WIP: Combination of #1 and #2 #10

Ericson2314 · 2020-01-11T09:16:34Z

Credit to @xplat for the heterogeneous patch idea for use with dmap.

Co-Authored-By: Alexandre Esteves <2335822+alexfmpe@users.noreply.github.com>

…p-inner-patch

Do this without changing strictness using ~.

This makes it match the new PatchMapWithMove, and obviates PatchDMapWithReset.

This isn't good enough. I think there is a bug in that we don't store enough information in the two field: we need both inner patches with the join so we can combine them together. In the non-dependent case, nothing stops us from making this misake, but in the dependent case we would need to use a `Category` instance on the inner move patches. We plainly aren't combining two inner patches in the double move case, and the types let us know.

Ericson2314 · 2020-01-11T09:21:19Z

As described in d72221c I think all PatchFooWithMove are wrong currently in the double-move case of their Semigroup instance.

I'm not sure it is a good idea in general, so I am putting in another library instead.

xplat · 2020-01-13T20:38:32Z

src/Data/Monoid/DecidablyEmpty.hs

+module Data.Monoid.DecidablyEmpty where
+
+class Monoid a => DecidablyEmpty a where
+  isNull :: a -> Bool


according to conventions this should be either null or isMempty ...

xplat · 2020-01-13T20:40:25Z

src/Data/Patch.hs

-                                       unsafePatchDMapWithMove, weakenPatchDMapWithMoveWith)
+import Data.Patch.DMapWithMove as X
+  ( PatchDMapWithMove, const2PatchDMapWithMoveWith, mapPatchDMapWithMove
+  , patchDMapWithMoveToPatchMapWithMoveWith


sprained my tongue :(

(i have no suggestions though)

xplat · 2020-01-13T20:44:53Z

src/Data/Patch.hs

-  negateG _ = Proxy
-  _ ~~ _ = Proxy
+  negateG ~Proxy = Proxy
+  ~Proxy ~~ ~Proxy = Proxy


is this actually necessary to be in this PR?

xplat · 2020-01-13T20:48:37Z

src/Data/Patch/Class.hs

-  type PatchTarget p :: *
+class ( PatchHet p
+      , PatchSource p ~ PatchTarget p
+      ) => Patch p where


i like the way you tied these two together, it should make life easier in a lot of cases.

xplat · 2020-01-13T20:49:37Z

src/Data/Patch/Class.hs

  apply (Identity a) _ = Just a

+-- | 'Proxy' can be used as a 'Patch' that always fully replaces the value


this comment is wrong -- a 'Patch' that never alters the value?

xplat · 2020-01-13T21:05:07Z

src/Data/Patch/DMapWithMove.hs

+  :: forall k v
+  .  (GCompare k, GShow k)
+  => DMap k (NodeInfo k v)
+  -> Bool
 validPatchDMapWithMove = not . null . validationErrorsForPatchDMapWithMove


has anyone ever used this? the sense of the test looks reversed to me ...

xplat · 2020-01-13T21:14:25Z

src/Data/Patch/DMapWithMove.hs

+          --    [ toAfter :=> Fixup_Delete ]
+          --  _ ->
+              [ toAfter :=> Fixup_Update (This $ From_Move $ fromBefore :=> (Flip $ p1 `o` p0))
+              , fromBefore :=> Fixup_Update (That $ To_Move $ toAfter :=> (p1 `o` p0))


p1 o p0 should be pulled out into a let to avoid recomputation

xplat · 2020-01-13T21:19:52Z

src/Data/Patch/DMapWithMove.hs

-moveDMapKey :: GCompare k => k a -> k a -> PatchDMapWithMove k v
+moveDMapKey
+  :: GCompare k
+  => k a -> k a -> PatchDMapWithMove k (Proxy3 v)


These can be generalized to any patchy Category pretty safely ...

xplat · 2020-01-13T21:35:49Z

src/Data/Patch/DMapWithMove.hs

+  -- | Move the value from the given key @k a@ to this key. The target key
+  -- should also have an entry in the patch giving the current key in
+  -- @_nodeInfo_from@, usually but not necessarily with @To_Delete@.
+  To_Move :: !(DSum k (p from)) -> To k p from


It would probably be better actually, once the types have been verified as lining up, to force one of these to Proxy so there aren't two copies of the patch that can end up being unshared.

That is what the original MapWithMove2 patch used to do, more or less. Maybe check out that PR. I couldn't get version to type check but perhaps I needed to just rebuild some values more to get things to type check.

xplat · 2020-01-13T21:40:11Z

src/Data/Patch/DMapWithMove/By.hs

+-- This type isn't used directly as the from field patch, but is instead wrapped
+-- in an existential. However, it is nice to be able to reason about this in
+-- isolation as it is itself a @Semigroupoid@ when the underlying patch is.
+data By (k :: a -> *) (p :: a -> a -> *) :: a -> a -> * where


Reading this, how is it different from From?

The only difference is no existentials and so has a Category instance.

It's not actually used at all; I made it in the process of figuring out what I was doing. I meant to call out that we probably should delete it but still keep it for the some review unless it helped anyone think about it too.

…-inner-patch

Ericson2314 · 2020-02-12T20:20:54Z

Should be made to include #14, and do something similar for a DMapWithMove vs DMapWithPatchingMove.

New one is called `DMapWithPatchingMove`

This is the one that doesn't store patches on the "to" side.

Now that it doesn't contain patches, it doesn't need so many params.

Expanded some thing fighting the type checker, can put back now.

danbornside and others added 10 commits October 8, 2019 13:24

add a couple of types for working with patches

d66d5e9

Update src/Reflex/Patch/Patchable.hs

732cb14

Co-Authored-By: Alexandre Esteves <2335822+alexfmpe@users.noreply.github.com>

Merge branch 'develop' into patch-dmap-reset

b742380

Merge remote-tracking branch 'patch/develop' into patch-dmap-reset

0918153

Add change log entry

0ec95ff

Merge remote-tracking branch 'origin/patch-dmap-reset' into dpatch-ma…

a74f716

…p-inner-patch

Get rid of entirely-_ pattern in Group instances

cf7b8fb

Do this without changing strictness using ~.

Add notion of heterogenous patch

7fd18c0

WIP: Patch in PatchDMapWithMove's From_Move

aa7311b

This makes it match the new PatchMapWithMove, and obviates PatchDMapWithReset.

Ericson2314 and others added 9 commits January 11, 2020 15:03

WIP Generalize DMap patching

f78dd16

Over the hill

69a7ab8

Get everything type checking again

3f037af

Export Data.Patch.DMapWithMove.By

6b8b77b

Remove PatchDMapWithReset now that it has been accounted for

ab5bd5d

Add decidably empty things

c99f2d2

Appease hlint

98634c9

Remove Patchable

5a13577

I'm not sure it is a good idea in general, so I am putting in another library instead.

Update ChangeLog.md

08f13f2

xplat reviewed Jan 13, 2020

View reviewed changes

Ericson2314 added 3 commits January 21, 2020 12:01

Merge branch 'develop' of github.com:reflex-frp/patch into dpatch-map…

6fcf213

…-inner-patch

WIP make old in terms of new

c3cac44

Merge branch 'patch-map-inner-patch' into dpatch-map-inner-patch

584334f

Ericson2314 added 5 commits April 22, 2021 22:22

Merge branch 'patch-map-inner-patch-desugar' into dpatch-map-inner-patch

6ef716a

Restore the old DMapWithMove,

ac8f411

New one is called `DMapWithPatchingMove`

DMapWithPatchingnMove: Restore old algo

48711bc

This is the one that doesn't store patches on the "to" side.

Simplify type of "To"

8e3015a

Now that it doesn't contain patches, it doesn't need so many params.

Simplify type of fixup

4da27a8

Ericson2314 added 7 commits April 23, 2021 13:17

Delete old code and format a bit

76f010e

Put back Cat.DecidablyEmpty

efb8a9b

Remove redundant geq

62bd0fc

Simplify

055f256

Expanded some thing fighting the type checker, can put back now.

Improve error

76033e9

Merge remote-tracking branch 'origin/develop' into HEAD

ccd725a

Replace Proxy3 with something more appropriate

873bad6

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

WIP: Combination of #1 and #2 #10

WIP: Combination of #1 and #2 #10

Ericson2314 commented Jan 11, 2020 •

edited

Loading

Ericson2314 commented Jan 11, 2020

xplat Jan 13, 2020

xplat Jan 13, 2020

xplat Jan 13, 2020

3noch Jan 14, 2020

xplat Jan 13, 2020

Ericson2314 Jan 14, 2020

xplat Jan 13, 2020

xplat Jan 13, 2020

xplat Jan 13, 2020

xplat Jan 13, 2020

xplat Jan 13, 2020

xplat Jan 13, 2020

Ericson2314 Jan 14, 2020

xplat Jan 13, 2020

Ericson2314 Jan 14, 2020 •

edited

Loading

Ericson2314 commented Feb 12, 2020

		apply (Identity a) _ = Just a

		-- \| 'Proxy' can be used as a 'Patch' that always fully replaces the value

WIP: Combination of #1 and #2 #10

Are you sure you want to change the base?

WIP: Combination of #1 and #2 #10

Conversation

Ericson2314 commented Jan 11, 2020 • edited Loading

Ericson2314 commented Jan 11, 2020

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Ericson2314 Jan 14, 2020 • edited Loading

Choose a reason for hiding this comment

Ericson2314 commented Feb 12, 2020

Ericson2314 commented Jan 11, 2020 •

edited

Loading

Ericson2314 Jan 14, 2020 •

edited

Loading