Thread by @petterik_

Amazed by the first day at Clojure Conj.
1. Massively improved dynamic development experience with REBL.
2. Unified Declarative Domain Model library, hodur. I think this could be big.
3. Serious projects with AWS AppSync & Datomic Ions.
4. Future of clojure.spec

Another takeaway is that GraphQL won.

I'm excited about a UI library that uses something like hodur or core.spec.next selectors for view data requirements, datascript as app-state db & GraphQL remote requests.

View data requirements can make UI's performant. I think an improvement from om.next is that we can separate child queries and routing from specifying the data requirements of a view

Data requirements can be merged which can lead to more efficient client querying

DataScript as app-state solves a lot of problems, but it requires that custom root resolvers has all the data it needs to extract data from the database. This is a separate data requirement from what the view's want. We can merge these two reqs to create a remote request.

GraphQL is so big now. If one could just declare the above and it would translate to a GraphQL query, one would write only the view code, optimistic mutations and some root resolvers to create clients.

Optimistic updates is achieved by branching off each version of an app-state and rebasing it on the remote response. I think an even better analogy (and implementation) is to treat each version of the app-state as a Photoshop layer. One can then switch out layers to get a rebase.

Storing each remote response as a "layer", makes remote responses queryable, which makes it possible for a view to ask the client state the status of a mutation (pending, error, success) and whether a read has happened or not.

I started work on a library like this early this year and some of these ideas have been implemented, but I think hodur and spec's selector API completes my thoughts in this space. I'm excited to - maybe - get further on this project.

</rant>

Looking forward to day 2 and 3 of the Conj!