that's precisely it

the functions you create with mkremote don't return state

state goes into cells, and formulas react to changes in state in the normal javelin way

what the mkremote functions do return is a jQuery promise you can use to handle errors, for example

because errors want to be handled at the place where the request was initiated usually

like if you have something like `

how do you mean? where is state generated then?

(defc stem-cell1 nil)
(def login! (mkremote 'my.namespace/login! stem-cell1 ...))

you can call the login! function now in the client

like for example

(let [user (cell nil)
      pass (cell nil)]
  (form :submit #(login! @user @pass)
    (label
      "User: "
      (input :change #(reset! user @%)))
    (label
      "Pass: "
      (input :type "password" :change #(reset! pass @%)))
    (input :type "submit")))

when you submit that form the login! function is called with two arguments

but notice how the return value of that call is discarded

the result will actually end up in the stem-cell1 cell

asynchronously sometime later

and presumably some user data will end up in there, as that's what the my.namespace/login! function in the server will return

if the username and password are okay, that is

then you will have formulas looking at stem-cell1 that might hide the login form and show something else

but the form you see there doesn't care what happens when the user successfully logs in

or not

it just needs to know how to collect the arguments and call that functions

where does stem-cell1 get defined? in the hoplon page?

you'd need to define that in your application

there i updated it

ok so it could be any cell

yes

you would likely wrap the login form like this:

(div :toggle (cell= (not stem-cell1))
  (let [user ...
    ;; login form as above here

so the login form is only shown when the user isn't logged in

i.e. when the cell containing user data doesn't have any user data in it

and that goes for every bit of the application which needs to show or hide itself

you'd do it via formulas, yes

generally you can have two containers, one for logged in and one for logged out

however when doing so avoid nesting cell=

(html
  (head ...)
  (body
    (div :toggle logged-out?
      ;; login form here
      )
    (div :toggle logged-in?
      ;; logged-in content here
      )))

with

(defc= logged-in? stem-cell1)
(defc= logged-out? (not stem-cell1))

it's pretty easy to reason about this way

ok im tracking

since no state is being built up by pieces

the stem-cell1 cell contains some value

and the formulas relationship to it is invariant

compare to making an async AJAX call and attaching callbacks to show/hide all the various things

that way would tightly couple the login form to all those other things

this way the login form doesn't even know that an ajax call is being made, it just calls a function

and it doesn't know that anything needs to be hidden or shown

so you can factor the whole login form into a separate component that can be used in any application

this is the basic model for factoring a hoplon application more or less

ok i will need to do some more digging on castra

@flyboarder: https://github.com/hoplon/demos/tree/hoplon6/castra-chat is updated to use hoplon6 but not the latest neither castra3. But it runs on boot2 😉

@mynomoto: thanks i was looking at that a bit earlier

how far off is it from a castra3 implementation?

is 2.2 very different from 3?

@flyboarder: just a couple of hours, there is not that many differences. For a really simple project you can run lein new hoplon-castra your-project-name to generate a project that uses the latest hoplon and castra.

The main differences are internal, cljson -> transit, a castra session. The main practical difference is that castra is now a middleware instead of a handler.

as in it does things on both side of the response?

No, as you can use like https://github.com/tailrecursion/hoplon-castra-template/blob/master/src/leiningen/new/hoplon_castra/handler.clj#L17

*can use it like

It intercepts the requests and short circuits it if it's a post request.

@mynomoto: so I can mixin any other ring middlewares then as well

how does that work with other things such as using the rest routes for an API?\

You need to apply the castra middleware only on some of the routes since it intercepts all post requests, so don't apply it on routes that will receive those requests. Other than that you can mix and match at will.

sorry not sure what you mean by that?

how is it applied to only some?

would I be creating an api app with a different route list?

In the end you need only one app, but you can have middleware applied only on part of it. http://stackoverflow.com/questions/28016968/serving-app-and-api-routes-with-different-middleware-using-ring-and-compojure

got to go now, good night.

night, also git the sack

😛

gotta hit*

@mynomoto: running boot dev on the hoplon-castra lein template gives me this stacktrace - https://www.refheap.com/109669 Running on windows if that's important

@mynomoto: I just realized I changed clojure's version to 1.8.0-alpha5, 1.7.0 works fine

@micha: I'm looking at spacemacs as you mentioned a couple days ago, man it's huge!

Just got my first castra rpc call working

i just updated and posted a minesweeper game i abandoned, https://github.com/alandipert/hoplon-minesweeper

http://tailrecursion.com/~alan/hoplon-minesweeper/

haha awesome

it needs the expand-adjacent-non-mines function still, but that would be fun to write

no wonder that game is a classic

also mathiasx's roguelike, that will rule

@flyboarder: we are sharing sessions stored in mongodb between our castra backend and our legacy nodejs app i think i can share the related code if it's not already shared. (though i was just deleting something like that today... i hope it was pushed to some other server

@alandipert: the original version had a couple more features, most notably middle-clicking a cell that already had all the flags it needs to set up would flip all the non-flipped neighbors at once

other than that it feels just like the original 😛

@micha: I remember you said formulas referencing other formulas are a code smell to you, yet here in Alan's minesweeper I see many formulas like that

@xifi: i think he was saying something like creating new formula cells within formula cells are a problem (causes memory leak if im correct)

like:

(def stem (cell 123))
(def double-squared
  (cell=
    (let [double (cell= (* 2 stem))]
         (* double double))

(def stem           (cell 123))
(def double         (cell= (* 2 stem)))
(def double-squared (cell= (* double double)))

@xifi: referencing other formulas is not bad

that's the whole point of having formulas

the bad thing is creating new formulas in a formula

that's very different from just referencing an existing formula

if you do like (cell= (cell= foo)) or something you're creating new formulas each time the thing is recomputed

and nothing will ever garbage collect them

and there is no need to do this

@micha: i wonder if cell= should complain if it has cell= in it?

i guess it couldn't really know

@onetom: I would appreciate any code you can share, im still new to clojure so seeing real world applications helps connect all the pieces together, @micha really got me tracking yesterday with the whole application development model in hoplon

@micha: am I correct that creating an anonymous cell= at runtime within a cell= would also cause problems

@flyboarder that's what i was saying yes

the rule of thumb is just don't create new formulas in the body of a formula

I see, thanks @onetom and @micha

I just launched the nukes on my linux notebook, the damn thing won't load my /home partition 😞

@micha ah I see that was probably the issue I was having before, refactoring yet again 😛

nukes worked 😅

just a few questions about hoplon, there used to be a documentation but since the new design where has all the data gone?

- is it normal that the "A stateful custom element" example the timers show different times?

- if I wanted to do a SPA that were to live in cordova, can HLisp be compiled to say app.js and be included as a lib? (like OM)

- updates from the server, how are the done?

@sjol: hi, yes, the old docs are in history but they were too out of date for us to keep public

to answer your questions: 1. yes it is normal, the point of that demo is to demonstrate that different instances of the same component can maintain their own state

2. i haven't made a mobile app with hoplon but but others here have

re: 3. and example is https://github.com/hoplon/demos/tree/hoplon6/castra-chat

we advocate https://github.com/hoplon/castra for server comm but any technique you use that can get values into cells is an option

@alandipert: so I should be able to compile the templates to pure Javascript?

everything compiles to javascript, yes

nice! And have you tried to use javelin with lib like D3.js? are there any caveat I should know about? ( OM has a few, which is why I ask )

i have done some d3, and there aren't any gotchas i'm aware of

@alandipert: i cant seem to get castra-chat to build

great, my intent is to use a cell as the data source and have the data change thus the d3 visualization change

it complains about missing hoplon, do the new versions not build yet?

Could not locate tailrecursion/boot_hoplon__init.class or tailrecursion/boot_hoplon.clj on classpath

flyboarder: can you paste your build.boot file please?

right outta the hoplon6 repo

i don't see any hoplon dependency in there

oh the build/deps

this is from the hoplon/demos repo hoplon6 branch

i remember now

if you look on clojars you can see the versions that will work

https://clojars.org/tailrecursion/hoplon

that demo is probably out of whack with latest developments

those things won't be batteries-included until we finish all the moving of things

ok sounds good, but im confused as to why it is looking for the tailrecursion namespace at all? it's not in the deps file anywhere

wait I see it

it was in the boot file above looking for boot-hoplon