update for linear #40
Conversation
replace R2 with R2 n, suitable constraints on n
| dmod :: RealFrac a => a -> a -> a | ||
| -- | Interpolate linearly between two values. The first argument is | ||
| -- the parameter. A parameter of @0@ results in the second argument; | ||
| -- with a parameter of @1@, @lerp'@ returns its third argument. |
There was a problem hiding this comment.
I think you've got these the wrong way round :)
There was a problem hiding this comment.
Thanks for catching that!
|
I was actually working on this too :). nvm, it's nice to compare. I've also converted |
|
I should have checked with you before I started. Sorry. I'd like to work on a Backend or two. What have you not started on? I just gave you push rights to more repos, including |
|
I'd like to linearize at least one backend as well. Let me know whats left. |
|
Great, thanks. So far I've worked on |
|
@bergey why dont you take -cairo or -canvas, your choice and i'll take the other. |
|
@cchalmers I gave you push rights to diagrams-rasterific |
|
@jeffreyrosenbluth I'll take |
|
@bergey sounds good. will do |
Follows the argument order in vector-space.
More type class constraints in Logo
In liftF & liftF2, `a` and `n` are distinct type variables, since `f` must operate both on `n ` and `Complex n`
| @@ -101,8 +99,8 @@ data Q236 = Q236 Rational Rational Rational Rational | |||
| deriving (Eq, Ord, Show, Read) | |||
There was a problem hiding this comment.
This Ord instance is wrong? It's used for Grid but I don't fully understand it. A proper Ord instance could be made by comparing Doubles:
instance Ord Q236 where
a <= b = toDouble a <= toDouble b
Same trick could be done for signum.
There was a problem hiding this comment.
That Ord instance has been there longer than I've been involved, so I hadn't really thought about it. I agree, it doesn't do what I'd expect (though it's adequate for Map and the like). Does Grid care which behavior is used?
There was a problem hiding this comment.
How does Grid use this? I don't see it. The Ord instance for Q236 was just intended for using them as keys in maps etc. I'd really rather not use toDouble a <= toDouble b since the whole point of using the Q236 representation is to avoid having to do floating-point comparisons.
There was a problem hiding this comment.
I ment Edge, specifically via mkEdgesorry. I haven't gone thought the code so I don't know why it puts it in order. Nvm I see now it's just to have an Eq instance for Polygon.
There was a problem hiding this comment.
It looks to me like all the uses of Ord just treat it as a canonical order, and don't assume that Q236 is an ordered field. I think it's worth adding a comment that this is deliberate, but we should leave the code alone.
|
@cchalmers Thanks for making nearly all the changes I planned to make today! |
Migrate from `vector-space` package to `linear` Make types more polymorphic.
Accompanies diagrams/diagrams-lib#212
This is nearly done; it remains to decide what to do about
force-layout, and the minor name collision ofsize. I think the choices withforce-layoutare port it to linear (it has no revdeps butdiagrams-contrib) or createvector-spaceclass instances forV2and other types, indiagrams-contrib.