesn't have to be that way.

Here's a great piece of exploratory research that imagines how things could be different, by an amazing visionary visual thinker and artist, Scott Kim:

"Viewpoint" is part of Scott Kim's PhD dissertation that he did at Stanford in 1988 (with Donald Knuth as principle advisor). It was programmed in Cedar, and ran on a color Dorado workstation at Xerox PARC.

https://www.youtube.com/watch?v=9G0r7jL3xl8

Demo and explanation of Viewpoint, a computer system that imagines how computers might be different had they been designed by visual thinkers instead of mathematicians. Caution: this is basic research, not a proposal for a practical piece of software. Part of my PhD Dissertation at Stanford University in 1988.

At 12:22, he starts with a blank screen, and performs a "visual boot", to use Viewpoint to build itself.

http://www.scottkim.com/

https://en.wikipedia.org/wiki/Scott_Kim

https://visualraccoon.wordpress.com/2013/08/14/the-viewpoint...

This is a visualraccoon Perspective on Scott Kim‘s very graphic and revolutionary exploration, Viewpoint: Toward a Computer for Visual Thinker.

What would it be like to go back to visual first principles and take a fresh look at graphic user interfaces?

The Viewpoint Thesis is that a small number of pixel manipulation primitives can be defined such that if they are bound to keyboard and mouse actions it is then possible to build a simple text-graphic editor by drawing it, and that that editor can be used to draw-build itself.*

The Viewpoint Thesis & Editor is part of a larger project founded on the hypothesis that:

“Only by treating the screen itself as a first class citizen will we be able to build computers that are truly for visual thinkers.” Scott, 1987.

This project includes building visual programming languages for such thinkers.

Viewpoint: Toward a Computer for Visual Thinkers

http://www.scottkim.com.previewc40.carrierzone.com/viewpoint...

http://www.scottkim.com.previewc40.carrierzone.com/viewpoint...

When I started my PhD project in 1981, the IBM PC was brand new, and it would be years before the Mac and Microsoft Windows would appear. Nonetheless I was familiar with graphical user interfaces because of my internship at nearby Xerox PARC, the visionary research center that spawned, among other things, the laser printer and the bitmapped display. I was well-acquainted with Alan Kay's vision of a dynabook (he envisioned the notebook computer in 1975), and understood the power of graphic user interfaces and graphic tools like paint programs to bring the power of computers to artists and visual thinkers.

At the time I was part of the digital typography program within the Stanford computer science department, which built computer programs for digital typeface designers. I was also enthralled by the Visual Thinking course at Stanford, which taught engineers in the product design program how to think visually.

It struck me as odd, and deeply wrong, that we were building tools for visual artists in a programming language that was utterly symbolic, and lacking in visual sophistication. I yearned for a programming language that had the same visual clarity that graphic user interfaces had.

So I set about wondering what a visual programming language might look like. If computers had been invented by artists and visually oriented people, instead of by mathematicians and engineers, how might they write programs? It seemed to me an important question, but one that hardly bothered most computer scientists. I read about a few attempts to build visual programming languages, and decided there was something fundamental I needed to understand.

My journey took me deep into the foundations of computer science, where I asked fundamental questions like “what is programming” and “what is a user interaction” — questions that often get passed over in computer science (any definition of “programming” that starts with “a sequence of symbols that…” is not deep enough to encompass visual programming languages). I never did build a visual programming language, but I did figure out a fundamental idea (things get interesting when the user and the computer share the same mental model of data), and built a rudimentary visual editor that demonstrated my ideas.

I’m posting my dissertation and the accompanying video demo to re-open the conversation. What in my dissertation is interesting to you? What other work do you know of along these lines? Do you know of foundational research in interaction design and visual programming? And most importantly, what would be a good next step to push the work forward? (My research stalled for lack of a juicy application domain.) Please email me your thoughts. I'd love to hear them.

I do not agree with the freeform nature of viewpoint.

The possibilities are boundless but you have to temper the boundlessness with the fact that our brains are biased and specialized to work in certain ways. Case in point can you communicate your entire post in terms of pictures? Is there any form of non linear diagramming you can do on a 2D canvas that can deliver your point better than text? If I give you the freedom of a 2D canvas to convey your point can you convey it better than the restricted form of text? No. In fact text was the best way to communicate your point. Restriction and lack of freedom are often more effective tools than ones that are completely unrestricted and boundless.

It doesn't matter whether you're a visual thinker. When it comes to parsing and conveying complex information our brains have a dedicated module for parsing information that comes in as a linear and serial string of symbols. If you can talk, listen read or write than you can be exceptional at writing programs that do this. Your brain is biologically biased to operate this way.

That being said if I show you a picture of a dog, you will learn more about what that dog looks like than any amount of textual description. There is no denying that certain things are communicated better visually and other things are communicated serially.

It's hard to delineate axioms that communicate exactly what is better communicated visually and what is better communicated textually. I can only talk in terms of examples: Your ideas on visual programming is better communicated through text, what your car or face looks like is better communicated through pictures.

One thing that has worked throughout history is a combination of text and pictures. Powerpoint. A linear slide show where each slide delivers serial text and pictures in parallel to the human brain.

The future of visual programming, if it occurs at all, will ultimately be some form of a combination of diagrams and restricted serial strings of symbols following the same principles as powerpoint. I don't believe it has anything to do with foundational computer science but rather the language must be built around the biases and limitations of the human brain. We build biased textual languages because our brains are partly biased towards text.

I have an idea about what that would look like and how that would work in terms of maintaining programmer productivity. Also I didn't see your dissertation (assuming your name is Don Hopkins) in your post.

Is there any form of non linear diagramming you can do on a 2D canvas that can deliver your point better than text?

I could write a JavaScript program just for you, that drew blinking animated sparkly text in a canvas which said "WATCH THE VIDEO I POSTED", but there would still be no point to that, since you can lead a horse to water, but you can't make him drink.

Because if you actually watched the video, you would realize that Viewpoint is all about text. Nobody claimed it was "better than text", or that it was trying to get rid of text. You'd know that if you had watched the video. But you don't know that, so I can conclude that me simply writing text was not enough to get you to watch the video.

I have an idea about what that would look like and how that would work in terms of maintaining programmer productivity.

That's very nice. And do you know how much that's worth? Nothing. Not only are ideas worthless without an implementation, but also ideas that you haven't even bothered to write down without implementing are worth even less, because they're a waste of everyone's time, and not tempered by reality or experience or empirical testing or iterative design and development or stakeholder requirements or user feedback.

What point is there to telling me that you have an idea about something you haven't implemented, or even written down, or drawn a picture of? Or did you forget to provide a link to your own dissertation, a video, and a runnable demo?

It's much better for you to spend your precious time doing difficult tedious things like writing down and implementing and testing your brilliant ideas yourself, instead of having so much fun talking about how great it would be if only you wrote it down for other people to read, payed somebody on Fiverr to draw a picture of your idea for you, or found some out-of-work but highly skilled programming with nothing better to do, who was willing to pair up with a starry-eyed "idea guy", and work night and day for free, on your sincere verbal promise of equity and exposure, implementing your grandiose but unwritten ideas, in the hopes that you don't get angry at them for not precisely interpreting and efficiently engineering your brilliant vision correctly, or suddenly change your mind once you see what they actually implemented that did not perfectly live up to your idealistic expectations and nebulous handwaving.

Good luck with that! Let me know when it ships.

You misunderstand. I watched it. I'm talking about information delivered in a serial format. A linear and ordered ingestion of information. I'm not talking about text per se, but the property of text that makes all programs follow a linear order of execution. Viewpoint is anything but.

Viewpoint is all over the place. It's basically a 2D canvas with no order. Sure it has text but that text is used more like a random label with random coordinates. There's no linear process of instructions delivered or ingested. What you see is basically a reactive canvas with arbitrary coordinates.

The universe we live in has a linear one dimensional property that follows the arrow of time. Viewpoint fails to represent this concept symbolically. The process of time, the process of execution. However you choose to represent this concept, all programming languages must have some sort of way of representing time as a spacial dimension. For text it is left to right, top to bottom reading order, for viewpoint it does not exist.

What's more all the symbols can be destroyed as text is just an abstraction over a pixel drawing interface. The axioms (pixels) of this interface are too low level. But this is just a minor flaw.

The main flaw I see with Viewpoint is that it's not clear how computation can be modeled with these pixels. How do you represent logic? Or, and, nor, xor? Is viewpoint Turing complete?

Time is an intrinsic part of logic, a programming must be able to represent logic and time by using only spacial dimensions.

I think you need to go back and watch it again, because what you're saying doesn't align with what the video shows, and it sure sounds like you misunderstand, and that you don't know what you're talking about.

Did you notice the part in the video description that I quoted in my post that said "Caution: this is basic research, not a proposal for a practical piece of software. Part of my PhD Dissertation at Stanford University in 1988"? While you're at it, read Scott's introduction and PhD dissertation that I linked to, as well.

Do you even know who Scott Kim or his thesis advisor Donald Knuth are, or how the research at Xerox PARC in the 80's contributed to the fields of programming language and user interface design? What were you doing in 1981?

So what's that great idea you claim to have? Where's the link to your dissertation and a runnable demo?

So what's that great idea you claim to have? Where's the link to your dissertation and a runnable demo?

That's mine. When it's ready I'll reveal it.

I think you need to go back and watch it again, because what you're saying doesn't align with what the video shows, and it sure sounds like you misunderstand, and that you don't know what you're talking about.

In his video he even called viewpoint an "editor" as opposed to a "visual programming language." It's clearly not Turing complete and I would say outside the topic of visual programming languages. I think he's more in the domain of researching user interfaces for the common computer user rather than a visual interface for programmers and programming in general.

That's mine. When it's ready I'll reveal it.

https://www.youtube.com/watch?v=RKMNPQ35OUc

Oh, it looks like we have ourselves a classic "idea guy"! I've met a whole lot of people like you before. It must be hard getting a job in such a crowded field, especially at a time like this.

So that's a hard no, you've never heard of Scott Kim or Donald Knuth or Xerox PARC, right?

I can see you have excellent soft "people skills", and it must be a delight to work with you. Good luck talking somebody else into implementing your ideas for you, but don't let anybody else know what they are, and keep them top secret, and never discuss them with anyone who's done similar work, and make sure everyone you discuss them with signs NDAs with No Complete clauses (and make sure you're in a state where they're enforceable) and Reach-Around Clauses (in case they fuck you up the ass), because your ideas are so easy to implement once you know them, that somebody might steal them from you before you get around to conning somebody else into implementing them for you for free, if you leak any details of all those original ideas of yours that are all yours and nobody else's and they belong to you and are yours. ahem ahem cough

https://www.youtube.com/watch?v=Xs7r5xfucPs

Better see somebody about that cough!

Their might be an under-explored middle ground here, somewhere between representing a program as ASCII text and a full blown graph ZUI. Language is already a very efficient way to layout a graph structure, so lets keep that and borrow from maths the notation for summation, integration, ect, and from physics maybe Dirac's bracket notation for matrixes.

I have thought of doing somethings like that in the browser with a WYSIWYG editor (document.designMode='on') and web components. The DOM is like the AST of the program and is fed into a runtime. Non-expression tags are basically just comments.

Not for regular programming, but for a type of scientific notebook interface to do exploratory programming. Anyway one of this days.

Here is a long list of reasons this is a bad idea: https://blueprintsfromhell.tumblr.com/

Having used my fair share of labview i know such examples are reality. Common mistake of the model is the sentiment that because it's visual everybody can do it! Biggest wtf i experienced was when system engineers wrote half completed "reference-implementations" in labiew, which coders then had to translate to Java for production, to sort out all the edge cases, because these visual tools (and the people that use them) are very bad at expressing edge cases. They might as well have given us graphs drawn on paper because the references were so far from reality you could never run them.

Together with that, when the only tool you have is a hammer, everything looks like a nail. Take https://blueprintsfromhell.tumblr.com/image/162487395171 as an example, it's actually one of the simpler graphs from that page. But if you take a look at it you see it's just doing standard function calls in a sequence, like trunctate(), toText() setText(), what's then the point of making it visual?! For simple signal control loops with only mathematical operations the labview model works really well, but when you start doing things like subString, groupBy or indexOf and don't realize you should change domain, that's when you get the worst spaghetti.

The benefit of a visual representation is that there are different ways to visualize things and you can choose which one fits the problem best. You have sequence, dependency, state, class, call-graphs, signals, lexical parsers, tables, charts, etc. Most visual environments only implement one of these and very soon you are using the wrong tool for the job.