Fork me on GitHub
#clojure-uk
<
2017-08-03
>
maleghast07:08:26

I am afraid I had gone to bed by the time you said that ^^

seancorfield07:08:51

And I am in bed now! Time zones suck 😸

maleghast07:08:11

Well, yes, they do, but they are also quite useful.

chrjs08:08:36

I gather you mean Sparkling as a noun, not a verb.

Rachel Westmacott08:08:36

…or an adjective

agile_geek09:08:39

Yep. I don't sparkle in any shape or form

maleghast08:08:52

@agile_geek - it’s only a complete waste if you’re not being paid… 😉 (Srsly I do get how frustrating this kind of situation is)

maleghast09:08:38

Is “Sparkling” the new hotness re Clojure + Spark?

maleghast09:08:57

(I thought it was all Flambeau or somethin’)

dominicm09:08:32

https://github.com/HCADatalab/powderkeg is very good, it works with live coding. Huge improvement over Flambo/Sparkling.

dominicm09:08:43

The difference between Flambo & Sparkling is much of a muchness tbh.

dominicm09:08:51

powderkeg is unfortunately not as mature (yet)

geek-draven09:08:58

will have to have experiment with that next time I have a spark job, Sparkling is my usual go to

malcolmsparks10:08:34

Sorry for this rant. But I used to love the BBC and what it stood for (a very long time ago, granted). If the reason for demanding a login was for user benefit, surely that's an opt-in choice the user can already make. Why can't they be honest and say they want to spy on their users?

maleghast10:08:43

@malcolmsparks - I would have to guess that the large proportion of the BBC does not want to (spy on their users) but that there are those who are responding to commercial and market pressures in a short-term-ist way without considering the long view of how it will affect trust between the public and the BBC apart from to avoid being nakedly honest about it. I’ve known a bunch of people involved in the “my BBC” and “BBCiD” teams over the years and privately a lot of them have expressed a feeling of complete irrelevance, that “the Corporation” wants personalisation, because “that’s what websites / companies on the web do” and not for any reason like “it’s what the audience wants and here’s the numbers”.

rhinocratic10:08:59

I feel your pain @malcolmsparks - I used to work for their Open University Production Centre many years ago, and it was a great place to be. However, I’ve seen few positive changes since the rot set in with John Birt’s appointment. 🙁

maleghast10:08:14

So… Does anyone in here have any strongly held opinions about message queues, Clojure and using them on AWS? I have some (positive) experiences of using SQS with Python, Ruby and (gasp) PHP, so I am reasonably happy to just default to SQS if__ I decide I need to offload some work from the app I am currently working on… Still I don’t like simply defaulting to a decision. I know that “other message queues are available” and so I was wondering what people have to say on the issue. Thanks in advance! 🙂

jasonbell10:08:13

@maleghast Sparkling is nice, Flambo is nice. Both have plusses and minuseseseses.

maleghast10:08:09

@jasonbell - are you answering my Q from up there ^^ that led to the “discovery” of PowderKeg?

mccraigmccraig10:08:40

@maleghast if you use something like onyx for processing off of your queue then it won't matter much whether you use sqs or kafka or whatever

maleghast10:08:43

@mccraigmccraig - I have been looking for an excuse to learn / use Onyx for a while, but it feels as though it would be a nuclear warhead to crack a walnut in my current use-case and I don’t want to incur the infrastructure cost(s) of putting up even a small onyx cluster at the moment, but I will bear that in mind… I was already ignoring Kafka, for this, as order is unimportant to me and my understanding is / was that Kafka’s real selling point is order preservation..?

jasonbell10:08:55

@maleghast yeah, powderkeg is on the to-try list

maleghast10:08:11

@jasonbell - Yeah, for me too now 🙂

mccraigmccraig10:08:01

if you are looking to remove the need to deploy a cluster then you probably want to use something like lambda fns to process your sqs messages then @maleghast ?

maleghast10:08:00

@mccraigmccraig - I have been thinking about that, yes… I like Lambda. I’ve only used it in Python and nodeJS, but I imagine writing Lambdas in Clojure is well supported by now…

maleghast10:08:41

@agile_geek - How can we help? (No, I am not taking the piss, I am seriously offering help)

chrjs11:08:45

@maleghast no strongly held opinions at all, but SQS has been very straightforward with Amazonica.

maleghast11:08:56

@chrjs - That sounds encouraging, thanks 🙂 I am a little intrigued by Squeedo (https://github.com/TheClimateCorporation/squeedo) which builds on top of bandalore rather than Amazonica, but I think that if I do go the SQS route, I will use Amazonica to begin with, for a nice “simple” approach.

mccraigmccraig11:08:33

@maleghast you can also do your lambdas on node with clojurescript - https://github.com/nervous-systems/cljs-lambda

maleghast11:08:41

@mccraigmccraig - Ooh, that’s interesting…

maleghast11:08:49

I will have a squint at that too.

maleghast11:08:20

That’s very cool 🙂

maleghast13:08:47

@dotemacs - I won’t, now__

mccraigmccraig13:08:36

@dotemacs does macchiato have to make similar changes to the ring-handler assumptions as yada does, to cope with responses often being promises of a response in js-land ? (yada lets you return a manifold Deferred response as well as a plain value)

dotemacs13:08:28

@mccraigmccraig just playing with it now, not delved into it properly

Rachel Westmacott13:08:37

if I was looking at AWS lambda I’d definitely try cljs before clj

maleghast14:08:34

@peterwestmacott - Yeah, I think that I would too… 😉

benedek18:08:06

jvm is still faster re aws lambdas I think. altho can't point to any hard evidence

dominicm18:08:15

I think with lambda you can use appcds method and have better performance than cljs on node, both to start and to run

glenjamin20:08:57

it might be faster in some scenarios, but its definitely slower to boot up

glenjamin20:08:19

and the jar you have to upload is much larger than the equivalent compiled cljs, i’m told - which can affect deployment speed