How do I deal with a handler depending on another. Something like having a handler called get-user that request the user from the server, another got-user, that gets the data and puts it in the state, but now I have one called get-friends. Get-friends should only be run after get-user/got-user finished but I don’t want got-user to run get-friends as it’s not needed most of the time? Should get-friends do a dispatch-sync of get-user?

@pupeno: You can dispatch several times?

profil: what do you mean? I guess the answer is yes, but I don’t see how that helps.

get-friends dispatches get-user/got-user and then dispatches itself

How do I wait for got-user to run to then dispatch itself?

profil: here you have some sample code: http://stackoverflow.com/questions/31985649/how-to-have-handlers-run-sequentially-in-re-frame

@pupeno: ahh, I guess you would have to dispatch it from got-user then, you could parameterize it

or use dispatch-sync, which is a hack 😄

profil: but got-user is used in every single page while I want to fetch the friends only in one page.

dispatch-sync can’t be used inside a handler

Every piece of information fetched from the server will depend on the user, which means that if I fetch that information in got-user, got-user will downloading a humongous amount of data that will mostly not be used.

I could pass around next-handler but it feels like continuations, which are hard to debug, and I bet I’m not the first one to deal with this. I could copy code, but then I think I’ll end up with a lot of duplication.

dispatch itself then, until you have got a valid user, in get-friends (when (nil? (:user db)) (dispatch [:get-friends]))

Do you mean:

@pupeno: https://www.refheap.com/108170

(if (nil? (:user db)) (do (dispatch [:get-user]) (dispatch [:get-friends])) (ajax-call-to-get-friends))

?

Ah, that’s better.

It still feels a bit convoluted… for example, if I have 5 or 6 sequential calls.

why would you have 5 or 6 sequential calls?

I don’t know yet as I haven’t build a complex SPA before, but my server apps got complex enough where we needed to interact with 10 different models or more in a single render. I can see that turning into several requests to the server API.

what is the correct way of having a parent component P conditionally display child component A or child component B (based on atom data and secretary route), when i switch from one state to the other the view does not get updated properly?

http://paste.ubuntu.com/12070421/

(defn p []
  (let [a-or-b-state (subscribe [:something-important])]
    (fn []
      (if @a-or-b-state
        a
        b))))

escherize: have a look at the paste, i think i am using the exact same approach , bbut the view only updates the first time

Mmm

after that , i get partial updates or none at all

as this: (defn p [state] (if @state a b))

do you mean for project-page?

[body-main]

is the parent component

I would try wrapping your page-function on line 28 with []'s

Uncaught Error: Invalid arity: 1

after wrapping with vector

Hmm. https://github.com/reagent-project/reagent/issues/39

It's hard to troubleshoot without more context

yep i have been struggling with it for a few hours

@shaym: i think page-function needs to either be a reaction, or it needs to be in the render part of a form-2/3 component

@antishok: i was also wondering if i need the 3rd form (create-class) ?

form-2 is enough here i think

making page-function a reaction seems easier, ill try it

you think that form 2 without another reaction will work, or it needs the page-function reaction

i think you should make body-main a form-2 anyway

even for page that's needed

form-2 without a reaction would work, if you do the @page part inside the render function

also line 14 seems maybe off

not sure

here is body-main now , revised

(defn body-main[] (let [page (subscribe [:current-page]) page-function (get pages @page)] (println "page is " @page) (fn [] page-function) ))

that's still not good

@page needs to be done in the (fn [] .. ) part

(that's the render function)

like so :

(defn body-main[] (fn [] (let[[page (subscribe [:current-page]) page-function (get pages @page)]] page-function) )

and it should also probably return [page-function] as was suggested before

heh

the subscribe should not be inside the render function

look at the examples in the re-frame readme :P

so what did you mean by "@page needs to be done in the (fn [] .. ) part "

@page needs to be inside it, subscribe needs to be outside o it

are you referring to the print?

(defn body-main []
 (let [page (subscribe ..)]
  (fn []
    (let [page-function (get pages @page)]
     [page-function]))))

subscribe is outside the fn, @page is inside

i used the above form , if i return a vector i get an arity error , without it i still get partial rndering

does the page-function component take arguments?

oh pages is there, didn't notice

yep

def pages {:project-list [project-table] :project [project-page]}) (defn body-main[] (let [page (subscribe [:current-page]) ] (fn [] (let [page-function (get pages @page)] (println "page is " @page) page-function) ) ))

donno why you're getting an arity error ;S

@shaym: what do you mean by partial rendering?

ohh your pages map already has vectors in it that's why you get the arity error

on first navigation from :project -list to :project , the data from :project is displayed but one button from :project-list is also showing , when i click browser back , and it switches back to :project-list , the table which was originally there is missing

yes pages has a vector

after multiple back and forth navs i get ......Uncaught Error: Invariant Violation: ReactMount: Two valid but unequal nodes with the same

well not sure what's going on but,

2 things catching my eye are lines 9 and 14 (in your original paste)

what about them?

9 should maybe have [] instead of ()

[entity-table]

and 14 hmm it's wrong because again you're dereffing a subscription outside the render function

not sure what the best approach is to fix it though

what is that subscription pulling out?

(subscribe [:project @project-id])

a specific project from a list, that part actually works

i can see the need of returning a function rather then calling a function , but i must admit i am confused as to how many times it needs to be wrapped in a vector , when having a chain of function calls

doesnt it end up [[[[comp]]]] ?

i suppose ;p

so all derefs must be in the render function? i am fixing it now

yes, or in a reaction. i think things can still seem to be working ok even if you do the deref outside, but it will cause the component to be re-rendered on any change to app-db

which you don't want

i might be confusing things though, maybe that's what happens if you do a subscribe inside the render function :P

been a while since i've tested

i am confused by now also

i feel like back in my C++ days, pointers to pointers to pointers **my-ptr 😉

ok the 2nd thing i said was correct, as per re-frame readme: Eek! subscription in renderer ... Why is this wrong? Well, this component would be re-rendered every time app-db changed

makes sense, as a subscription means dereffing app-db

yep

i am wondering which is better/easier a chain of functions where the last one returns a vopm , or a chain of componenets nested within each other?

right now it looks like im mixing both

where the last one returns a what?

component , sorry

vomp=comp

ah ;)

well a chain of components feels "safer" to me, but i've yet to turn this feeling into an actual explanation i can reason about ;p

i feel the same , but yes it needs to be substantiated

i think it reduced paradigm confusion , all defined the same way

@mikethompson is usually good with putting stuff into words ;)

am sure he'll have some input when he's available

@antishok: thanks for all the help so far