Yilin's powerful syntax technique

[kenbot]

Introduces Yilin's powerful new syntax technique, replacing PR #1055.

• Starts from Yilin's commits, preserving it in the history
• Has cake and also eats it, 🎂 nom nom nom
• Moves the Focus keywords from monocle.syntax.FocusSyntax into a monocle.Focus.KeywordContext trait, because it not like the other xxxSyntax traits and shouldn't live with them or look like them. Alternative naming suggestions welcome
• Moves the extension ... def focus pre-applied Focus into an AppliedFocusSyntax, which is like the other syntax traits
• Haven't mixed that into monocle.syntax.all because I haven't figured out how to do that correctly with the different Scala version directories, suggestions welcome I figured it out. Lmk if you want me to remove it from object Focus as well
• Fixed up error messages that still somehow saying "GenLens"
• Moved the keyword stripping logic inside the macro, because I considered that it belongs in there, it is not truly an external bolt-on. You can't separate the internal logic from the fact that the keyword context was passed in
• Split the general lambda function configuring out from ParserLoop into a new thing LambdaConfigParser, which also does the keyword magic stripping. These boundaries make sense, because this initial config stuff is very slow moving and should not change often. The feature loop stuff in ParserLoop however, is fast moving and often needs to change.
• ParserParams renamed to ParserConfig, and unused type argument removed.

[yilinwei]

LGTM.

Raised #1080 to track the rest of the issue.

[simlei]

Hi, I'm a total outsider here; but I got very interested seeing the KeywordContext ?=> stuff in Monocle3. I tried to figure it out by reading #1055 — but I think an insider's explanation would be super helpful. Would someone care to elaborate or link me to elaborations of how cake is had and eaten too, and how this is probably a Scala 3 feature to be desired by generic library authors? I shall be most happy :)

[kenbot]

Hi @simlei!
The problem we had was that we wanted both:

• to have type-safe "keywords" inside the focus expression (eg each in Focus[Basket](_.items.each.price))
• we didn't want to have to import obnoxious "extensions on everything" , or force the user to separately import the Focus syntax, which we thought would be an unacceptable UX penalty.

We define our "keywords" in the KeywordContext trait; so the context function lambda: (KeywordContext ?=> From => To) means that (a) the From => To lambda is typechecked with respect to an implicitly provided KeywordContext, allowing the keywords to be used within; however (b), the Focus.apply function can't be run without providing a given KeywordContext instance.

So we want the juicy keyword goodness from (a), but we want to avoid paying for (b).

Fortunately, because this is a macro, we can do this. Firstly, the Scala compiler typechecks the lambda term provided by the user, which works, because it sees the KeywordContext ?=> bit. The typechecking we wanted is done - but we don't want the KeywordContext ?=> bit anymore, and certainly don't want to instantiate or provide one. So secondly, inside the macro we match on the user's lambda and throw away the KeywordContext ?=> bit. The real code that gets generated and run is created from the information in the user's supplied lambda argument, but the KeywordContext argument and the keyword method calls (ie .each, .index, etc) are all gone.

Hopefully this is a bit clearer, please feel free to ask more questions.

[simlei]

Hi @kenbot, thank you very much for taking the time to outline this. I think this clears it up for me :) Also, this is the example most useful for me for learning about context lambdas that I have seen to date.
Cheers :)

[julien-truffaut]

@kenbot I wonder if there would be a benefit to mark KeywordContext as erased https://docs.scala-lang.org/scala3/reference/experimental/erased-defs.html

[kenbot]

We probably can, but the feature is still experimental as far as I know. Probably something to check back in on periodically

…

On Sun, 16 Jan 2022, 3:53 am Julien Truffaut, ***@***.***> wrote: @kenbot <https://github.com/kenbot> I wonder if there would be a benefit to mark KeywordContext as erased https://docs.scala-lang.org/scala3/reference/experimental/erased-defs.html — Reply to this email directly, view it on GitHub <#1079 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AACNIY56JWR5BP4FA4JNOYLUWGRB7ANCNFSM4X2OA56Q> . Triage notifications on the go with GitHub Mobile for iOS <https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675> or Android <https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub>. You are receiving this because you were mentioned.Message ID: ***@***.***>