Various new theorems in univalent mathematics written in Agda ------------------------------------------------------------- Martin Escardo 2010--2020--∞, continuously evolving. https://www.cs.bham.ac.uk/~mhe/ https://github.com/martinescardo/TypeTopology The main new results are about compact types, totally separated types, compact ordinals and injective types, but there are many other things (see the clickable index below). * Our main use of this development is as a personal blackboard or notepad for our research. In particular, some modules have better and better results or approaches, as time progresses, with the significant steps kept, and with failed ideas and calculations eventually erased. * We offer this page as a preliminary announcement of results to be submitted for publication, of the kind we would get when we visit a mathematician's office. * We have also used this development for learning other people's results, and so some previously known constructions and theorems are included (sometimes with embellishments). In our last count, this development has 40000 lines, including comments and blank lines. * The required material on HoTT/UF has been developed on demand over the years to fullfil the needs of the above as they arise, and hence is somewhat chaotic. It will continue to expand as the need arises. Its form is the result of evolution rather than intelligent design (paraphrasing Linus Torvalds). Our lecture notes develop HoTT/UF in Agda in a more principled way, and offers better approaches to some constructions and simpler proofs of some (previously) difficult theorems. (https://www.cs.bham.ac.uk/~mhe/HoTT-UF-in-Agda-Lecture-Notes/) Our philosophy, here and in the lecture notes, is to work with a minimal Martin-Löf type theory, and use principles from HoTT/UF (existence of propositional truncations, function extensionality, propositional extensionality, univalence, propositional resizing) and classical mathematics (excluded middle, choice, LPO, WLPO) as explicit assumptions for the theorems, or for the modules, that require them. As a consequence, we are able to tell very precisely which assumptions of HoTT/UF and classical mathematics, if any, we have used for each construction, theorem or set of results. We also work, deliberately, with a minimal subset of Agda. * There is also a module that links some "unsafe" modules that use type theory beyond MLTT and HoTT/UF, which cannot be included in this safe-modules index: The system with type-in-type is inconsistent (as is well known), countable Tychonoff, and compactness of the Cantor type using countable Tychonoff. (https://www.cs.bham.ac.uk/~mhe/TypeTopology/UnsafeModulesIndex.html) * A module dependency graph is available, updated manually from time to time. (https://www.cs.bham.ac.uk/~mhe/TypeTopology/dependency-graph.pdf) * There are some somewhat obsolete comments at the end of this file, explaining part of what we do in this development. See instead the comments in the various modules. * This has been tested with Agda 2.6.1. Click at the imported module names to navigate to them: \begin{code} {-# OPTIONS --without-K --exact-split --safe #-} module index where \end{code} Some of the main modules and module clusters: \begin{code} import Compactness import TotalSeparatedness import InjectiveTypes-article import TheTopologyOfTheUniverse import RicesTheoremForTheUniverse import Ordinals import LawvereFPT import PartialElements import UF import Types2019 import MGS -- Modular version of https://github.com/martinescardo/HoTT-UF-Agda-Lecture-Notes import PCFModules -- by Tom de Jong import Dyadics -- by Tom de Jong \end{code} The UF module (univalent foundations) has been developed, on demand, for use in the preceding modules (and the modules below, too). The modules UF-Yoneda and UF-IdEmbedding contain new results. All modules in alphabetical order: \begin{code} import ADecidableQuantificationOverTheNaturals import AlternativePlus import ArithmeticViaEquivalence import BasicDiscontinuityTaboo import BinaryNaturals import CantorSchroederBernstein import CantorSchroederBernstein-TheoryLabLunch import Codistance import Compactness import CompactTypes import CoNaturalsArithmetic import CoNaturalsExercise import CoNaturals import ConvergentSequenceCompact import ConvergentSequenceInfCompact import DecidabilityOfNonContinuity import DecidableAndDetachable import DiscreteAndSeparated import Dyadic import DyadicOrder import Dyadics import Dominance import DummettDisjunction import Empty import Escardo-Simpson-LICS2001 import ExtendedSumCompact import Fin import frame import FailureOfTotalSeparatedness import GeneralNotation import GenericConvergentSequence import HiggsInvolutionTheorem import Id import InfCompact import InitialBinarySystem -- More work than needed! import InitialBinarySystem2 -- No need to work with subtype of normal elements. import InjectiveTypes-article import InjectiveTypes import LawvereFPT import LeftOvers import LexicographicCompactness import LexicographicOrder import LiftingAlgebras import LiftingEmbeddingDirectly import LiftingEmbeddingViaSIP import LiftingIdentityViaSIP import Lifting import LiftingMonad import LiftingMonadVariation import LiftingSize import LiftingUnivalentPrecategory import LPO import Lumsdaine import NaturalNumbers import NaturalNumbers-Properties import NaturalsAddition import NaturalsOrder import Negation import NonCollapsibleFamily import OrdinalArithmetic import OrdinalCodes import OrdinalNotationInterpretation import OrdinalNotions import OrdinalOfOrdinals import OrdinalOfTruthValues import OrdinalsClosure import Ordinals import OrdinalsShulmanTaboo import OrdinalsType import OrdinalsWellOrderArithmetic import PartialElements import Pi import Plus import PlusOneLC import Plus-Properties import PropInfTychonoff import PropTychonoff import QuasiDecidable import RicesTheoremForTheUniverse import RootsTruncation import Sequence import Sigma import SimpleTypes import SliceAlgebras import SliceEmbedding import SliceIdentityViaSIP import Slice import SliceMonad import SpartanMLTT import SquashedCantor import SquashedSum import Swap import sigma-sup-lattice import sigma-frame import TheTopologyOfTheUniverse import TotallySeparated import Two import Two-Prop-Density import Two-Properties import UnivalenceFromScratch import Unit import Unit-Properties import Universes import WeaklyCompactTypes import W import WLPO import UF import UF-Base import UF-Choice import UF-Classifiers import UF-Classifiers-Old import UF-Connected import UF-Embeddings import UF-EquivalenceExamples import UF-Equiv-FunExt import UF-Equiv import UF-ExcludedMiddle import UF-Factorial import UF-FunExt-from-Naive-FunExt-alternate import UF-FunExt-from-Naive-FunExt import UF-FunExt import UF-hlevels import UF-IdEmbedding import UF-ImageAndSurjection import UF-Knapp-UA import UF-KrausLemma import UF-LeftCancellable import UF-Miscelanea import UF-Powerset import UF-PropIndexedPiSigma import UF-PropTrunc import UF-Quotient import UF-Retracts-FunExt import UF-Retracts import UF-Size import UF-StructureIdentityPrinciple -- Old, probably delete. import UF-SIP -- New, better, version. import UF-SIP-Examples import UF-SIP-IntervalObject import UF-Subsingletons-FunExt import UF-Subsingletons import UF-UA-FunExt import UF-Univalence import UF-UniverseEmbedding import UF-Yoneda \end{code} Old blurb. I want to completely rewrite this eventually, and update it, as it is very old. However, the linked files already have up-to-date information within them. September 2017. This version removes the module CurryHoward, so that we use the symbols Σ and + rather than ∃ and ∨. This is to be compatible with univalent logic. We also make our development more compatible with the philosophy of univalent mathematics and try to streamline it a bit. The original version remains at http://www.cs.bham.ac.uk/~mhe/agda/ for the record and to avoid broken incoming links. December 2017. This version includes many new things, including a characterization of the injective types, which relies on the fact that the identity-type former Id_X : X → (X → U) is an embedding in the sense of univalent mathematics. January 2018. This again includes many new things, including 𝟚-compactness, totally separated reflection, and how the notion of 𝟚-compactness interacts with discreteness, total separatedess and function types. We apply these results to simple types. April 2018. The univalence foundations library was monolotic before. Now it it has been modularized. We extended the Yoneda-Lemma file with new results. 29 June 2018. The work on compact ordinals is essentially complete. Some routine bells and whistles are missing. 20 July 2018. Completed the proof that the compact ordinals are retracts of the Cantor space and hence totally separated. This required work on several modules, and in particular the new module SquashedCantor. You can navigate this set of files by clicking at words or symbols to get to their definitions. The module dependency graph: http://www.cs.bham.ac.uk/~mhe/TypeTopology/manual.pdf The following module investigates the notion of compact set. A set X is compact iff (p : X → 𝟚) → (Σ x ꞉ X , p x ≡ ₀) + Π x ꞉ X , p x ≡ ₁ The compactness of ℕ is a contructive taboo, known as LPO, which is an undecided proposition in our type theory. Nevertheless, we can show that the function type LPO→ℕ is compact: \begin{code} import LPO \end{code} See also: \begin{code} import WLPO \end{code} An example of an compact set is ℕ∞, which intuitively (and under classical logic) is ℕ ∪ { ∞ }, defined in the following module: \begin{code} import GenericConvergentSequence \end{code} But it is more direct to show that ℕ∞ is compact, and get compactness as a corollary: \begin{code} import CompactTypes import ConvergentSequenceCompact \end{code} An interesting consequence of the compactness of ℕ∞ is that the following property, an instance of WLPO, holds constructively: (p : ℕ∞ → 𝟚) → ((n : ℕ) → p(under n) ≡ ₁) + ¬((n : ℕ) → p(under n) ≡ ₁). where under : ℕ → ℕ∞ is the embedding. (The name for the embedding comes from the fact that in published papers we used an underlined symbol n to denote the copy of n : ℕ in ℕ∞.) \begin{code} import ADecidableQuantificationOverTheNaturals \end{code} This is used to show that the non-continuity of a function ℕ∞ → ℕ is decidable: \begin{code} import DecidabilityOfNonContinuity \end{code} Another example of compact set is the type of univalent propositions (proved in the above module Compact). Given countably many compact sets, one can take the disjoint sum with a limit point at infinity, and this is again a compact sets. This construction is called the squashed sum of the countable family compact sets. It can be transfinitely iterated to produce increasingly complex compact ordinals. \begin{code} import SquashedSum import OrdinalNotationInterpretation import LexicographicCompactness import ConvergentSequenceInfCompact \end{code} As a side remark, the following module characterizes ℕ∞ as the final coalgebra of the functor 1+(-), and is followed by an illustrative example: \begin{code} import CoNaturals import CoNaturalsExercise \end{code} The following module discusses in what sense ℕ∞ is the generic convergent sequence, and proves that the universe U of types is indiscrete, with a certain Rice's Theorem for the universe U as a corollary: \begin{code} import TheTopologyOfTheUniverse import RicesTheoremForTheUniverse \end{code} The following two rogue modules depart from our main philosophy of working strictly within ML type theory with the propositional axiom of extensionality. They disable the termination checker, for the reasons explained in the first module. But to make our point, we also include runnable experiments in the second module: \begin{code} -- import CountableTychonoff --unsafe (see above) -- import CantorCompact --unsafe (see above)` \end{code} The following modules return to the well-behavedness paradigm. The first one shows that a basic form of discontinuity is a taboo. This, in fact, is used to formulate and prove Rice's Theorem mentioned above: \begin{code} import BasicDiscontinuityTaboo \end{code} The following shows that universes are injective, and more generally that the injective types are the retracts of exponential powers of universes: \begin{code} import InjectiveTypes \end{code} This uses properties of products indexed by univalent propositions, first that it is isomorphic to any of its factors: \begin{code} import UF-PropIndexedPiSigma \end{code} And, more subtly, that a product of compact sets indexed by a univalent proposition is itself compact: \begin{code} import PropTychonoff \end{code} And finally that the map Id {X} : X → (X → U) is an embedding in the sense of univalent mathematics, where Id is the identity type former: \begin{code} import UF-IdEmbedding \end{code} The following generalizes the squashed sum, with a simple construction and proof, using the injectivity of the universe and the Prop-Tychonoff theorem: \begin{code} import ExtendedSumCompact \end{code} The following modules define 𝟚-compactness and study its interaction with discreteness and total separatedness, with applications to simple types. We get properties that resemble those of the model of Kleene-Kreisel continuous functionals of simple types, with some new results. \begin{code} import TotallySeparated import CompactTypes import SimpleTypes import FailureOfTotalSeparatedness \end{code}