Date: 18-2-2020
Status: draft specification
This document attempts to describe a interchange format for morphisms of monoidal categories.
Categories are algebraic structures that allow you to "compose" maps (also called morphisms).
A map is a thing with a certain source and target type (called domain and co-domain). This can be graphically depicted as a box with input and output wires.
Composing a map is only possible if the target of one map is the source of the other. Graphically:
Monoidal categories (also called tensor categories) are categories with an additional way of combining morphisms, called the tensor product. Instead of composing in sequence we compose "in parallel". Graphically this amounts to putting boxes next to eachother.
Which itself can be interpreted as a box with two in and two outgoing wires.
There is no meaning assigned to the boxes!
This is on purpose. By seperating the semantics from the syntax, monoidal categories become a useful language to describe many different systems, all of them using the same base language of boxes and wires.
There are many different flavours of monoidal category (syntax): braided, symmetric, etc.
We don't attempt to capture each such category out there, but this draft specification is an attempt to address some of the most commonly used monoidal categories.
Consider this string diagram.
This could be represented as a term,
f ; (h * g)
(; is actually called "compose")
In this format, that would be represented as.
{
"type": "compose",
"terms": [{
"type": "generator",
"outputTypes": ["b", "c"],
"name": "f",
"inputTypes": ["a"]
}, {
"type": "tensor",
"terms": [{
"type": "generator",
"outputTypes": ["e"],
"name": "h",
"inputTypes": ["b"]
}, {
"type": "generator",
"outputTypes": ["f"],
"name": "g",
"inputTypes": ["c"]
}]
}]
}We have compose and tensor. Compose works in semicolon order.
The terms[] array should contain at least two terms.
A box has a name (can be any unicode string, also emoji's of course), domain and codomain.
{
"type": "generator",
"outputTypes": ["c", "d", "e"],
"name": "😎",
"inputTypes": ["a", "b"]
}- TODO
- TODO
In certain cases the is a unique morphism from X^n -> X^m, which is rendered as a dot with n in and m outgoing "legs".
{
"typeParam": "α",
"type": "spider",
"outputs": 5,
"name": "g",
"inputs": 4,
"color": "white"
}{
"typeParam":"α*",
"type":"cup",
"name":"("
}- TODO
root ::= term
term ::= unit | tensor | identity | compose | generator | permutation | spider | cup | cap
type ::= string
unit ::=
{ "type": "unit"
, "inputTypes": type[]
, "outputTypes": type[]
}
tensor ::=
{ "type": "tensor"
, "terms": term[]
, "inputTypes": type[]
, "outputTypes": type[]
}
identity ::=
{ "type": "identity"
, "inputTypes": type[]
, "outputTypes": type[]
}
compose ::=
{ "type": "compose"
, "terms": term[]
, "inputTypes": type[]
, "outputTypes": type[]
}
generator ::=
{ "type": "generator"
, "name": string
, "inputTypes": type[]
, "outputTypes": type[]
}
permutation ::=
{ "type": "permutation"
, "name": string,
, "permutation": number[]
, "inputTypes": type[]
, "outputTypes": type[]
}
spider ::=
{ "type": "spider"
, "name": string
, "inputs": number
, "outputs": number
, "color": "white" | "black"
, "typeParam": type
, "inputTypes": type[]
, "outputTypes": type[]
}
cup ::=
{ "type": "cup"
, "name": string
, "typeParam": type
, "inputTypes": type[]
, "outputTypes": type[]
}
cap ::=
{ "type": "cap"
, "name": string
, "typeParam": type
, "inputTypes": type[]
, "outputTypes": type[]
}