Tutorial: Basics

This basics tutorial will teach you how to write and run Nibbles code. You will have a feel for the language and be able to solve real code golf problems, but some things will still be awkward and imperfect. It assumes basic familiarity with and ability to run unix commands and that you know how to code. Knowledge of Haskell is useful but not required.


Running code

To follow this tutorial install Nibbles or run it online. Follow instructions in Try It for either option.

The Big Idea

In Nibbles, each instruction only requires 4 bits (also known as half a byte - aka a nibble). But it's nice to read and write in ascii (literate form), so we write code in ascii and compile it, packing 2 instructions per byte (binary form). In the binary form, ops are heavily overloaded by type, allowing for more than 16 possible instructions.

Typically the literate code you write will correspond 1 ascii character to 1 nibble, so you can basically divide your code length in half. There are some exceptions to this (numbers, strings, etc). But don't worry about this for now.

Let's see this in action.

Create a simple program

Here's a simple Nibbles program in literate form:

+2 1

Save that to intro.nbl (nbl is the file extension for literate Nibbles programs).

Run it

> nibbles intro.nbl
size = 5 nibbles (3 bytes) 3

In addition to running the program, it also output the compactified size of your program to stderr.

You may have also noticed that a file named out.hs appeared. For now, Nibbles works by compiling to Haskell.

Compress it to binary form

> nibbles -c intro.nbl

A file appears named intro.nbb (nbb is the file extension for binary Nibbles programs) which has a size of 3 bytes. That compactification was not as good as the promised divide by 2... but it would have been 2.5 bytes if we were measuring that way (and small numbers are a bad example). FYI you can automatically expand .nbb files back to literate form using -e.

Syntax

Programs are written in Polish (prefix notation). Instead of writing 1+2 we write +1 2. The advantage of this is that you do not ever need parenthesis. (1+2)*3 would be written as *+1 2 3. 1+(2*3) would be written as +1*2 3. This may seem strange for math operators but it is actually quite familiar for other functions. For example C uses prefix notation for function calls. e.g. add(1, 2). But unlike C we don't need commas or parentheses since all functions have a fixed arity.

Simple Mode

If you play around with those examples, you'll find that +1 2 doesn't actually work! This obviously makes no sense yet, but the reason is that Nibbles tries to only have one way to do things in the binary form, other ways are remapped to something else, and these are called extensions and are covered later. But the whole purpose of the literate form is to make it easy and not error prone to write code, so we error rather than do what you didn't mean. In this case ]1 2 would generate the same binary code that +1 2 would have and it means max. You should be protected from accidentally doing the wrong thing, but this could be quite annoying so I suggest using nibbles with nibbles -simple to disable all extensions and implicit operations until you have mastered the basics.

Exercise

Write a program that computes (1+2)*(3-4)

Solution
*- 3 4+2 1

You may be distraught about those spaces, but they don't affect binary size. The first space is to stop -3 from being interpreted as negative 3.


Other Types

You've seen integers, other data types are chars and lists. Chars behave like integers in most cases but are displayed differently (the ops table linked below makes this more precise). They can be created with single quotes, e.g. ' ' to create a space char.

Lists can be created with the : (append) operator. However unlike Haskell, : coerces its operands, turns non lists into singleton lists and then concatenates them. For example :1 :2 3[1,2,3]. Note that this bracket/comma notation is just for output display purposes and not valid Nibbles code.

Strings are actually just a list of chars, and can be created using double quotes. Escapes are valid for strings and chars in the same style that Haskell uses. E.g.

"Hi\nthere"
Hi there

A list of strings can be created by listing multiple strings together without spaces between them, e.g. "hi""there"["hi","there"]

A note on example format in this tutorial

Ops

Here is a table of the basic Nibbles ops. There is a more complete quick reference, but for now it would have some confusing things. A couple things to note:

Exercise

Use the ops table to write a program which outputs your name as a square, e.g.

Darren Darren Darren Darren Darren Darren

Hint: Since we haven't introduced assignments yet, it is ok to hard code the length.

Solution
^6:"Darren" '\n'

That newline could have been included in the string but I'm just showing off how not to code golf.


Functions

There were a couple things in that table we haven't seen yet, functions and args. For example map's (.) second argument is a function and the first is the list it will map over.

Functions do not require any syntax, some ops take a function argument and automatically treat that operand as a function. Any expression can be treated as a function, for example if +1 2 is to be treated as a function with 1 argument, then it is simply a function which ignores its argument and returns 3!

Args are referenced by number with the following identifiers: $ @ _. If you are nesting functions then these count upwards in scope. For example in an inner function of 1 argument, $ would be its argument as usual, but @ would be the first argument of the outer function. This is also known as DeBruijn indices.

Let's look at the examples from the ops table for map and foldr1.

."abc"+1$"bcd"

This corresponds to the Haskell code:

flip map "abc" (\a -> chr ((+) 1 (ord a)))

The chr and ord are implicit because Nibbles supports + on chars. Let's imagine the map function just took the list first and also supported math ops on chars then the code could have been:

map "abc" (\a -> (+) 1 a)

Which is more obviously related to the Nibbles code.

And for foldr1:

/,3+@$6

This corresponds to the Haskell code:

flip foldr1 [1..3] (\elem accum -> (+) elem accum)

Note: If you need a DeBruijn index > 3, then preceding an identifier with a ; adds 3 for each ;. E.g. ;;@ is DeBruijn index 8.

Tip: Using DeBruijn indices on more complicated programs can become extremely tedious, so Nibbles also supports explicit lambdas with identifiers in the literate form. Just put a \ followed by the names of the identifiers you wish to use. This will be automatically converted to DeBruijns during conversion to binary. For example this program will generate the same binary code as the earlier example:

/,3 \element accum +accum element
6

Exercise

Compute the product of all even numbers less than 50 (the answer is 10409396852733332453861621760000). And yes this number is > 264 but Nibbles uses arbitrary precision, so don't worry about that.

Hint1: the range operator (,) is 1 indexed, which may seem bad, but, as it turns out in code golf, is much more commonly used than lists starting from 0. So range and all other list operators in Nibbles are 1 indexed.

Hint2: the filter operator (|) uses the truthiness of the function result. Integers are considered truthy if they are > 0 (and false if 0 or negative, this is nonstandard but useful since there are no comparison operators). Lists are truthy if non empty. Chars are truthy if they are non whitespace (and >0).

Solution
/|,49%+1$2*@$

Or more verbosely:

/ # flip foldr1 | # (flip filter ,49 # [1..49] %+1$2 # (\a -> (mod) ((+) 1 a) 2 > 0) *@$ # ) (\a b-> (*) a b)

Finally we are seeing nice looking programs, can you do better? We will learn more ways to shorten this.

Those #'s act like typical scripting language comments.


Input

All programs actually start off with args available for use, that read from stdin.

You may also refer to these by name (fstInt fstLine ints sndInt sndLine allInput intMatrix allLines).

For example, if the input is a list of integers we could find the sum as so:

> echo "+_" > intro.nbl > echo 1 2 3 | nibbles intro.nbl
6

Keep in mind that these are DeBruijn indices too! So after you start using functions they will shift. It can be hard to keep track of what DeBruijn index corresponds to what, so you can always use ct anywhere to see context types and info.

Arg Input

You may also pass inputs to your program as command line args and they will be parsed as if they are Haskell values (tuples, lists, strings, chars, numbers are ok). These will take up the lower DeBruijn indices than the regular input variables if present. For example:

> echo 5 | nibbles filename.nbl 6 [(7,'c'),(8,'d')]

Will assign 6 to $, [(7,'c'),(8,'d')] to @ and 5 to _.

Note that (with z shell and likely others) you would need to \ backquote all ( ) " ' symbols (not shown in this example for clarity).

A caveat is that if you wish to use args, you must pass these args when generating the Haskell code that executes your program. This is because arg types must be known when parsing nibbles code! You don't have to worry about that if you just run your programs through nibbles instead of to out.hs first.

This is super handy if the problem allows it as you won't need to do any parsing even for complex input types.

Exercise

Input a number n, then repeat the second line n times.

Solution
^$;;@

Were you expecting it to be harder than that?