Elliptical Python Programming

TIL that in python, 1--2==3

For those who are curious, `...` is a placeholder value in Python called Ellipsis. I don't believe it serves any real purpose other than being a placeholder. But it is an object and it implements `__eq__`, and is considered equal to itself. So `...==...` evaluates to `True`. When you prefix a `True` with `-`, it is interpreted as a prefix negation operator and implicitly converts the `True` to a `1`, so `-(...==...)` is equal to `-1`. Then, you add another prefix `-` to turn the `-1` back into `1`.

`--(...==...)--(...==...)` evaluates to `2` because the first block evaluates to 1, as previously mentioned, and then the next `-` is interpreted as an infix subtraction operator. The second `-(...==...)` evaluates to `-1`, so you get `1 - -1` or `2`.

When chaining multiple together, you can leave off the initial `--`, because booleans will be implicitly converted to integers if inserted into an arithmetic expression, e.g. `True - -1` -> `1 - -1` -> `2`.

> There should be one-- and preferably only one --obvious way to do it.

This article is obviously completely tongue-in-cheek, but I feel the need to point out that this sentence is not meant to be a complete inversion of the Perl philosophy of TIMTOWTDI. The word "obvious" is crucial here - there can be more than one way, but ideally only one of the ways is obvious.

It's not a python thing. 1-(-2), distribute the negative.

In most C-like languages that would be a syntax error. E.g. in C and C++ as a rule you tokenize "greedily", "1--2" would be tokenized as "1", "unary decrement operator", "2", which is illegal because you're trying to decerment an rvalue.

Python doesn't have "--", which allows the tokenizer to do something else.

Numpy actively uses … to make slicing multidimensional arrays less verbose. There are also uses in FastAPI along the lines of «go with the default».

Ok do this but for JavaScript

If you're curious, the code in ellipsis results in executing:

    print('hello, world')

I think we're really starting to over crowd pythons syntax and I'm not a fan.

excellent explanation, to add to this since I was curious about the composition, '%c' is an integer presentation type that tells python to format numbers as their corresponding unicode characters[1] so

'%c' * (length_of_string_to_format) % (number, number, ..., length_of_string_to_format_numbers_later)

is the expression being evaluated here after you collapse all of the 1s + math formatting each number in the tuple as a unicode char for each '%c' escape in the string corresponding to its place in the tuple.

[1] https://docs.python.org/3/library/string.html#format-specifi...

In C, that is really because Unary minus (negation) has precedence over binary operations.

    +a - b; // equivalent to (+a) - b, NOT +(a - b)
    -c + d; // equivalent to (-c) + d, NOT -(c + d)

https://en.cppreference.com/w/cpp/language/operator_arithmet...

    +-e; // equivalent to +(-e), the unary + is a no-op if “e” is a built-in type
     // because any possible promotion is performed during negation already

The same doesn't apply to, !! Which is applied as iterated binary operations (IIRC)

I am pretty sure the decriment operator came around well after that quirk was established.

Thank you!

I noticed some ** and * in the thing sent to eval(), which (given that the building blocks are small integers) seemed related to prime factorizations.

The initial %c is duplicated 21 times (3*7, if I read correctly), and then string-interpolated (%c%c%c...) against a long tuple of integers. These integers themselves are composed of products of factors combined using * and **.

There is also one tuple "multiplication" embedded within that long tuple of integers -- (a,b)*2 = (a,b,a,b). That is for the 'l' 'l' in "hello".

It's all very clever and amusingly mathy, with a winking allusion to the construction of natural numbers using sets. It made me Godel.

you're telling me you never program in python elliptically??

This behavior can be replicated with any class that has two special methods: __neg__ that returns -1 and __sub__ that accepts ints and returns 1-other.

For example if you make this class:

  class _:
       def __neg__(self):
           return -1
       def __sub__(self, other):
           return 1-other

You get similar behavior:

  >>> --_()
  1
  >>> _()--_()
  2

Fun python for everyone.

Peter van der Linden’s book “Expert C Programming” (which is awesome btw) says that one of them (Kernighan, Richie or maybe Ken Thompson I forget) realised early on that the c compiler had the wrong operator precedence for bit twiddling and unary and boolean operators but “at that stage we had a few thousand lines of C code and thought it would be too disruptive to change it”

You can do this on JavaScript too.

  alert(1)
  // equals to:
  [][(![]+[])[+!+[]]+(!![]+[])[+[]]][([][(![]+[])[+!+[]]+(!![]+[])[+[]]]+[])[!+[]+!+[]+!+[]]+(!![]+[][(![]+[])[+!+[]]+(!![]+[])[+[]]])[+!+[]+[+[]]]+([][[]]+[])[+!+[]]+(![]+[])[!+[]+!+[]+!+[]]+(!![]+[])[+[]]+(!![]+[])[+!+[]]+([][[]]+[])[+[]]+([][(![]+[])[+!+[]]+(!![]+[])[+[]]]+[])[!+[]+!+[]+!+[]]+(!![]+[])[+[]]+(!![]+[][(![]+[])[+!+[]]+(!![]+[])[+[]]])[+!+[]+[+[]]]+(!![]+[])[+!+[]]]((![]+[])[+!+[]]+(![]+[])[!+[]+!+[]]+(!![]+[])[!+[]+!+[]+!+[]]+(!![]+[])[+!+[]]+(!![]+[])[+[]]+([][(![]+[])[+!+[]]+(!![]+[])[+[]]]+[])[+!+[]+[+!+[]]]+[+!+[]]+([]+[]+[][(![]+[])[+!+[]]+(!![]+[])[+[]]])[+!+[]+[!+[]+!+[]]])()

https://jsfuck.com/

Also worth noting that `1 - -2` works and produces 3 in C because the space breaks the operator.

Expanding on this a little, I will be replacing all occurrences of 2 with two blobs fighting, with shields:

    >>> 0^((...==...)--++--(...==...))^0
    2

>> There should be one-- and preferably only one --obvious way to do it.

Except for package management, of course. There, we need lots and lots of ways.

Pretty sure this would have been possible in Python 2.6. The Ellipsis object has been around for a very long time.

And apparently string formatting which should have an ever growing number of ways to handle it. :shrug: