A program that "passes around state" is easier to test than a program whose 9,999 methods alter an atom. And after a failure wouldn't the atom have to be reverted back to some prior state in order for the next possibility to be tried? Of course if the state's appearance in the code is completely routine then you could automate it away with a macro, or if the code overall is quite routine then you could either code-generate it or otherwise make it data-driven to reduce the surface area of source code. But - in general - "passing around state" is one of the great things about not-OO.

Not sure if this will work for your situation, but take a look at interceptors. Some links http://pedestal.io/guides/what-is-an-interceptor https://lambdaisland.com/episodes/interceptors-concepts https://www.youtube.com/watch?v=bDN898hu_wQ https://ericnormand.me/article/a-model-of-interceptors https://quanttype.net/posts/2018-08-03-why-interceptors.html https://github.com/metosin/sieppari https://www.proctor-it.com/papillon-a-new-interceptor-library-for-clojurescript/

Thanks! Had forgotten about Pedestal's interceptors .. have used Pedestal before for a web service and quite liked the pattern there.

RE: code generation from data. I had thought about making a parser generator from the bnf grammar (since this is a toy) .. but decided I want to get to done before veering off into the weeds 🙂

Looks like YAML, maybe you can use a YAML parser? There are multiple ways to represent this, your suggestion is one example. For consistency, you could represent every (non-leaf?) item as [key children], turning C into [C []] and so on. Another option is to use maps:

{"A" ["a" "b"]
 "B" ["c"]
 "C" []}

great suggestion! thanks!

I was about to suggest the reverse, define a set (def formats #{"webp" "avif" "jpeg"}) and use it as a function: (formats imgfmt)

ah yes, that's much nicer!

thanks

will need to get used to using sets

I might not be the best person to answer your question, but I've made a few commits to MrAnderson, so I might be able to help. What problem are you having?

thanks, so basically it never compiles for me. when using [thomasa/mranderson "0.5.3"] and running lein inline-deps i get

RuntimeException: No reader function for tag Inf, compiling:(mranderson/inlined/parallel/v0v10/parallel/core.clj:323:31)

i get a sense that 0.5.4-SNAPSHOT might be still unstable, while 0.5.3 doesn’t play well with something in my lein environment

@U04H5V7URR9 is your library public? If so, can you share a link?

it’s a fork of https://github.com/fullcontact/full.aws/tree/0.10.9 by FullContact, based off tag 0.10.9

with deps modified to:

:dependencies [[org.clojure/clojure "1.8.0"]
                 [com.amazonaws/aws-java-sdk "1.11.16"]
                 [com.taoensso/faraday "1.7.1"              ; DynamoDB sugar
                  :exclusions [com.amazonaws/aws-java-sdk-dynamodb joda-time]]
                 ^:inline-dep [fullcontact/full.http "1.0.9"]
                 ^:inline-dep [fullcontact/full.json "0.12.0"
                               :exclusions [com.fasterxml.jackson.core/jackson-core]]
                 [fullcontact/full.async "0.9.0"]
                 [fullcontact/full.core "0.10.1"
                  :exclusions [org.clojure/clojurescript]]
                 ^:inline-dep [javax.xml.bind/jaxb-api "2.4.0-b180830.0359"]
                 ^:inline-dep [com.fasterxml.jackson.core/jackson-annotations "2.6.0"]]

Not at a laptop at the moment to check, but I see Clojure 1.8.0 in the list there, and the error message mentions Inf. Reader syntax for ##Inf was added in a later Clojure version.

of interest: with @U0508JT9N's blessing I've just created a #C04HWRD2D32 channel.

Good eyeballs @U0CMVHBL2, no unit tests for that lib, eh @U04H5V7URR9? Would be nice to know if it works before applying MrAnderson.

yeah it works without mranderson, I've tried with clojure 1.9.0, and also clojure.tools that supports Inf and couldn't get it to compile still

If nobody else beats me to it, I can have a deeper peek sometime soon (maybe tomorrow) @U04H5V7URR9. Is your fork public?

that would be great! my fork is private, but other than the dependencies changes I didn't change a thing

Can you make it public? I would want to make sure I'm working with what you are working with.

I can try running mranderson on the official repo just to make sure I didn't screw anything up myself

also, I would mention I'm running with apple silicon M1 which has been the source of the most unusual bugs twice already, so maybe that's somehow related to this conundrum as well

good to know!

But maybe you found a bug in MrAnderson too. Whatever the case, I expect we'll come to some solution.

let’s move to #C04HWRD2D32? it might get verbose 🙂

Ya, sure would be nice to have some traffic over there.