Physics on the other hand very much has types, the unit to type correspondence is dopetastic.
Lately there's been a resurgence in foundations because of what new foundations like Homotopy Type Theory bring to the table. Funny how something as basic as equality can be revisited time and time again for thousands of years!
Was this decision ever explicitly expressed and agreed on by mathematicians at the time, or was it just implicitly adopted?
tl;dr: types are propositions and values are proofs of those propositions. Constructive logic is not concerned with (true xor false), but rather "I can prove A" or "I can prove not-A". There's also the middle ground of "I don't know how to prove anything about A".
https://en.wikipedia.org/wiki/Curry%E2%80%93Howard_correspon...
But the connections don't really end there.
But I preffer to keep that stuff in, even when doing pretty abstract stuff. Essentially because I wanted the type annotations. For example my favorite convention for the Minkowski metric is:
[c^2 0 0 0]
[ 0 -1 0 0]
[ 0 0 -1 0]
[ 0 0 0 -1]
That way your vectors are just $(t, x, y, z)$, where $t$ has units of time and the rest of space. Your covectors do come out oddly as $(c^2t, -x, -y, -z)$, but that's what makes them covectors.Another reason is that people instinctively feel that the elements of the coordinate tuple all have to have the same units. But that's not necessary when you have a metric around. Indeed keeping irregular units is helpful because it prevents you from accidentally forgetting to multiply through by the metric.
Mathematics is often expressed in language, That language us usually a mix of a human natural language English/French/German/etc plus symbolic extensions, and that is very poorly typed.
In the recent failed proof of P!=NP, the flaw was a type error:
https://cstheory.stackexchange.com/questions/38803/is-norber...
Type errors are rather common in math proofs that end up retracted.
But it's hard to call them type errors, because it's not like a computer program where the typechecker detected the collision or failed in some way: the proofs make very little mention of types at all, and it's up to the readers to apply a type-system from their own heads.
The link you gave does not support this statement.
For example suppose a geometer explains something to another geometer just by drawing diagrams with no verbal or textual annotation. You can plausibly claim that "geometry diagrams" are a language, but now you are defining language so broadly that the term is almost useless.
Google, define language:
lan·guage ˈlaNGɡwij/Submit noun 1. the method of human communication, either spoken or written, consisting of the use of words in a structured and conventional way. "a study of the way children learn language" 2. the system of communication used by a particular community or country. "the book was translated into twenty-five languages" synonyms: tongue, mother tongue, native tongue; More
I would say "geometry diagrams" are a language, by the second definition.
Mathematics is largely based on set theory, so every element has a set it belongs to, and every function has a source-set (domain) and a target-set (co-domain). Mathematicians are very careful to always qualify the encompassing sets, domains and co-domains. You would always write:
* Let x \in X be a point in Space. * Let f:A->B be a fuction with ceratin properties.
This is even more pronounced in Category theory, which invaded algebra and geometry quite heavily. Where the only thing you know about a function is it's domain, codomain and how it composes.
So, yes. Mathematics on a practical level is _very_ type aware.
And it also prevents you from doing stupid stuff: f:A->B, g:A->C, can't be composed, so f(g(x)) is a "type error".
This is generally on purpose, to avoid making explicit things that are obvious, but as someone developing and using statically typed languages, reading Math papers is often quite frustrating.
As others have mentioned, Mathematics isn't really a language. It's at the very least absolutely certainly not one language. Certain parts of Mathematics use a typed language. Certain parts rely on other kinds of containment properties which serve nicely as types.
Finally, generally, math is very interested in being completely clear the properties of objects and very interested in speaking abstractly, so even if the language of use is untyped the logic is often very rigorous around typing.
Above it all, perhaps, is the idea that mathematics is concerned with proof and proofs may contain reference to the properties being discussed---metamathematical ideas pervade normal mathematical thought. This means that even absent of types the content of a mathematical utterance is often going to explicitly posit, discuss, and prove properties similar to type propositions.
No, but static type hinting is strongly encouraged. If I say in my paper that "x" is a vector, if I later use "x" to mean a scalar it would cause tremendous cognitive dissonance and confusion for the reader. There is no compiler error, in that the information can still be read and understood. It is just much more difficult.
Mathematical notation is also aggressively polymorphic. The expression "f x" can have wildly different meanings depending on what "f" and "x" are.