what's going on if i get this from a hoplon (page ...)?

clojure.lang.ExceptionInfo: No reader function for tag js

(clj->js ...) works

@onetom: looks related to https://github.com/boot-clj/boot/issues/47

there is a boot.core/load-data-readers! function, perhaps it needs to be strategically called?

I getting this error in the https://github.com/exicon/foreign-lib-with-cljs-init repo which is a super minimalistic repo. it doesn't have datomic, doesn't have boot-reload, nothing fancy, but the latest (boot-)hoplon. I hit this issue sevaral times in the past 2 weeks but I was always lazy to track it down.

data: {#object[clojure.lang.Keyword 0x38105db8 ":type"] #object[clojure.lang.Keyword 0x4ffa1257 ":reader-exception"]}
                             clojure.core/ex-info          core.clj: 4593
   clojure.tools.reader.reader-types/reader-error  reader_types.clj:  330
                                              ...
                 clojure.tools.reader/read-string        reader.clj:  788
       tailrecursion.boot-hoplon.util/read-string          util.clj:   31
      tailrecursion.boot-hoplon.compiler/as-forms      compiler.clj:   46
tailrecursion.boot-hoplon.compiler/compile-string      compiler.clj:  112
                                              ...
                               clojure.core/apply          core.clj:  632
  tailrecursion.boot-hoplon.compiler/compile-file      compiler.clj:  124
                                              ...
            tailrecursion.boot-hoplon.impl/hoplon          impl.clj:   46
                                              ...
                               clojure.core/apply          core.clj:  630
                            boot.pod/eval-fn-call           pod.clj:  184
                                boot.pod/call-in*           pod.clj:  191
                                              ...
                                boot.pod/call-in*           pod.clj:  194
       tailrecursion.boot-hoplon/eval374/fn/fn/fn   boot_hoplon.clj:   81
              cljsjs.boot-cljsjs/eval778/fn/fn/fn   boot_cljsjs.clj:   30
              pandeiro.boot-http/eval480/fn/fn/fn     boot_http.clj:   71
...
              clojure.core/binding-conveyor-fn/fn          core.clj: 1916

@onetom: that actually makes sense

since hoplon is using the clojure reader

when it parses

we no longer really have to parse the page actually

so we could modify the compiler to not do that anymore

and avoid those errors

hoplon just needs to parse the first expression, the page declaration

i hope that brings some speed up too, because i was just working on our boot1/hoplon5 homepage today (hopefully for the last time) and it was compiling like lightning compared to boot2/hoplon6...

the hoplon compiler doesn't take much time

it's cljs compiler where it's spending most of its time i think

sure, yet the old one was noticeably/obviously faster and when it comes to recompilations, then it was comparable to the cljs compilation time

but i was working on the layout namespace which ~15 other hoplon pages depended on, so all of them were recompiled every time...

wait you're saying it's taking longer for hoplon to generate cljs namespace than it does for cljs to compiler them?

so it's just our degenerate case magnifies all these things i guess

not longer, but it was on par with the cljs incremental compilation

yes with the older versions of cljs, correct?

with newer versions of cljs hoplon compiler is still fast to generate cljs namespaces, but it takes the cljs compiler a long time to recompile them

no. boot1+hoplon5 <-- fast boot2+hoplon6+cljs1.7.48-3 <-- slow

no understand

i mean i understand that you're saying something is being slow

but it's not clear which thing is slow

boot1 packages its own cljs compiler, and it's an old version

i think i should just record a short work session for you to get a feel for it...

so dont even worry about this now

it's possible that the stuff we did to accomodate hot reloading of test namespaces could be slowing down the cljs compiler, too

sure, that's something i also felt when i was switching to 1.7.4x

but what i was saying is the .cljs.hl -> .cljs was noticeable faster with hoplon5

that's interesting

it's doing less work now, which you'd think would make it faster

that's why it's rather sad, not interesting

maybe some judicious memoization is in order

but still it mostly bothers myself only really. my other 6 colleagues seem pretty enduring in this regard, so no worries

how can i produce attributes without a value, like: <script async src="..."> ? (script :async true) produces <script async="async"> and i can see that in the code too, but is there a way not to provide a value? it's a purely aesthetic question though. eg if i do this in js:

s = document.createElement("script"); s.async = true; s

<script async></script>

i tried (script :async nil) and (script :async "") but none yields the "desired" result

@micha: you need to read the hoplon file for multiline strings too right?

@mynomoto: hey man!

no i removed that

i think the best solution there is to use something like the haml preprocessor

a much more powerful way to achieve the same

That makes me sad, but if it makes things easier I'm for it. I only need the multiline thing, don't want to use haml for it.

hmm

a preprocessor is i think still a good way to go

And about compilation speed I saw this: https://twitter.com/nikitonsky/status/636192947168485376

i think i have something that could almost work, i have the basis implemented already as an instaparse grammar

And https://twitter.com/nikitonsky/status/636193057931722752

it's not fs copying, because filesets don't copy anything

the fileset only makes hardlinks

We probably could have a task that only looks for something like multiline scripts and replaces it with the real clojure thing.

and all the cljs compiler stuff is cached

I had problems with lein cljsbuild and wouldn't trade reliability for speed, but if those measurements are precise it's a pretty big difference.

yeah

i suspect it has something to do with that testing problem

and the fix for that

@mynomoto: that sounds ideal

since I don't see any better docs I went through the http://hoplon.io docs and have a couple of questions: 1) there's :on-click and :do-class used. Is there a list of available events and what they do? 2) why are all castra requests POST? If you had a db and wanted to query it would you route it through castra? Wouldn't that be an example of a GET request? (as you might tell from these questions I'm still not sure what is castra's use case.) 3) does one first check for errors when calling through rpc? The doc says the state cell is only updated when the request is successful 4) since state, error and loading are cells, does one usually just hook up some code on top of that with cell=? This seems like you don't really need async channels and stuff like that 5) is there an example app like hoplon-minimal from @mynomoto that includes a server?

@xifi: ok the first question first

ka-boom, right

the :on-* and :do-* convention has been dropped in h6 for efficiency reasons

did you see the do! multimethod?

and on!?

yes

ok great

so consider the following

(div :foo bar "this is some text")

div here is a function

or it could be a DOM element, since hoplon makes DOM elements into functions by implementing IFn on them

the function application semantic on DOM elements denotes setAttribute and appendChild, as it does in HTML markup

tracking so far?

I think so

go on

ok so the IFn implementation knows certain things about which arguments make sense

like for instance only text nodes, comments, and elements can be children of a DOM element

this is a restriction imposed by the DOM in the browser

likewise attributes on DOM elements must have string keys and values

so hoplon elements can parse the arguments by using these constraints,

like for example a keyword can't be a child of a DOM element

so what happens in the example is the element takes its arguments

and when it sees a keyword it knows that that's an attribute key

it continues to iterate over all keyword/value pairs in the arglist until it gets to something that could be a child, like a DOM element or a string

for each attribute it can determine what to do by looking at the type of value associated with it

bar in the example above

if bar is a function then we know how to interpret the attribute

it must be an event handler, because attributes can't be set to functions in the DOM

if it's not a function then the attribute is interpreted as either setting an attribute (in the case where the value is a string)

or if the value is a javelin cell it wires up the attribute reactively such that the attribute will be updated wenever the cell changes

so for example:

(button :click #(js/alert "click!") "click me")

this creates a button that calls the handler when you click on it

I still don't see how does this relate to my question 😄 But go on

(a :href (cell= (str "" some-cell)) "go somewhere")

that craetes an anchor whose href attribute changes as some-cell changes

ok so now for a list of events

when you do something like the button above, what actually happens is the on! method is called on the element

(on! elem :click #(js/alert "click!"))

if you haven't yourself defined a dispatch for :click in the on! multimethod it will end up here: https://github.com/tailrecursion/hoplon/blob/master/src/tailrecursion/hoplon.cljs#L560-L562

notice it simply delegates to jQuery, setting up a jQuery event handler for "click"

so to finally answer your question: the list of available events is the same as the list of available jQuery events

you can add your own by implementing a dispatch for the on! multimethod

how do we feel?

I see, so I have to look into jQuery (remember I'm new at this)

I'll be afk for a long while now but keep writing I'll read it later

@xifi: about question 5) there is https://github.com/tailrecursion/hoplon-demos/tree/hoplon6/castra-simple

Almost all demos are converted to hoplon 6 on that branch.

@mynomoto: 👑

!m mynomoto

lol

@micha forgot to say - thanks!

@mynomoto: thanks too, will check the examples

@xifi: anytime!

@micha: cool! I'll set up a call tomorrow

haha sure

seriously though we could probably do a workshop or something

http://tmate.io and google hangouts or something

(= we ?)

whoever is interested

alan might be able to join, too

wouldn't mind that if it was during my work hours I don't have much time after that, you know, family

reading docs or source code is nice but having an education/workshop is great

think about it Bedtime here, bye guys and thanks

later!