👋

Ok Hoplon, quick question : In do-watch, what is the init argument?

(defn do-watch
  ([atom f]
   (do-watch atom nil f))
  ([atom init f]
   (with-let [k (gensym)]
     (f init @atom)
     (add-watch atom k (fn [_ _ old new] (f old new))))))
 

any examples?

(we need a docstring here!)

Let's build a docstring, okay?

😉

Other question

I have a cell content-state that gets updated once, when the db loads the content-state data.

I have this in my views. (cell= (map #(view-item :id %) (map :id content-state))

And it gets slow

when click on a button, it takes a while to update, etc.

Chrome profiler is great for these kinds of things

i.e. are you sure it's that line that takes all the time?

not sure

didn't know about Chrome profiler

thanks

@dm3 ^^^

It seems like (program) takes some time, what is this?

you should see better on the flame graph

this just says that it spends all of its time checking equivalence and creating seqs

you want to see your functions in the profile

@dm3 r.handle the jquery thing takes a while

ok, no I get it better

scrolling down, I see a bit more

fns

(do-watch fetch-db-success
          #(when %2
             (reset! content-state
                     (vectorify-all form-content-vectors %2))))

Can I close a do-watch after the db is fetched, so after it's first f call (reset!) ?

https://clojurians.slack.com/archives/hoplon/p1463026078003421

^ @alandipert @micha interested to learn more

Oh, the loop-tpl thing seems slow too

ideally you'd extract some of you functions to top level from cells so that you could see them in the trace

none of the Hoplon fns are slow by themselves

ok

rather you're doing a huge number of recomputations

Yes, in (cell= (map #(view-item :id %) (map :id content-state)) my view-item-fn is pretty slow

But the thing is, if I could at least deref content-state when db-fetch is done, I could save some time I think

But How to deref-once-loaded?

do-watch?

not sure how

I already have this

(do-watch fetch-db-success
          #(when %2
             (reset! content-state
                     (vectorify-all form-content-vectors %2))))

Not sure where to deref what

not sure what you will gain by derefing content-state

@donmullen: we just deploy to ASGs now, without beanstalk

we have some simple scripts that we use to create ASGs with a user-data script that knows how to download application and run a Makefile

ASG - auto-scaling group (?)

but the overall organization is similar to beanstalk, eg. we have the concept of "applications" which can contain "environments"

yes ASG -> autoscaling group

that's the basic unit we allocate in AWS now

even if it will only contain a single instance, we still have that instance created by an ASG

environments for us are simply ASGs

ASGs are basically instnace factories

@micha - Interesting - thanks. Anything OSS yet - or planned? — similar to boot-beanstalk?

so once you create them you can easily create and manage instances

i can show you the scripts, they're really very simple

trivial even

all they do is call the aws api to create things and give those things sane names

like if you have an environment named foo-prod, all the things in AWs will be named foo-prod

cool - looking for pointers to give to a dev spinning up aws.

so you don't have like awseb-9832457fnsdjfwertoisghsd names

i can't stand getting alerts on pagerduty with those kinds of eb names

useless

indeed

@dm3: was 10x faster when content-state was hardcoded and static

my state machine is bad 😞

it seems you're waiting for the network call to finish in order to make the first render, no?

yes

@dm3 ^^^

I could do it server side and pass in params to hoplon page?

depends on what you want to do

you need to decide on the pre-loaded state first

yes

What could I do on the client?

(pretending I rely on rpc call)

(no server prerendering)

show "Loading..." state

I do it already

It is slow, but my problem is more, when I interact with stuff like buttons

the delay is huge

because content-state get updated

so you press a button, then content-state gets updated, then you recompute the view-items which takes all the time?

yes

well, if that's really the case - you have to reduce the work you're doing

if the network call eats time - you can only work around it by adding loading states

if it's actual computation - you need to cache/limit the things you're recomputing

Thanks @dm3

I solved the lagging problem 1.4s -> 0.15 s

I create a second cell that has the same data but is not updated

works like a charm!

I now need to find a way to pre-fetch data...

will study castra