Interaction Trees
A Library for Representing Recursive and Impure Programs in Coq
Introduction
For a quick overview of the core features of the library, see
examples/ReadmeExample.v
.
See also the tutorial.
The coqdoc documentation for this library is available here.
Top-level modules
ITree.ITree
: Definitions to program with interaction trees.ITree.ITreeFacts
: Theorems to reason about interaction trees.ITree.Events
: Some standard event types.
Installation
Via opam
opam install coq-itree
Dependencies
See coq-itree.opam
for version details.
Axioms
This library currently depends on UIP, functional extensionality, excluded middle,
and choice; see also theories/Axioms.v
.
UIP
This library depends on UIP for the inversion lemma:
Lemma eqit_inv_Vis
: eutt eq (Vis e k1) (Vis e k2) ->
forall x, eutt eq (k1 x) (k2 x).
There are a few more lemmas that depend on it, but you might not actually need
it. For example, the compiler proof in tutorial
doesn't need it and is
axiom-free.
That lemma also has a weaker, but axiom-free version using heterogeneous
equality: eqit_inv_Vis_weak
.
The axiom that's technically used here is eq_rect_eq
(and also JMeq_eq
in
old versions of Coq), which is equivalent to UIP.
Functional extensionality
The closed category of functions assumes functional_extensionality
,
in Basics.FunctionFacts.CartesianClosed_Fun
.
Excluded middle and choice
The theory of traces (theories/ITrace/
)—which Dijkstra monads for ITree
depend on (theories/Dijkstra
)—assumes excluded middle, to decide whether an
itree diverges, and a type-theoretic axiom of choice, which provides a strong
excluded middle in propositional contexts:
Theorem classicT_inhabited : inhabited (forall T : Type, T + (T -> False)).
Remark: excluded middle implies UIP, but we still consider UIP as a separate dependency because it's used in a more central part of the library.
Exported: strong bisimulation is propositional equality
The library exports the following axiom for convenience, though it's unlikely you'll need it, and the rest of the library does not depend on it:
Axiom bisimulation_is_eq : t1 ≅ t2 -> t1 = t2.
Contributions welcome
Feel free to open an issue or a pull request!
See also DEV.md
for working on this library.