New blog post: EDSLs as functions #120

adelbertc · 2016-10-07T18:51:22Z

Scheduled for Wednesday October 12.

larsrh · 2016-10-07T22:08:47Z

_posts/2016-10-12-edsls-part-2.md

+}
+```
+
+This works.. to the extent that we want only to work with


larsrh · 2016-10-07T22:10:58Z

_posts/2016-10-12-edsls-part-2.md

+```
+
+We've parameterized our language with a [higher-kinded type][hkt] which
+represents the context of our set. We can then write mini-programs which talks


Not a native speaker, but shouldn't this read "mini-programs that talk …"?

it is the same

I agree with @larsrh, it should be talk and not talks.

larsrh · 2016-10-07T22:13:30Z

_posts/2016-10-12-edsls-part-2.md

+set (think `Set[Int] => A` or `Set[Int] => Future[A]`).
+
+```scala
+trait SetLang[F[_]] {


IMHO this additional abstraction step could use some more thorough explanation. The two examples are good, but why not implement the trait for one of them?

Or maybe emphasize that you're now talking about the effects of performing operations on something set-like, instead of dealing with the set-like things themselves.

larsrh · 2016-10-07T22:14:17Z

_posts/2016-10-12-edsls-part-2.md

+and si2712fix-plugin 1.2.0 using [tut][tut].*
+
+[cats]: https://github.com/typelevel/cats "Typelevel Cats"
+[hkt]: http://typelevel.org/blog/2016/08/21/hkts-moving-forward.html "Higher-kinded types: the difference between giving up, and moving forward"


use post_url here too

larsrh · 2016-10-07T22:14:53Z

_posts/2016-10-12-edsls-part-2.md

+Fortunately we are now in a position to leverage a powerful tool:
+[monads][monads]. By extending our set language with monadic powers,
+we recover composition in an elegant way. The [Cats][cats] library is used
+for demonstration purposes, but discussion applies equally to [Scalaz][scalaz].


"but discussion applies to"?

should be "...but the discussion..."

larsrh · 2016-10-07T22:17:26Z

_posts/2016-10-12-edsls-part-2.md

+// Instance
+
+type SetState[A] = State[Set[Int], A]
+implicit val stateInstance: MonadSet[SetState] with MonadCalc[SetState] =


Here, you have successfully modularised and composed programs, but not the underlying implementations. It could be worth mentioning that you can also split up this instance into two.

larsrh · 2016-10-07T22:18:54Z

I'm planning to finish #119 by Sunday, so hopefully you'll be able to contribute this as a tut document.

larsrh · 2016-10-09T21:04:42Z

I changed the base of this pull request to development, so you have to move your post into the posts folder. Sorry about the hassle!

adelbertc · 2016-10-12T18:17:19Z

Probably not happening today, got a talk I got to prepare for on Monday and lots of stuff happening at $JOB. Put please still leave comments! *nudges @S11001001 * ;)

S11001001 · 2016-10-22T16:13:20Z

_posts/2016-10-12-edsls-part-2.md

+all traces of `Set`.
+
+```scala
+trait SetLang[F[_]] {


Such a type parameter would be *-kinded in a sensible refinement of this algebra in a direction other than the one you're about to take. 😊 You're paving the way for the monadic variant, but as a foundation, this one's a little shaky.

Hm could you expand on your comment a bit? I'm definitely biasing the post towards the monadic variant, but would be interested in hearing more about a different direction/guidance on how I can better present this.

Ah wait I think I know what you mean now, sounds good. Will adjust wording, thanks!

S11001001 · 2016-10-22T16:14:46Z

_posts/2016-10-12-edsls-part-2.md

+Interpretation of our language is straightforward, by implementing `SetLang` and
+passing in the corresponding implementation to `program`.
+
+Our language is still limited. `F` is arbitrarily designated to


The problem is that the language is too powerful, not that it's limited. The capabilities it provides are so strong as to prevent you from writing the kinds of interpreters you are interested in writing, but there is no “better than” rule here—your redesign admits those interpreters only by forbidding many programs that formerly could be written in SetLang.

S11001001 · 2016-10-22T16:15:22Z

_posts/2016-10-12-edsls-part-2.md

+```
+
+Our new encoding introduces a new but very important problem: how do we
+compose? In the previous encoding we could add and then remove by threading


“combine the results of multiple calls to SetLang methods?”, perhaps? I have a peeve about the mention of composition without specifying what kind of composition is meant.

S11001001 · 2016-10-22T16:16:08Z

_posts/2016-10-12-edsls-part-2.md

+
+While `App1` and `App2` are both valid compositions, the
+semantics of the compositions differ. `App1` describes a program where
+the computation of a *value* at each transition may fail whereas `App2`


…may fail—but any changes in state are preserved—whereas…

S11001001 · 2016-10-22T16:16:55Z

_posts/2016-10-12-edsls-part-2.md

+  def monad: Monad[F]
+
+  def raiseError[A](e: E): F[A]
+  def handleErrorWith[A](fa: F[A])(f: E => F[A]): F[A]


A simpler encoding of handleErrorWith may be

[A](fa: F[A]): F[Either[E, A]]

Assuming Functor[F], it's equally powerful.

Er, Monad.

Ah interesting, I'll play around and see which one ends up being easier to talk about. Thanks!

adelbertc · 2016-10-25T04:14:17Z

Feedback incorporated. Ran into a weird state with the new development branch so I reset --hard to upstream/development and redid the patch :/

Assuming everything looks OK let's schedule this for Wednesday Oct 26.

larsrh · 2016-10-25T04:53:25Z

@adelbertc Sorry I made you go through a Git hassle 😟 But the changes look good.

adelbertc · 2016-10-25T05:34:08Z

No worries, change was worth it I think :)

adelbertc · 2016-10-26T18:04:39Z

Thanks everyone for the review and feedback, really helpful, I really appreciate it! Probably the main reason I write for this blog, to get peer review, fact checking, and very useful comments.

Going to merge this now.

larsrh suggested changes Oct 7, 2016

View reviewed changes

larsrh changed the base branch from master to development October 9, 2016 21:04

S11001001 reviewed Oct 22, 2016

View reviewed changes

New blog post: EDSLs as functions

1d9ced6

adelbertc force-pushed the edsl-pt-2 branch from 0e7c0dd to 1d9ced6 Compare October 25, 2016 04:13

larsrh approved these changes Oct 25, 2016

View reviewed changes

adelbertc merged commit 52ee311 into typelevel:development Oct 26, 2016

New blog post: EDSLs as functions #120

New blog post: EDSLs as functions #120

Uh oh!

Conversation

adelbertc commented Oct 7, 2016 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

larsrh commented Oct 7, 2016

Uh oh!

larsrh commented Oct 9, 2016

Uh oh!

adelbertc commented Oct 12, 2016

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

adelbertc commented Oct 25, 2016

Uh oh!

larsrh commented Oct 25, 2016

Uh oh!

adelbertc commented Oct 25, 2016

Uh oh!

adelbertc commented Oct 26, 2016

Uh oh!

Uh oh!

adelbertc commented Oct 7, 2016 •

edited

Loading