@numberq: server errors also happen if your server isn't running properly. Have you tried a REPL and evaluated your whole API? Usually that tells me that my database initialization went wrong.

if I have (defc foo 1) is there any way to get at the foo cell inside a cell=, or can i only get at the value 1?

not that I know of. Why would you want to do that?

cell= derefs the cells used inside

you can have nested cells though

although it then becomes very hard to reason about the program (and easy to have memory leaks)

because i have a datascript connection as a cell

and i want to pass the connection around rather than use a global connection or pass the db

having cells kind of presumes that you have a root cell somewhere with the value that gets set there

so for a cell-based API I'd expect to have a cell with the value of the db

yes

that’s what i did

that’s what the problem is

why is it a problem?

because if the value of the cell is the db

that means the cell is the connection

so if i want to work with the connection, i can’t just have the db value

like, if i want to do (d/transact! conn tx)

you want to do that inside the cell?

cell that reacts by transacting?

I have conn a separate variable, actually a Mount defstate

well no, i’m not doing that atm

but i am running queries against it

then I have a root cell that is set to :db-after value on every transaction

i wanted to have a consistent usage of conn throughout the functions i wrote to wrap my queries

rather than passing a db for reads and a conn for writes

that seemed weird to me

I wouldn't pass anything

just have one var with a conn and a cell with a db in a db namespace

then say (db/transact! datoms)

but if i pass the conn around, i can easily write tests

using with

and make a bunch of different conns on the fly during the tests

you can also do with-redefs in clojurescript I think

I mean you can probably do it your way, only it doesn't really play well with the concept of cells

yeah, i have some complicated logic resulting in a transact! but the actual transaction is pretty simple

so i have a test where I pass in the situation(s) that would result in a transact

then compare the conn after the change, to a fresh db using with to get at the end-state

i might try just wrapping the conn cell in another cell, for usage in cell=

hopefully that doesn’t break everything

(with-redefs [db/db (cell (test-db))
             db/conn (new-conn)]
     (logic!)
     (is (= @db/db (db/with (test-db) ...)))

only nicely wrapped into some helper function

I'd have to do it like that if I wrote any tests

haha, i like to refactor

so i write tests

I like to refactor

so I don't

(I do write tests for scary parts though)

i’m always surprised by what i manage to break

especially when I’m sure I won’t break anything

the only reason I picked up on this issue that I’m now asking about is because of a test i wrote 😛

@dm3 maybe it’s not so bad to do db for reads and conn for writes

i think the point you make about transact! inside cell= is pretty much right

i do want to keep the functions that work on conn and db only working with the arguments they take

i can already see that i might want two or more different dbs to track the state of different things

so if i do that, i might also want functions that can operate equally well on the different dbs

I find that the best structure is when you have ns'es that encapsulate your state and logic fully from the views

so you have you defelems and whatever which make calls to users/..., orders/.. both for queries and for "transactions"

(ns cart.view
   (require [cart.state :as cart]))

(defelem cart [...]
   (div :click #(cart/remove! @%)
       (text "~cart/status")))

yeah, i’m slowly working towards that

i think i might need one state that persists and one that doesn't

for tracking things like “keys pressed"

you can have that as cells in the view ns

or locally to the element

I have a bunch of states managing different concerns, some of them in datascript db, some in cells

very easy to reason about if all of the logic is contained in a single ns

yeah

i’ll try it out

but i suspect it will be good to have a “state that persists” which is basically the user’s data

and “state that tracks transient things” which is the user’s input

which would actually be bad to accidentally persist

@dm3 i got inspired by the select-id cell from the other night 😉

i want to take this (and (not @id) (which-key/pressed? :backspace %)) (reset! select-id (-> @state:sorted last first)) from the keyup

and turn it into something that just uses the information about what was pressed from an existing cell

but I don’t want to put too much in the one PR >.<

but yeah… it helped me write some tests that i hadn’t figured out how to write previously

i didn’t know how to easily get at selected text, or the cursor position in selenium

but writing tests against the value of a cell is easy

endorsed Hoplon on HN https://news.ycombinator.com/item?id=11186983

what are Om and Reagent like

have you used them?

yes, they're like biting your nails off when you need to integrate third party JS

also a lot of ceremony around component states

and lifecycles

Reagent seems simple but you have to know there are 3 (!) forms of components that you can create

even though the difference is undiscoverable at first sight

the only tricky thing that Hoplon has is loop-tpl. React + Reagent/Om has dozens

(I mean tricky as in the effect of the code on DOM is hard to reason about without knowing the implementation details)

hurrr

considering i’m learning cljs as i go

glad i chose the least complicated option

I think React is only needed if you're planning on using React Native

or you have a huge team and then React+Om.Next gives a nice UI + server architecture

out of the box

i do not have a huge team

with Hoplon you have to kinda discover it yourself

there is a lot of discovery going on, lol

but not sure how much of that is me and how much is the framework

what are the advantages of the different components in reagent?

not the advantages, you have to know that you need to use one way over the other in different contexts

https://github.com/Day8/re-frame/wiki/Creating-Reagent-Components

like look at the last point: https://github.com/Day8/re-frame/wiki/Creating-Reagent-Components#appendix-a---lifting-the-lid-slightly and compare with what you get in Hoplon - a function that will create a DOM element. That's it.

is Hiccup a templating language?

wah, that refactor just took me 13 days to wrap up but fixed 4 tx, two of which i had no idea how to tackle otherwise, and paves the way for another 5 tx that i didn’t really know how to do either

epic

thanks again for helping me @dm3

no problem

i’ve been trying to show people hoplon

but it’s really hard to explain to someone who doesn’t know any clj

@thedavidmeister: re: cell w/ db in it, have you considered using a lens?

it exists for the case when you want to export a cell value and also export a particular way of updating it, without exposing too much about what contributes to the cell value

@alandipert: i haven’t looked at lenses yet

i need to learn how they work first

hehe of course 😄

if you think it’s worth looking at, i’ll set aside some time to investigate

I think in this case instead of transact! you'll be able to (reset! db txs)

right - the lens cell can have the db update logic inside it

@dm3: thanks for the kind words on HN

hmm, so i could make a lens version of transact!?

and just work with the db everywhere

and not worry about the conn?

unless you need to listen! and such

i could still listen to the conn if i referenced it directly?

or once i make a lens-transact! i am stuck?

yes, the lens would only proxy the transact! call

the particular case that lenses address are situations like these

@alandipert - someone needs to promote the work

ok, from what i read, i might need listen to get the data into something like datomic later

(def root (cell {:x 123}))
(def num (cell= (root :x)))

(defn inc! [c]
  (swap! root update :x inc))

so imagine we created num because we only want to expose :x to some consumer, or component within our app

because the more about the structure of root we expose, the more coupled to it dependent cells become

so num is a "line in the sand" so to speak

yeah, for me, num would be the query i’m running

cool

so where num is this effort to narrow what's exposed to consumers, inc! stands in contrast

as it takes the root cell

so the way a component would view and a way it would update are not symmetrical

yeah

that’s exactly what i ended up with

but in datascript language it is conn for write and db + a query for reads

i didn’t mean to do that, it just sort of happened

yeah

i think it’s ok

i only have one query and one loop-tpl atm

so i’m kind of over-analysing 😛

i’ll let you know if something comes up that i didn’t think of

cool

it's really awesome what you're doing btw

i'm not aware of anyone going as far w/ the datomic ideas and cells

haha, it’s only because i suck at clojure

i think i’ll find it easier to deal with a db

than mess around with raw clj data

yeah

that's still mindblowing though

building a reactive network of databases... in the browser?

❗

i would also refute that you suck lol

ok, well let’s say that clj is my least-good language >.<

@dm3: have you used reframe btw?

yes

I also built something like re-frame with Om before that

but it sucked

and I now have a weird mix of Javelin-based state, Datascript and event-based sync with backend

in a Hoplon app

what is the difference between "Reactive-Atom” and a cell?

so i’m reading https://github.com/Day8/re-frame

Hoplon gets mentioned a lot

but I don’t understand the problem that Hoplon has that this is solving?

doesn't use React I think

and uses weird build tooling (boot)

so they picked a framework, and it made a problem

I think the implicit argument is that they "need to" use React for whatever reason

and the framework “at it’s best” is close to Hoplon

and the solution is… 20 pages of stuff

to explain

maybe they did hit a perf bottleneck with Hoplon at the time the re-frame was born

and this diagram

is that common?

perf issues in Hoplon?

I haven't, but I've heard Colin Yates complain when he added ~20k things to his app (whatever those things were)

didn't work with Hoplon, worked with React

i don’t think i’ll need 20k

is there any low hanging fruit for perf in Hoplon

or is it about as good as it will get?

@alandipert: do you have any feedback for me on https://github.com/hoplon/hoplon/issues/85?

@alandipert: happy to make changes, but i’d need some guidance to go much further than i did on the PR

@thedavidmeister: i think there is a lot of opportunity for perf improvements

we haven't put much effort there so far

@micha: ok cool, good to know

have you run into anything?

not at all

like have you had to work around any performance issues?

actually, i keep expecting to, but i don't

i’ve had to work around race conditions that were really confusing me

haha yeah that's why we haven't spent much time on profiling

but the solution was to just keep making things more abstract

until i could no longer reproduce the issue

not having perf optimizations allows us to keep the code more high level

so it's easier to add features or change things

i’m a big fan of knowing what you could optimise, but only optimising once you have a problem

the only perf optimizations i've really done is to remove little used features

that were part of the hot section

perf stuff often makes things complicated and buggy

like keeping the number of things hoplon needs to do when the page loads to a minimum

yeah exactly

i think javelin is pretty much stable now, so that might be a place to look for perf

we haven't added or removed anything in over a year i believe

maybe more than that even

that sounds pretty stable

to the point where the repo kind of looks abandoned 😛

that's how we like it

that means it's well factored

like i think the priority-map implementation could perhaps be shaved a bit for perf

that would speed up javelin, and hoplon would benefit from that

yeah, that’s the way all projects probably should work

but i gotta say it’s weird

haha

i have a perf test harness for javelin i think

ah, i found something to nitpick

you could have a cache in your travis setup

it would go like 10s faster

sweet!

what's your github user?

thedavidmeister

i just sent a github invite

sweet

i should put up a PR then

actually… it’s nearly 2am

i should go >.<

talk to you later!

catch

you should have access to push travis changes etc

awesome, but i’ll still test on my account before i change anything

feel free to push changes to those repos, just things that add or change the api maybe should be done with community input

but like tests, CI, things like that, fire at will

I do actually want to stick up a bunch of basic tests for hoplon

like, just checking that all the elements actually render the right HTML

mynomoto made some tests i believe

on a particular branch?

he devised the setup for doing the async tests etc

in a separate repo

aaah ok

but then how do you review the tests before rolling a new release?

the cljs testing stuff has improved a lot since then

you’d have to go back to the other repo each time

it was a proof of concept

ah ok

might be useful to look at though

yeah, i’ll look at it when i have time for sure

it’s good to get in the habit of writing tests as you find/fix bugs, at the least

even if you don’t go for comprehensive coverage, at least you know the stuff that tripped you up in the past won’t come back

yeah javelin has hundreds of tests

awesome

we haven't had any regressions that i can remember

the way i see it, is writing 10 tests upstream saves 100 tests in my projects 😛

yep

and browser testing stuff is tougher than just asserting data out of functions

and the model is more open ended too

so i’d like to know the browser stuff is rock solid from the framework, and i can just test my data in my project

javelin is all factored out, so it's easy to test it

yeah, it’s the right approach for that

hoplon is a little more vague because there is setup involved

and setup phase uses the things you're testing

mmm

i guess it depends what you’re testing

we can definitely do it tho

making sure that (div) spits out a div and not a dov

that we can lock down easy

hah yes

a macro could emit all of those tests, too

so we'd only need to maintain one test for all the elements

that sounds cool

you’ll have to show me how to do that

i haven’t written my own macros yet 😉

just been using regular functions

basically if you find yourself cutting and pasting code in the editor, a macro is an option

because macros generate code for you

ah ok, i copy and paste data, but not code

i usually refactor code

yeah like for tests:

(emit-primitive-tests div span a p ...)

that would expand into all the individual test clauses

ah yes

i often split my tests into 2 functions

1 just holds all the setup and expected outcomes

yeah i think in this case you could use a funtion actually

and the other does the “stuff"

ah no

because you need div as both a symbol (so you can read the name) and as a function

like if you had a function that would loop over the things and test them you'd need to do:

(test-primitives
  div 'div
  span 'span
  ...

which is still not bad

but a macro could do that for you

ah i see

most of the tests i’ve written were in PHP

you probably don’t want to know how it works there 😉

oh i know, unfortunately

https://github.com/tailrecursion/wigwam

although i can't say we had too many tests for the 250K lines of PHP in that application

maybe 2 lol

hah

i had to make my own repl

you can just run your strings as functions

there was no php repl that worked

aw, i use psysh at work

it’s pretty good

not sure how old it is though

when i looked around none of them could handle global variables correctly

because they all evaluated code in a function context

because of the way they did the error catching

my repl works at the global context

the app i was working on had tons of global variables

and require() and so on

ah, well...

thankfully, if you ever find yourself back in PHP world, there’s a lot less of that going on now, with PSR and composer being common

and there are REPLs that work

i remember looking at the facebook repl

it was the worst

http://psysh.org/

and i realized that facebook is populated by true php programmers

they don't use anything

it was the most naive thing ever

hah

then they went on to re-invent php7, before it was released

fb do weird things sometimes

yeah i have not seen a single thing from them that isn't terrible

i can only imagine the environment their developers work in

hellish

i had enough clients get mad at me over the years because fb silently dropped/changed/added/blocked something i was using...

and their docs always are wrong/missing

lost all faith

they remind me of like the "chavs" i think they call them in england

lol

well, i barely get to write code at work any more

and all my evening stuff is in cljs

so i might be slowly kicking the PHP

https://s-media-cache-ak0.pinimg.com/564x/c1/30/67/c13067da6da379571bc2a0aedb8653b4.jpg

lol meet the fb team

lawl

repulsive

on that note, i really have to leave

2:30am now

later dude!