clojure-uk 2016-08-09 | Slack Archive

@otfrom: currently i have a version 3.7.0 pulled in by onyx through ZK, and a 4.1.0.CR3 by aleph... they actually seem to work together, so i'm tempted to leave it, because afaics i can't solve it without removing the aleph dependency (which is only for an http client in this case, so doable if really necessary)

mccraigmccraig13:08:12

also netty changed their "all" dep from io.netty/netty to io.netty/netty-all which is the only reason i could have two versions around since i always use :pedantic? :abort

mccraigmccraig13:08:28

whatever the weaknesses of node's archive or package deduping or lockfile, their runtime solution is so much better than this "find the golden version combination" hunt which java/mvn forces on you

mccraigmccraig14:08:43

@otfrom: datastax do the right thing with the cassandra java-driver - they provide a version with a "shaded" netty i.e. a netty which has been re-written to a private package https://github.com/datastax/java-driver/tree/3.0/manual/shaded_jar

mccraigmccraig14:08:59

but the smell of eau de hack is very strong

glenjamin14:08:51

I don’t think I’ve yet seen a package manager that contains all the best features around

glenjamin14:08:19

allowing differently versioned dupes in subtrees, allowing parents to substitute, providing lockfiles for the whole tree are the big ones imo

elise_huard14:08:08

@mccraigmccraig: wondering if you have any views on what the best way is to talk to cassandra using components (session components) - we (Mastodon C) have an old project lying around which uses a protocol on the component itself for execute, prepare-statement, execute-prepared and a with-session wrapper which does some clever things to pass in the session object, but not sure what the best practice is nowadays (first messy bits of our project here https://github.com/MastodonC/kixi.heimdall/pull/1)

mccraigmccraig14:08:47

@elise_huard: we use this https://github.com/employeerepublic/er-cassandra/ for all our cassandra interactions ... there's a high-level session protocol, and a low-level session protocol

mccraigmccraig14:08:30

we don't really use prepared statements - there's only really an advantage if you are doing boatloads of the same statement (and then there is a massive advantage) but our statements are very mixed

otfrom14:08:42

mccraigmccraig: I think this auth(n|z) service will hit the same query reasonably frequently.

mccraigmccraig14:08:55

we pass the session explicitly - our system is entirely async, so with-x doesn't work so well

elise_huard14:08:39

mccraigmccraig: thank you, I'll have a look at the code! inspiration is welcome

otfrom14:08:54

thx mccraigmccraig

mccraigmccraig14:08:24

yw @elise_huard - it's very much a work in progress, and the testing stuff isn't really there yet... soon though 🙂

otfrom14:08:29

hugs

otfrom14:08:55

(っ⌒‿⌒)っ

mccraigmccraig14:08:38

oh, and the README is radically out of date 😊

elise_huard14:08:16

ok 🙂

mccraigmccraig14:08:59

feel free to ask questions... the general idea is that the model level stuff allows you to automatically maintain a primary table, secondary tables, lookup tables and unique-key tables (lookups with a unique constraint) for a model entity type - you get select/upsert/delete for entity maps, and the lib does all the required inserts/updates/deletes of all the different tables (including unrolling collections for lookup tables), runs LWTs for the unique keys etc

quentin14:08:30

erc? is it a new implementation of slack?

mccraigmccraig14:08:50

uh i got a (っ⌒‿⌒)っ @otfrom - looks like it's working

mccraigmccraig14:08:18

@quentin: erc is an emacs irc client

quentin14:08:39

of course it is 😮 (damn emacs really got everything 😄 )

minimal14:08:48

emacs only comes with 3 irc clients IIRC

minimal14:08:10

oh “just" erc and rcirc built in