Morning

Good Morning!

mogge

morning

Bore da

@agile_geek the Welsh flag is missing from the emoji's 😞

@thomas it's more disappointing to me that there's no massive Welsh dragon right in the middle of the Union flag...but that's what you get for being a conquered Principality.

true... it always kinda bothered me as well that Wales isn't recognised in the Union Flag.

Maybe once Scotland and NI become a separate country they can have St. Andrew's and St. Patricks cross and England/Wales can have a St. George's cross with a big badass dragon on it.

<<NI become a separate country>> as declarations of war goes, that's a pretty good one.

🙂

I wonder if Scotland (or Wales or whoever) don't have a union jack on their flag - maybe Australia can finally lose theirs?

@korny hey, Australia had it's chance to become a Republic...

and interesting that NZ didn't ditch the Union flag bit of the flag either

yeah. It was a lesson in how to run a referendum while simultaneously undermining it. Why didn't the #brexit planners learn how to do that?

If the majority of people want to do X, put up a vote with "do nothing", "do a poor version of X", and "do a different poor version of X" as options. \

The sad thing about brexit is, it was precisely the opposite - "do nothing" vs "do a vague idealistic undefined version of X"

Hi all, if anyone wants to spread some UK Clojure love and good feelings the way of this company, allow me to plug for a RT: https://twitter.com/juxtpro/status/842307041309913088

Oh reminds me of my PA/Sporting Life days.

that’s an old picture of sheffield 😄

oh neat, i’m in that team photo

hmm, if it's really old I could change it 🙂

it’s only noticeably old if you know sheffield pretty well

it’s the tram livery that gave it away

i had a quick look for a better one, but nothing jumped out

thanks for setting up @glenjamin

what happened to the tram?

they got repainted

without the bands

@glenjamin you do look like you're about to be eaten by the revolving door tho!!

Morning.

I have a rather weird problem.... I am getting a stacktrace in our Immutant app. but not when I call the same function from the repl.

and even weirder.. when I don't remove a key from the map it works as well.

removing keys from records can turn them into ordinary maps?

it is not a record... just an ordinary map

I think

no defrecord in the code

Has anyone here ever use walk or postwalk ?

postwalk but only once

I use postwalk all the time… It's good 😄.

You'll likely find that the inputs are subtly different when using Immutant. If you're using CIDER I suggest plugging the debugger into the web request itself if you can. If not, it's time to sprinkle in some printlns 🙂.

Types can subtly change with maps & # of keys. But I would hope the code isn't coded to clojure.lang.TheLargerPersistentMap or something similar to that.

@thomas ^^

hmmm ok.... I'll try a few things more.

I found a few examples that change particular values in a map.. but I need to do it in nested maps as well... postwalk would be really good for this.

Yeah, postwalk is designed for that 🙂

Should just be:

(postwalk (fn [x] (if (map? x) (update-map x) x)) input

@thomas of course you could also look at traversy and spectre for that.

I think that postwalk should be more than sufficient for this kind of task.

Specter is targetted at mostly known deep paths into a structure (Update every odd element, nested 5 levels deep inside this structure)

and it is with postwalk that we get the stacktrace 😞

a java.lang.AbstractMethodError of all things... doesn't even make sense (at the moment at least)

and that looks very similar to our code @dominicm

And we can't really switch to Spectre either... running an older version of clj at the moment.

Your getting some strange things here....could this be some odd AOT side effect?

could be.... but how would I determine that is the case?

@thomas Are you able to run the uberjar locally?

@thomas Are they definitely maps? And not, e.g. datomic entities?

I can run it locally... but then I don't see it... I am running a newer version of Java though. and yes Datomic is involved.

:page {:db/id 17592186050855} is the key that is offending.

Yeah, this sounds like that then

Yeah, datomic entities are not maps

arghhh.

what to do then?

Do this: (into {} my-entity) as part of the postwalk?

at the moment we do this (clojure.walk/postwalk (fn [x] (if (map? x) (into {} (map f x)) x)) m))) where f is our function

so what about doing a into {} when we get back the data from Datomic?

Could do, as long as you don't need to use it as an entity later.

Are datomic entities passing map? I wonder :thinking_face:

dunno...

I mean, tbh, you probably want to do a d/pull to turn them into a map

ok let me have a look at the code (I didn't write it of course)

thank you

A datomic entity does not pass the map? predicate

it looks like we do this:

(defn export-entity
  [entity]
  (into {} entity))

If you perform (d/pull db '[*] eid) it will get all the attributes off the member and put them into a map (which passes map?)

^^ That also works if you have a whole entity

I wonder if that's sufficiently recursive…

or you can do what we do a style and specify only the attributes required in the pattern

@agile_geek It depends.

and you can use pull syntax in the query too

We've moved away from that. It limits consumers of the querying functions.

And you get ~50 attributes randonly added.

Not if they pass the pattern

That's the other option, but at that point, you may as well just use entities.

not if they're remote

You'd have to be more explicit about what remote means. But I suppose yes. Though I'd still make that an endpoint feature rather than db.

@thomas So yeah, the answer is to either add an extra (into {}) into your postwalk, or explicitly update that key before passing it in (with (update :page #(into {} %)). Which you do depends on the appropriate level of abstraction for yourr use-case.

ok let me try that. thank you

@dominicm I agree it's definitely an end point feature to allow fine control over what a client can 'pull' rather than just exposing the datalog syntax to a client but having pull syntax type semantics across the client-server boundaries can be useful

You need control at the end point if only for validation and security (which also involves validation btw to ensure you can't do the equivalent of SQL injection)

@agile_geek The database functions should probably be agnostic of whether they're being exposed to the world or just used internally in the same JVM. The Endpoint can do the (d/pull) equivalent on the entity it receives if it chooses.

Hmm, what's the overhead in terms of memory or performance if you have to map the pull over a large collection of entities rather than asking datomic to resolve it in one query? (I don't know Datomic's internals well enough to judge)

It's approximately the same if you just use (d/pull-many) 🙂

Though you'd need a large number of entities coming back from queries to even notice in the first place.