-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathTODO
89 lines (63 loc) · 2.9 KB
/
TODO
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
= Current stack
== Finish tensors
- instance SrN Monoidal...
- Propagate instances for Sr2:
- Either move monadic (nested product) combinators to a subclass (refactoring
headache with low apparent value right now), or shove 'Sr2' into its own
module as it is, because 'SrN' ought to do everything 'Sr2' can.
= Anytime dead code cleanup/maintenance
- Cat.Orthotope
- Cat.Unsized.Category.Free.Data - cleanup @foldMap@, @fixed@
- Cat.Sized.Semigroupoid.Free.Data - cleanup @foldMap@
- Cat.Sized.Category.Free.Data - cleanup @foldMap@
- Cat.Sized.Monoidal.Free.Data - cleanup @foldMap@
- Cat.Sized.Braided
- Cat.Sized.Diagonal
= Foundations / high-value experiments
1. Sort out what the relevant constraints on foldMap should be by comparing
- Cat.Unsized.Semigroupoid.Free.Data
- Cat.Unsized.Category.Free.Data
2. Bimonoidal
- Revisit Cat.Sized.Circuit with Cat.Sized.Cartesian.Free
- Could you actually use Codiagonal ± Semicocartesian methods?
- DistributiveLR instances
- R1/VS.Vector
- Maaaaybe Free Semiadditive (unless in the course of writing it you discover
Free Additive would be a better use of time?)
3. Sketch a profunctor-based typeclass hierarchy in a new branch when your
examples include biproducts and enough of the basic functionality with
biproducts (e.g. recursion schemes) seems to be working.
4. Feature integration / testing:
- Boolean circuits as tensors over a semiring
- Consider a numhask reimplementation
- Recursion scheme over circuits/products of constant boolean morphisms for
normal forms?
- Do you need/want different types (and perhaps a different,
not-just-an-example module) for this?
- Incremental computation ± traces?
- Check refs to see what you need to take a beeline to encoding SAT problems
- Boolean relations
- ((Self-dual?) compact) closed categories (/dagger categories?)
- Traces
- Zusatz: ersatz terms as a DSL.
= Maintenance that might pay dividends
- Refactor: Because of module coupling (currently manifesting with instance
derivations and obnoxious constraints), move '{Unsized/Sized}.HFunctor' into
a module separate from '_.Functor'.
- Expected payoff might increase with the volume/complexity of future code
that depends on recursion schemes.
= Low priority
- comonad instances for R1 and VS.Vector/TODO stubs
- non-empty leftmost/rightmost/center comonad for rank ≥ 1 — all defined by a lens that points to a *location* independent of contents
— instance for any functor over a monoid
— "search"/"selection" that points to a first/last value satisfying a predicate (or a default)?
- pointer/finitary representable store comonad
- tangent bundle/finitary discrete derivative comonad
== Infrastructure
- repository CI / fix cachix setup
- Multi-cradle project:
- cat-herder
- cat-herder-test
- cat-herder-base
- cat-herder-vector-sized
- cat-herder-orthotope