I believe the answer is currently no. As I understand it, the entire instance resolution mechanism drops away after type checking and is therefore not available to the simplifier. So if you need to add a constraint on the RHS of a rule, I think you're mostly out of luck. The only thing I can think of is that you can look at what the specializer does and try to look around to find the right dictionary, but that sounds hard and brittle. OTOH, I'm not an expert, so maybe there's something major I've missed. On Oct 2, 2017 8:04 PM, "Conal Elliott" wrote:

Thanks very much for the reply, Joachim.

Oops! I flubbed the example. I really `morph` to distribute over an application of `comp`. New code below (and attached). You're right that I wouldn't want to restrict the type of `morph`, since each `morph` *rule* imposes its own restrictions.

My questions:

* Is it feasible for GHC to combine the constraints needed LHS and RHS to form an applicability condition? * Is there any way I can make the needed constraints explicit in my rewrite rules? * Are there any other work-arounds that would enable writing such RHS-constrained rules?

Regards, -- Conal

``` haskell {-# OPTIONS_GHC -Wall #-} -- Demonstrate a type checking failure with rewrite rules

module RuleFail where

class C k where comp' :: k b c -> k a b -> k a c

instance C (->) where comp' = (.)

-- Late-inlining version to enable rewriting. comp :: C k => k b c -> k a b -> k a c comp = comp' {-# INLINE [0] comp #-}

morph :: (a -> b) -> k a b morph = error "morph: undefined"

{-# RULES "morph/(.)" forall f g. morph (g `comp` f) = morph g `comp` morph f #-}

-- • Could not deduce (C k) arising from a use of ‘comp’ -- from the context: C (->) -- bound by the RULE "morph/(.)" ```

On Mon, Oct 2, 2017 at 3:52 PM, Joachim Breitner

...

wrote:

...

Hi Conal,

The difference is that the LHS of the first rule is mentions the `C k` constraint (probably unintentionally):

*RuleFail> :t morph comp morph comp :: C k => k1 (k b c) (k a b -> k a c)

but the LHS of the second rule side does not:

*RuleFail> :t morph addC morph addC :: Num b => k (b, b) b

A work-around is to add the constraint to `morph`:

morph :: D k b => (a -> b) -> k a b morph = error "morph: undefined"

but I fear that this work-around is not acceptable to you.

Joachim

Am Montag, den 02.10.2017, 14:25 -0700 schrieb Conal Elliott: > -- Demonstrate a type checking failure with rewrite rules

...

module RuleFail where

class C k where comp' :: k b c -> k a b -> k a c

instance C (->) where comp' = (.)

-- Late-inlining version to enable rewriting. comp :: C k => k b c -> k a b -> k a c comp = comp' {-# INLINE [0] comp #-}

morph :: (a -> b) -> k a b morph = error "morph: undefined"

{-# RULES "morph/(.)" morph comp = comp #-} -- Fine

...

class D k a where addC' :: k (a,a) a

instance Num a => D (->) a where addC' = uncurry (+)

-- Late-inlining version to enable rewriting. addC :: D k a => k (a,a) a addC = addC' {-# INLINE [0] addC #-}

{-# RULES "morph/addC" morph addC = addC #-} -- Fail

-- • Could not deduce (D k b) arising from a use of ‘addC’ -- from the context: D (->) b

-- Why does GHC infer the (C k) constraint for the first rule but not (D k b) -- for the second rule?

_______________________________________________ Glasgow-haskell-users mailing list Glasgow-haskell-users@haskell.org http://mail.haskell.org/cgi-bin/mailman/listinfo/glasgow-haskell-users -- Joachim Breitner mail@joachim-breitner.de http://www.joachim-breitner.de/

_______________________________________________ Glasgow-haskell-users mailing list Glasgow-haskell-users@haskell.org http://mail.haskell.org/cgi-bin/mailman/listinfo/glasgow-haskell-users

Hi, Am Montag, den 02.10.2017, 17:03 -0700 schrieb Conal Elliott:

My questions:

* Is it feasible for GHC to combine the constraints needed LHS and RHS to form an applicability condition? * Is there any way I can make the needed constraints explicit in my rewrite rules? * Are there any other work-arounds that would enable writing such RHS-constrained rules?

if you are fine writing one RULE per _instance_ of C, the following works: {-# LANGUAGE ExplicitForAll, TypeApplications #-} {-# OPTIONS_GHC -Wall #-} module RuleFail where class C k where comp' :: k b c -> k a b -> k a c instance C (->) where comp' = (.) instance C (,) where comp' (_,a) (c,_) = (c,a) -- Late-inlining version to enable rewriting. comp :: C k => k b c -> k a b -> k a c comp = comp' {-# INLINE [0] comp #-} morph :: forall k a b. (a -> b) -> k a b morph _ = error "morph: undefined" {-# NOINLINE morph #-} {-# RULES "morph/(.)/->" forall f g. morph @(->) (g `comp` f) = morph g `comp` morph f #-} {-# RULES "morph/(.)/(,)" forall f g. morph @(,) (g `comp` f) = morph g `comp` morph f #-} Let’s look at the rules: $ ghc -O -c -ddump-rules RuleFail.hs ==================== Tidy Core rules ==================== "morph/(.)/(,)" [ALWAYS] forall (@ b) (@ b1) (@ a) ($dC :: C (->)) (f :: a -> b) (g :: b -> b1). morph @ (,) @ a @ b1 (comp @ (->) @ b @ b1 @ a $dC g f) = comp @ (,) @ b @ b1 @ a $fC(,) (morph @ (,) @ b @ b1 g) (morph @ (,) @ a @ b f) "morph/(.)/->" [ALWAYS] forall (@ b) (@ b1) (@ a) ($dC :: C (->)) (f :: a -> b) (g :: b -> b1). morph @ (->) @ a @ b1 (comp @ (->) @ b @ b1 @ a $dC g f) = comp @ (->) @ b @ b1 @ a $dC (morph @ (->) @ b @ b1 g) (morph @ (->) @ a @ b f) As you can see, by specializing the rule to a specific k, GHC can include the concrete instance dictionary (here, $fC(,)) _in the rule_ so it does not have to appear on the LHS. This is pretty much how specialization works. Is that a viable work-around for you? It involves boilerplate code, but nothing that cannot be explained in the documentation. (Or maybe TH can create such rules?) If this idiom turns out to be useful, I wonder if there is a case for -rules specified in type classes that get instantiated upon every instance, e.g. class C k where comp' :: k b c -> k a b -> k a c {-# RULES "morph/(.)/(,)" forall f g. morph @k (g `comp` f) = morph g `comp` morph f #-} Greetings, Joachim -- Joachim Breitner mail@joachim-breitner.de http://www.joachim-breitner.de/

Hi, Now that I think about it: You can probably even generate these rules in a core2core pass that looks for instances of C, and then adds the rules to the mod_guts. That would solve the problem neatly, I’d say. Greetings, Joachim Am Dienstag, den 03.10.2017, 08:45 -0700 schrieb Conal Elliott:

Hi Joachim. Thanks very much for the suggestions and the `-ddump- rules` view. I wouldn't want to make people write `morph` rules for all combinations of operations (like `(.)`) and categories, but perhaps as you suggest those rules can be generated automatically.

Regards, - Conal

On Tue, Oct 3, 2017 at 7:52 AM, Joachim Breitner wrote:

...

Hi,

Am Montag, den 02.10.2017, 17:03 -0700 schrieb Conal Elliott:

...

My questions:

* Is it feasible for GHC to combine the constraints needed LHS and RHS to form an applicability condition? * Is there any way I can make the needed constraints explicit in my rewrite rules? * Are there any other work-arounds that would enable writing such RHS-constrained rules?

if you are fine writing one RULE per _instance_ of C, the following works:

{-# LANGUAGE ExplicitForAll, TypeApplications #-} {-# OPTIONS_GHC -Wall #-} module RuleFail where class C k where comp' :: k b c -> k a b -> k a c

instance C (->) where comp' = (.) instance C (,) where comp' (_,a) (c,_) = (c,a)

-- Late-inlining version to enable rewriting. comp :: C k => k b c -> k a b -> k a c comp = comp' {-# INLINE [0] comp #-}

morph :: forall k a b. (a -> b) -> k a b morph _ = error "morph: undefined" {-# NOINLINE morph #-}

{-# RULES "morph/(.)/->" forall f g. morph @(->) (g `comp` f) = morph g `comp` morph f #-} {-# RULES "morph/(.)/(,)" forall f g. morph @(,) (g `comp` f) = morph g `comp` morph f #-}

Let’s look at the rules:

$ ghc -O -c -ddump-rules RuleFail.hs

==================== Tidy Core rules ==================== "morph/(.)/(,)" [ALWAYS] forall (@ b) (@ b1) (@ a) ($dC :: C (->)) (f :: a -> b) (g :: b -> b1). morph @ (,) @ a @ b1 (comp @ (->) @ b @ b1 @ a $dC g f) = comp @ (,) @ b @ b1 @ a $fC(,) (morph @ (,) @ b @ b1 g) (morph @ (,) @ a @ b f) "morph/(.)/->" [ALWAYS] forall (@ b) (@ b1) (@ a) ($dC :: C (->)) (f :: a -> b) (g :: b -> b1). morph @ (->) @ a @ b1 (comp @ (->) @ b @ b1 @ a $dC g f) = comp @ (->) @ b @ b1 @ a $dC (morph @ (->) @ b @ b1 g) (morph @ (->) @ a @ b f)

As you can see, by specializing the rule to a specific k, GHC can include the concrete instance dictionary (here, $fC(,)) _in the rule_ so it does not have to appear on the LHS. This is pretty much how specialization works.

Is that a viable work-around for you? It involves boilerplate code, but nothing that cannot be explained in the documentation. (Or maybe TH can create such rules?)

If this idiom turns out to be useful, I wonder if there is a case for -rules specified in type classes that get instantiated upon every instance, e.g.

class C k where comp' :: k b c -> k a b -> k a c {-# RULES "morph/(.)/(,)" forall f g. morph @k (g `comp` f) = morph g `comp` morph f #-}

Greetings, Joachim -- Joachim Breitner mail@joachim-breitner.de http://www.joachim-breitner.de/

-- Joachim Breitner mail@joachim-breitner.de http://www.joachim-breitner.de/

Hi, I think creating your rules as built-in rules is a good way to go. You could reduce the complexity somewhat by generating all n*m rules as “normal” rules in the plugin when you see the instancs. This way, the plugin does not have to do anything when you want the rule to actually file (…and maybe the plugin does not have to be loaded at all). But I am not sure if it is worth the effort. And the built-in rules are more efficient, as you say. Greetings, Joachim Am Dienstag, den 03.10.2017, 09:01 -0700 schrieb Conal Elliott:

Thanks for the suggestion, Joachim.

Since I'm writing a core-to-core plugin anyway, it wasn't so hard for me to implement all of these n*m rules (for n operations and m instances) at once via a "built-in" rewrite rule that explicitly manipulates Core expressions. Doing so is probably also considerably more efficient than matching against many rewrite rules (whether generated manually or automatically), at least the way rewrite rule matching is currently implemented. As you & I discussed at ICFP, I'm looking for ways to reduce the complexity of the plugin to make it easier to maintain and extend, and I thought that dictionary synthesis from rewrite rules might be one.

Regards, -- Conal

On Tue, Oct 3, 2017 at 8:49 AM, Joachim Breitner wrote:

...

Hi,

Now that I think about it: You can probably even generate these rules in a core2core pass that looks for instances of C, and then adds the rules to the mod_guts. That would solve the problem neatly, I’d say.

Greetings, Joachim

Am Dienstag, den 03.10.2017, 08:45 -0700 schrieb Conal Elliott:

...

Hi Joachim. Thanks very much for the suggestions and the `-ddump- rules` view. I wouldn't want to make people write `morph` rules for all combinations of operations (like `(.)`) and categories, but perhaps as you suggest those rules can be generated automatically.

Regards, - Conal

On Tue, Oct 3, 2017 at 7:52 AM, Joachim Breitner wrote:

...

Hi,

Am Montag, den 02.10.2017, 17:03 -0700 schrieb Conal Elliott:

...

My questions:

* Is it feasible for GHC to combine the constraints needed LHS and RHS to form an applicability condition? * Is there any way I can make the needed constraints explicit in my rewrite rules? * Are there any other work-arounds that would enable writing such RHS-constrained rules?

if you are fine writing one RULE per _instance_ of C, the following works:

{-# LANGUAGE ExplicitForAll, TypeApplications #-} {-# OPTIONS_GHC -Wall #-} module RuleFail where class C k where comp' :: k b c -> k a b -> k a c

instance C (->) where comp' = (.) instance C (,) where comp' (_,a) (c,_) = (c,a)

-- Late-inlining version to enable rewriting. comp :: C k => k b c -> k a b -> k a c comp = comp' {-# INLINE [0] comp #-}

morph :: forall k a b. (a -> b) -> k a b morph _ = error "morph: undefined" {-# NOINLINE morph #-}

{-# RULES "morph/(.)/->" forall f g. morph @(->) (g `comp` f) = morph g `comp` morph f #-} {-# RULES "morph/(.)/(,)" forall f g. morph @(,) (g `comp` f) = morph g `comp` morph f #-}

Let’s look at the rules:

$ ghc -O -c -ddump-rules RuleFail.hs

==================== Tidy Core rules ==================== "morph/(.)/(,)" [ALWAYS] forall (@ b) (@ b1) (@ a) ($dC :: C (->)) (f :: a -> b) (g :: b -> b1). morph @ (,) @ a @ b1 (comp @ (->) @ b @ b1 @ a $dC g f) = comp @ (,) @ b @ b1 @ a $fC(,) (morph @ (,) @ b @ b1 g) (morph @ (,) @ a @ b f) "morph/(.)/->" [ALWAYS] forall (@ b) (@ b1) (@ a) ($dC :: C (->)) (f :: a -> b) (g :: b -> b1). morph @ (->) @ a @ b1 (comp @ (->) @ b @ b1 @ a $dC g f) = comp @ (->) @ b @ b1 @ a $dC (morph @ (->) @ b @ b1 g) (morph @ (->) @ a @ b f)

As you can see, by specializing the rule to a specific k, GHC can include the concrete instance dictionary (here, $fC(,)) _in the rule_ so it does not have to appear on the LHS. This is pretty much how specialization works.

Is that a viable work-around for you? It involves boilerplate code, but nothing that cannot be explained in the documentation. (Or maybe TH can create such rules?)

If this idiom turns out to be useful, I wonder if there is a case for -rules specified in type classes that get instantiated upon every instance, e.g.

class C k where comp' :: k b c -> k a b -> k a c {-# RULES "morph/(.)/(,)" forall f g. morph @k (g `comp` f) = morph g `comp` morph f #-}

Greetings, Joachim -- Joachim Breitner mail@joachim-breitner.de http://www.joachim-breitner.de/

-- Joachim Breitner mail@joachim-breitner.de http://www.joachim-breitner.de/

_______________________________________________ Glasgow-haskell-users mailing list Glasgow-haskell-users@haskell.org http://mail.haskell.org/cgi-bin/mailman/listinfo/glasgow-haskell-users -- Joachim Breitner mail@joachim-breitner.de http://www.joachim-breitner.de/