Guest post by figosdev
Summary: Historical perspective on computer languages and how to do better
I have stayed interested in friendly programming languages for decades now. My favourite language, fig, came from another language inspired by the way people talk to the computer on Star Trek. On the Enterprise, people create "programs" used in places such as the holodeck, though most of the lines from the show remind me more of search queries and similar parameters:
"Alexa, please stop spying on us."
"Unable to comply, authorisation from a CEO or higher is needed."
"Alexa, initiate self-destruct sequence, authorisation Bezos-Four-Seven-Alpha-Tango."
This could also be looked at as a command line sequence, with three parts:
1. [ So-called "Wakeup" Command ]
2. [ Initiate command | command parameters ]
3. [ Password syntax | parameter (the password itself) ]
In GNU Bash this might look something like:
# sudo systemctl selfdestruct
Password: Bezos-Four-Seven-Alpha-Tango
systemctl: self-destruct sequence unnecessary, Systemd already running
Most (not all) common and/or friendly programming languages have a simple convention of:
1. a command name
2. command parameters after the name
3. some symbol that goes between commands (or just a newline)
For example, to say "Hello, world" on the screen in Bash, Python, and BASIC:
echo "Hello, world"
print "Hello, world"
PRINT "Hello, world"
First comes the command, then the parameters-- the things you want the command to know or work with.
To the computer, this is all translated into instructions that can be expressed as a long numeric sequence. To simplify this a bit, you can think of the command to put text on the screen as a the number 8:
8 "hello, world"
Of course to the computer, there is no such thing as "put text on the screen." The screen shows the display buffer, the display buffer to the computer is a series of numeric locations, the real job of the computer is to copy certain numeric values to these locations. It doesn't matter whether you're listening to music or running an interactive online 3d simulation, the computer is just a glorified calculator moving numeric values around-- to and from numeric locations.
The job of a compiler author (especially the first of them) used to be to worry about getting simple instructions like "print" translated to these simpler numeric routines (simpler to the computer, more complicated to many of us.) Someone still has to worry about that, but most people don't. Thanks to decades of innovation, if you implement or create a conventional programming language, you will probably use another existing computer language to do so.
For example, CoffeeScript is implemented in (and has since influenced development of) JavaScript. JavaScript was originally written (implemented) in the C language. V8 is a JavaScript implementation in C++. These languages are in many ways easier to work with than the native, numeric "machine language" the computer actually uses.
The pioneer of compiled languages, Grace Hopper, taught university level mathematics-- yet still decided that computer languages would be more useful to a greater number of people if they were based on words, rather than "symbols."
This is a great boon to the general understanding of computing, because these "words" can perfectly abstract the computer hardware itself. How well can they do that? Alan Turing mathematically proved they can do so, but in more practical terms, an emulator program like Qemu that lets you run a second operating system in a window from a "host" operating system is better proof for most people. If we can teach more people to code, we will have more people who are happy to say they are computer literate. We must offer greater computer literacy to everyone, if we are going to make informed political decisions about life and liberty in the 21st century.
The history of the GUI is one of creating a clumsier, more vague language for certain tasks. A GUI is very useful for things that require pointing or drawing, and sometimes very useful for other tasks (such as managing term windows.) But just like when you try to use a hammer for every carpentry task, sometimes our workarounds for not using the command line are a cure worse than the disease. Teaching a GUI is closer to application training-- next year, Microsoft or Apple or IBM will move things around and you'll need to retrain again. You probably won't understand computing by poking at things on a menu. But you can, from learning just the basics of coding.
Believe it or not, the first code written by Linus Torvalds was not in C. His first program went more like this:
10 PRINT "Sara is the best"
20 GOTO 10
This example from the Basic programming language includes the much-maligned GOTO statement. This statement corresponds to the JMP command, one of those numeric instructions mentioned earlier. JMP is a "mnemonic" (useful imaginary association) of the number reserved for that fundamental computing task. Most modern languages avoid use of this command, because it makes it more difficult to maintain larger programs.
Again we see the command name followed by the parameters, and we also see the line numbers that the Dartmouth Time Sharing System (DTSS) used to mark each line of text, or code. Dartmouth is the university where Basic was developed in the early 1960s.
Prior to the existence of Basic for teaching, Grace Hopper influenced the design of COBOL that contained TERSE-ENGLISH-PHRASES that were JOINED-BY-DASHES. Some early computers, even some from Apple, Inc., relied on character sets that did not include lower case letters. Some of the earlier languages (such as Basic) tended towards commands and other text expressed in ALL UPPER CASE.
Around the same time as Basic, the language known as Logo was developed as part of an entire philosophy of teaching. In Massachusetts, the MIT Media Lab continues to explore applications of languages in the same family as Logo, such as AppInventor, Lego Mindstorms and Scratch.
The feature of Logo that is best known, as (like Basic) it was used to introduce many children to coding, is Turtle Graphics. Apart from not requiring graphics initialisation commands like the SCREEN command in Microsoft QuickBasic, Turtle Graphics does not (in many implementations) require much or anything in the way of punctuation in its syntax.
To draw a square is this simple:
UP 5 RIGHT 5 DOWN 5 LEFT 5
There is no symbol required between commands, it simply follows from left to right and continues taking parameters until it finds another keyword. (as with math, statements put in nested parentheses might change this.)
So in Logo, this text:
UP 5 RIGHT 5 DOWN 5 LEFT 5
Translates (conceptually) into English instructions such as:
"Draw upwards 5 points in length, then draw to the right 5 points in length, then draw downwards 5 points in length, then draw to the left 5 points in length."
Other dialects may offer additional options or simply different commands:
TURN 90 FWD 5 TURN 90 FWD 5 TURN 90 FWD 5 TURN 90 FWD 5
This violates a useful convention in coding known as D-R-Y or "Don't Repeat Yourself." A better way to code the same task is:
REPEAT 4 TURN 90 FWD 5
You can change this from a square to a diamond simply by turning 45 degrees first:
TURN 45 REPEAT 4 TURN 90 FWD 5
One of the things that truly limited Basic in the 1980s and early 90s, when the PC revolution was taking place, was not just the lack of processing power but the constraints on RAM and what size a program could be. As the PC moved from 8 and 16-bit platforms to 32-bit, it was more trivial to access large amounts of RAM and have larger programs with larger units of information handling.
For example, a string of characters such as "Hello, world" in the 80s was limited to somewhere between 256 characters and 32,768 characters. You could have string arrays, which are collections of strings with a common name, but loading a large file into an array was costly. Basic traditionally has very rudimentary commands for dealing with strings.
Python deals with strings in a more flexible, modern fashion. A "list" in Python, similar to an array in Basic, can for starters mix types of data-- instead of having to initialise an array as being for strings or numeric data, Python lists can contain one or both at once. Basic has a MID$ function to get part of a string, but Python has a syntax that lets you get part of a string a list.
With increases in RAM and ease of addressing it came a greater use for more flexible commands, which were more practical to implement.
This does not mean Python is easier to learn than Basic-- one of the nice things about Basic is that you could master a good portion of the entire language, and feel like you definitely knew how to write programs in it. Sure, you wouldn't actually learn every command. But you probably knew which commands did something you didn't need for your programs.
Python has in many ways supplanted Basic in education, while Logo continues to provide inspiration for new languages that are easy to learn as a first language. Python can also be taught as a first language, although probably in later school years than Logo-based languages.
Getting back to the earliest languages designed for general education, Basic and Logo both had strengths. Basic was more obviously practical, being used to create video games, accounting programs and even point-of-sale software. There were often better languages (such as C++) for creating commercial software, but Basic enjoyed notoriety from both its success via Dartmouth and its inclusion in chips on board 8-bit computers.
Logo was an easier language, but typically wasn't used (by most of its users, that is) to do anything very practical. Today, it is a good foundation for creating mobile apps or controlling robots, or animating a cartoon cat-- but Logo-based languages (though Logo was originally a general purpose language) are not typically used for a multitude of tasks.
As a teen, I often wondered what a cross between BASIC and Logo would be like, or if such a thing was even possible. To this day, I believe exploring this question and similar questions will make it possible for more people to implement reasonably complex and sophisticated programs, with relative ease. I do not think Python is going to get friendlier, in fact I find it going in a typical direction of expanding until it is much less easy to learn than in the beginning.
Basic too, has grown more complicated than it needs to be. But Python, JavaScript and Logo are all used in education.
For my own efforts added to the pot, I have finally explored the question of combining aspects of Logo and Basic, with fig. Like 1980s versions of Basic, it is case-insensitive. Like Logo, you can avoid most punctuation in syntax. You can also add it optionally.
It goes left to right, demonstrates a variety of tasks such as drawing, array and file handling, getting text from the Internet, and even transitioning to a more feature-rich language such as Python-- while allowing you to create native functions that contain inline Python code.
To make an installer less necessary, the fig translator is a single Python script. It takes a program written in fig, translates it into Python, and (as with interpreted languages in general) requires Python to run the output program.
Fig was featured, thanks to much-appreciated help from the late Robert Storey, in the first 2017 issue of DistroWatch Weekly: https://www.distrowatch.com/weekly.php?issue=20170102
But although I enjoy using it (quite often, actually) as a heavily simplified derivative of Python, fig was always an experiment and a prototype. I would be just as happy to see people explore the concepts of fig as an educational language, which I spent many more years thinking about even while I taught myself to code ("You know what would be cool? An array that uses a string index instead of a numeric one!" ...Linked lists, hash tables, Python dictionaries) than I would be for them to adopt fig itself.
The core idea of fig is simply this-- instead of teaching a programming class, try to sit down everyday people and teach them to code. That's what fig was designed to make possible-- as well as make easier.
I tried doing that with BASIC, it is possible. I tried it with Python, as well as JavaScript. That was harder, though overall I think Python has some advantages over Basic. It has some rough corners too-- "SCREEN 9: PSET(10, 10), 5" draws a dot in early 80s QuickBasic and GW-BASIC. Try doing that in Python, such as you might with Pygame. It's a bit more to begin with. Of course that particular matter could be solved with a different library-- but you still have to learn how to import it.
As you teach more people, and note what it is that gets in the way of them learning, you remove as many unnecessary requirements of coding as possible. I removed Python's case-sensitivity. Although a fan of indented code (I wasn't one until I learned Python) I decided to replace Python's mandatory indentation with Basic-like keywords. It's easier to implement than you might think-- I just use a variable to keep track of the indentation level and increase or decrease it when required.
Before fig, I attempted to create a language with no punctuation at all-- so the print command was like this:
todisplay
hello there, world! (you can put any text of any sort on this line)
display
This is tedious, but I was trying my hand at designing a Basic-inspired language to be spoken into a microphone instead of typed on a keyboard.
When I got bored with that effort and focused on fig (then "fig basic") instead, I still tried to keep punctuation and other syntax minimal. I used "quotes for strings" and # hashes for comments and as I received feedback, I thought it best to allow the option of putting a colon : between commands on the same line.
Ultimately I thought why not do like English, and have several characters available for grouping text together visually? You can teach fig as a way to introduce bash syntax and concepts:
p=5 ; ucase | print
or Basic:
P = 5 : UCASE : PRINT
or Python:
p = 5 ; ucase().print()
I realise these are caricatures, I can write actual code in the syntax of each of these languages.
But "why?" is a question I will answer in three parts:
1. We want more people to understand code, so they can understand computing and help write free software.
2. An educational language didn't taint Torvalds, and most languages (even Basic) now lack the very problems that caused Dijkstra to moan about them so much. Indeed, many of them took a lot of what he said to heart. We can teach more people, more easily, if we have languages better suited to that task.
3. We can learn more about what makes languages easier to learn if we aren't afraid to experiment and collaborate more.
This is not a solved problem-- like with other tasks related to computing, the solutions are ongoing. Despite its revolutionary contributions to computer education, Basic no longer holds the title for most popular or most useful educational language. That title probably goes to Python now-- or perhaps Logo, if you consider it as a language family. Of course I still find great value in Basic as a concept, I would never be happy just switching from existing versions of Basic to existing versions of Logo.
The non-standard, computer language / computer educational philosophy that is Logo may well continue its legacy, but we know the language will continue to change to suit the needs of learners.
Regardless, not everyone is going to learn Python no matter how we try, and some languages are still easier to learn. I think it is trivial to prove (for those who are interested in the idea) that we can make a language that is both easier to teach than Python, and more generally practical than Logo.
I'm very interested in multiple takes on this idea, not just one or two. What kinds of friendly languages with powerful features can we imagine, and perhaps implement? What can we learn from educators in the field (many of whom have their own difficulties with computers and programming.)
This is less about reveling in the past than first learning what we can from historical and contemporary efforts, and hopefully creating greater success in the future.
It is not a requirement, though I really do like the idea of translating to something like Python. You can technically implement a language in almost any other language (provided certain minimal features exist in it) but starting out with a friendly language means the implementation can also be tweaked or inspire a larger number of people. Someone who learns fig can even learn how to implement a similar language. (I have considered, but never gone to the extra trouble, of implementing fig in fig itself. With its inline python feature it is certainly possible.)
License:
Creative Commons CC0 1.0 (public domain)
⬆