@leontalbot: there is the *request* dynamic var i think

one sec i check

yep that should work

@micha! Excellent! Can you provide a minimal exemple so I can figure out how to use it?

could you try putting something like (prn castra.core/*request*) into one of your rpc functions in the server?

it should be populated with the request map from ring

it will contain the client's ip address and all the other things in the request map

(defrpc test! [] (str castra.core/*request*))
(test!)
=> ""

I must do something wrong...

it won't work in the repl

because there is no http request there

you would need to make a request from the client to have that var be bound

@micha You are right! it works, thanks!

Castra is a really nice piece of software

hopefully documentation soon

Count me in if you need help

much appreciated!

Hi all! Javelin question: Given a large data structure, would a change in a very specific and small part of that structure cause reevaluation for all cells that have that large data structure as a source? For example:

(defc data {:a 2
  :b {:c 3 :d 4}
  :e {:f 5 :g 6}})

(defc= bd (get-in data [:b :d]))
(swap! data assoc-in [:f :g] 7)

yes, because data changes

Does anyone actually uses the castra.core/*validate-only* dynamic var? It seems to me it's just some debug helper feature: https://github.com/hoplon/castra/commit/aebc5c37fc8558255c1756c0f6342a5619f84cd0

@dm3: Wouldn’t that mean that ultimately for large projects, datascript is untenable?

define large

@dm3 say… enterprise software 😛

@micha: @jumblerg: would it make sense to hook up the castra xhr responses to some cells earlier, instead of staying in promise land across mkremote, ajax, ajax-fn? we were trying to handle unauthorized responses to castra rpc requests. 1. make rpc call #1 2. it responds with a castra {:error {:message "Unauthorized" :data :unauthorized}} 3. show a login dialog until the user successfully logs in 4. make rpc call #2 (with the same parameters as rpc call #1) 5. return the response of call #2 as the response for call #1 to implement this with promises is quite unreadable and would require to redefine all of castra/core.cljs as i understood.

@levitanong: i’ve been thinking about this. One idea I had was filtered databases.

@levitanong: also, there’s nothing to say you can’t have as many databases as you want

@levitanong: e.g. I’m probably going to have a separate database for “status messages”

one reason to pass your conn around, so you can re-use utility functions for different dbs

https://github.com/tonsky/datascript/wiki/Tips-&-tricks

i haven’t tried this, but a filtered database would change less frequently than the whole db, so if you had queries that you know only applied against a subset of the main db, and they were more expensive than a filter, you could squeeze some efficiency out there

@thedavidmeister: oooh! that looks like it could be useful. I’ll give it a try one of these days. Thanks for the tip

org.sonatype.aether.resolution.ArtifactResolutionException: Could not find artifact com.datomic:datomic-transactor-pro:jar:0.9.5350 in clojars ()

@levitanong: @jumblerg any ideas?

running boot datomic -k XYZ

from https://github.com/tailrecursion/boot-datomic

looks suspicious… i don’t really know anything about maven though

Failed to execute goal org.apache.maven.plugins:maven-install-plugin:2.4:install-file (default-cli) on project standalone-pom: The specified file '/Users/thedavidmeister/estimate-work/datomic-transactor-pro-0.9.5344.jar' not exists -> [Help 1]

hurr

hmm, but https://github.com/jamesmartin/datomic-tutorial works in the repl

Hmm. I do something different with datomic.

(let [uri "datomic:"]
  (d/delete-database uri)
  (d/create-database uri)
  (def conn (d/connect uri))
  (d/transact conn schema)
  (sample/init-all! conn))

@thedavidmeister: u probably want to cd /your/datomic-pro/ and bin/maven-install

oh wait, you’re using pro

this is the free version

i can do either...

if thats the thing which doesnt work then it's because u tried to install it over the net and in your ~/.m2 folder there is some maven metadata file which tells maven to keep trying to get it over the network

if i have [com.datomic/datomic-pro "0.9.5344”] in boot

then where do i find it?

hmmm actually this problem sounds familiar

so first just rm -rf ~/.m2/repository/com/datomic/datomic-pro/

then bin/maven-install and boot whatever and it will work

what i do is have a :repositories arg in set-env!

(conj % ["my-datomic" {
                  :url      ""
                  :username “xxx"
                  :password “xxx"}])

@onetom: where actually is that bin/maven-install if boot pulled it down?

That way, you just get the repo from datomic’s servers

(i think)

@levitanong: which is the more complicated way and u have to inlcude your secret into it and not something u can remember easily

yeah @levitanong i did this

:repositories  [["clojars"       ""]
                  ["maven-central" ""]
                  ["datomic"       {:url      ""
                                    :username (System/getenv "DATOMIC_REPO_USERNAME")
                                    :password (System/getenv "DATOMIC_REPO_PASSWORD")}]]

@thedavidmeister: if boot has successfully downloaded it then u dont need that bin/maven-install

Okay, I think i’ll defer to @onetom in this matter. My knowledge about maven is limited.

ok so basically...

my end goal is just to have some datomic thingo running so i can save some data in it, locally

i was not following the conversation though. i just saw the error message and it looked familiar.

not trying to get anything up in the cloud yet

That’s weird. My :repositories is a function, while yours is a vector of vectors.

so if i do a repl

i can (require ‘datomic.api) and then do stuff

it's a bit off-topic on the #C08BDAPRA channel, btw

yeah that’s fair

I do suppose this could go to #C053K90BR

well maybe i don’t need boot?

i don’t know...

i was hoping to just have castra talking to datomic

as an mvp

if you are happy with trying things our with an in-memory db first, then u don't need anything else just the free version of the datomic client because it contains the "in-memory transactor"

yeah but i don’t know how to set up either

@onetom: the *validate-only* var is what i use to do serverside validation

some tutorials say one thing, others say another

@thedavidmeister: do u want to do a screen sharing session? i can help quickly

so if i have castra setup, will the in-memory db work between ajax calls?

thedavidmeister: yes

oh ok

well it’s probably working enough for now then 😛

@onetom: when you bind that to true in the client and you call an rpc function the backend rpc fn is called, but the body of it is not evaluated (only the preconditions)

when do i lose the memory db?

@micha: i saw the code, but i dont understand why would you want to do such a thing

@thedavidmeister: when u stop your jvm

i have a state machine that handles interaction with the backend

and forms in the client

the state machine can be in "validation" mode

@onetom: oooh ok, so as long as i have serve running with boot, i get the same db, so castra will work. That sounds good enough for experimentation for a bit

like if you submit the form and the server rejects it becaue of validation fail

@onetom: the crucial thing is that when you're validating input you must not evaluate the body of the rpc fn

@micha: i see

hi a question around focus, i have a use-case where users are editing a dom of code, i want to move focus around the dom. two ways I can think of to do this, 1) have a toggle of focus on each node, 2) after an update set focus explicitly, is there a ‘best practice’ approach?

so you are using it to reflect validation errors during form fill?

because sometimes you just want to validate input that the user is typing, but the user doesn't want side effects to happen

like the user isn't submitting the form, so you can't transact on the database

but you want to have the server validate what they have in the incomplete form

yeah it's mostly used like that, during form fill

why don’t you save the user input always, but have a way of knowing whether the input is “safe” or not for general usage?

we don't want to change the database unless the user intends to submit

that would be very confusing for the user if we were accepting input in the backend as they type

why would the user know what you put in the database?

well suppose they're running ad campaigns at adzerk

when they start seeing those ads show up on their website they know it got into our database

oooh, i mean, have an extra thing in the database

that could be very bad news if it happens accidentally

a place to put raw input

that the user doesn’t see

until they make it match validation rules

hm

currently that place is in the cell in the client

in the form state machine thingy

oh, but you do validation server side?

why would i want to save that in the database?

yes all serverside validation for now

there is a hook for having clientside validation but we don't use it

i dunno? i’m literally about to try this approach, i’ll let you know how i go 😛

because serverside validation is free and no perf impact yet

@micha: and what do u think about the automatic unauthorized response handling in castra?

the nice thing about serverside validation is that it's automatic

@onetom: what is that?

sorry a lot of stuff in the scroll buffer

well, what if a user was struggling with the validation

if you had the raw input in the db

@micha: i described it just below my question about the *validate-only*

your support staff could see the problem straight away

that’s one idea

if you rolled a new release that relaxed validation rules for something that was previously invalid, maybe you could retroactively allow things already in the system

@thedavidmeister: thats how @jumblerg is planning to extend hoplon. storing the whole frontend state in datomic so you can replay or join any user session any time

if people were consistently failing validation, you’d have data to do UX review on

@onetom: i’ve been using datascript for front end state

@onetom: it’s OK, but for my use case i’m finding that having a few different dbs and cells in different ns is better than one big one

@onetom: i don't understand what you mean with the two requests above?

like returning the response for #2 to #1?

and the connection to promises?

@thedavidmeister: i see whta you mean now with saving in the database

that would be pretty sweet, and you could activate that behavior in castra when validate-only is bound

if you're using datomic i can see how it would be pretty awesome to have that

unfortunately i don't think we can use datomic

i mean at adzerk

oh, fair enough

i’m playing around with this:

(defprotocol ISafe
  "Safe values can be extracted from raw user input"
  (raw    [this] "The raw user input")
  (safe   [this] "The safe version of whatever the user input")
  (raw?   [this] "Is there any user input at all?")
  (safe?  [this] "Is the user's input safe?”))

then an input on a form would have something like

(h/input
                :type "text"
                :data-raw (j/cell= (item-data/raw? v))
                :data-safe (j/cell= (item-data/safe? v))
                :value (j/cell= (item-data/raw v))

i was thinking like middleware perhaps

sure, i’m just messing around with it client side

in your :rpc/pre for example

you could have middleware there

like for instance:

(defrpc foo [{:keys [bar baz]} :as args}]
  {:rpc/pre (validate (record args))}
  ...

where record puts the things in the debug database if *validate-only* is bound

yeah totally

you could see what people are constantly getting wrong

yeah

that's aweosme when you can check out a copy of the database that is rewound in time to when the validation failed

when you don't have datomic it gets less useful i think, but still probably worthwhile to have

how come you can’t use datomic at adzerk?

is there a problem with it?

it's closed source and the licensing is complicated, which scares people

postgres or ms sql server is a less stressful choice for such a core thing for them

ah, so it’s byo db for your customers?

like how does datomic work when you have 150 peers?

no it would have to be a core thing for our business

i mean we'd need to store mission critical stuff in there

i thought the whole thing was you could have as many peers as you want?

the most you can buy is 8 peers last time i checked

then you start negotiating

oh sure

yeah i dunno

the close source part is the real problem for us i guess

assuming the cost for the number of peers we need would be reasonable

like what if cognitect goes under or what if there are intellectual property issues with datomic

in the contract it says they can pull your license with 10 days notice

we'd basically go out of business if that happened probably, if our whole thing was relying on it

yeah

that’s fair

so who is paying for it then?

i'd be very interested to talk to them, yeah

there must be some huge enterprise customers for it

hmm

I don't think they'd pull a license that way

i wonder if it would be OK to just stick entities somewhere

like, in rethinkdb or something

have you ever heard of that happening e.g. for Oracle/other db providers?

yeah there have been incidents, i forget the most recent one

some nosql database went belly up and disappeared i think

@micha: yes, 2nd response (if succeeds) would be the response of the 1st request. im just talking about retrying the rpc call, which would be "interrupted" by a login screen.

@onetom that seems like something you'd want on the frontend, maybe

not the backend

you could implement retries there, in the state machine

yup, im talking about xhr calls on the frontend

having a staet machine that mediates interaction with the backend is pretty key for that

you have mentioned it a few times but not sure what do u mean by that

last time u said u will show an example later but we never got back to it

ah yeah i started making a demo

then i got sidetracked trying to find better abstractions

what i have currently is a somewhat naive implementation that works

but doesn't seem as elegant as it should be

but the state machine representation problem is pretty deep i think

if u could share it i might be able to think it further

yeah i'll do that today

i went downt he rabbithole of UML statecharts and such

lol

publishing the code will be the best first step i think

@dm3: http://www.wired.com/2015/03/apple-pulls-plug-tech-company-runs/

the doomsday scenario is somebody buys datomic just to shut it down

yeah, heard about that one

haha "innovation"

my nodejs application with 10000 alpha release databases

my app to put mustaches on phone selfies

we must protect innovation

ah yeah, foundationdb

i was using that

luckily, not for much 😉

is apple going to buy adzerk and take away my hoplon?

haha it's open source and you have commit access to the repo

so i don't think that's possible

it's also a bit simpler to reimplement once you are familiar with the ideas

that’s fair

i am definitely not able to make database software

once it's open source you can fork it

it’s ok, i have other skills

the license you have is good forever

so if the hoplon repo were shut down you could just make a fork and coninue

yeah i’d fork it

the problem is when you don't have the source

i’d rather not have to though

well, if you consider a scenario where github decided to remove every fork

but that's probably fantasy

yeah you'd need to use bitbucket then or your own repo

that’s true, github is buyable by apple too

but you could stll do it

if you had the source

lol “apple bought github to shut down all competitors everywhere"

:)))

imagine tomorrow you find that notice on http://github.com

haha

yeah i have local copies of everything

that would be a bit worse than pulling an .isArray package from NPM

it would be widespread chaos

lol it would be like the dark ages

everyone computing in the dark in their hovels

haha

roasting human flesh on spits and drowning witches

If I was running Apple, I'd probably buy bitbucket, gitlab and github and shut down just for laughs

haha right

it's now on iTunes

lol

sign up fuckers or die

0.99 per repo

you would need to have it approved before you can push a commit

they might put a nsfw tag on it

anyone here used rethinkdb before?

looks like they have a query powered data streaming thing

wondering how hard it would be to plug into javelin

if it's event-driven you can just attach events to a function that does swap!

yeah i don’t know if that’s exactly how it works

i’ll just read up and play around

changes
(changes table)
Return an infinite stream of objects representing changes to a table or a document.

i think that’s where i’d do it

but actually, that happens server side

maybe castra is not the right thing here

having issues with cond-tpl, the test isn't firing

I have something like: (cond-tpl (string? some-cell) ...)

is that allowed?

@l1sp3r: i thin kyou need (cond-tpl (cell= (string? some-cell)) ...)

(string? some-cell) is always false

without the (cell= ...) wrapping it

Ah makes sense, thanks!

@micha quick question

With (env/def KEY "value") "value" will be use as KEY key only if there is no environnement variable set-up, correct?

yes, it's a default setting

thanks!

Hey guys! I just started with Hoplon and I really like it! I'm using the default template from the tutorial (`boot -d seancorfield/boot-new new -t hoplon -n address-book`), but somehow there's no output generated when I run boot prod or boot hoplon. What am I missing?

@beatngu13: hi! Temp solution: If you want to test hoplon, you can also do lein new hoplon address-book

You can then run with boot dev