So my take on the component/di thing is that in the systems that I've been working on, a lot of the interaction with outside systems happen over http, which means there is no state in terms of connections that need to be handled. Like when you're talking to aws stuff, you don't have a "connection". The only connections that come to mind in our current system is redis, the dbs, and probably one or two more that I can't seem to remember. Now, I believe that none to few of the connections which we need in our app are actually dependent, so it doesn't really matter if I start redis before the dbs or vice versa. Also, wrt the db's since we're so lucky as to be using mongo, we've implemented multitenancy as one mongo-db pr customer, which means that we don't know which dbs we'll need on system startup.

So for all intents and purposes, we (of course we're using component, why wouldn't we) end up with something like (assoc req :system (assoc @system :customer-db (connect-to (:customer req))) in a ring middleware somewhere.

And with that, I wish you a good morning!

Guten MOrgen!

Morning

mogge

Can I just say that I really like that fact that this channel has become a community in the last year or so. Talking all kinds of Clojure stuff, loads of other things and lots of people coming in each morning and saying hello!

So a big thank you to you all! 🙏

morning :)

Morning!

@thomas I can really second that. And I appreciate the "takhøyde" as we say in Norwegian in here. There seems to be more room for discussing off-topic-ish, "stupid", or philosophical Clojure questions in this little channel than other places.

I'd say this is almost like the hallway track at a good Clojure conf 🙂

thank you @slipset , that last comment sums it up perfectly.

I wonder if you could run a conference of only clojure philosophical talks (no exp reports, etc.)

I would love that.

Do them in a discussion style, 20m to present an idea, 20m to answer questions / discuss it.

Well, in the virtual style you could even pre-record the talk, and have the speaker answering questions and discussing it as it was played.

Presented at ffs-conf the one year it was run in London. It was the coolest conf I've attended so far.

Morning

https://ffstechconf.org The only downside to it was that it was https://en.wikipedia.org/wiki/Chatham_House_Rule, so no recording and no live-tweeting

But the format was exactly like you suggest @dominicm 20 minutes rant, 20 minutes discussion.

Dia duit!

Morning! I have to alternate between playing PM and figuring out what other people should do, and the urge to sit down at my editor and solve things myself and it’s driving me crazy :) anyone with similar experience could offer some tips?

I would gladly rant about this at a conf setting BTW :)

Give in and just code 🙂

Working with developers is like herding cats.

One of the guys at my work transistioned to being a developer to the "development manager". I don't envy his role.

saying "morning" is a bit like having clojure dojos. It gets everyone in the same place with similar interests having fun and getting to know each other. That just makes more chances for good things to happen

@dharrigan I do code, quite a bit :) in terms of making things happen though a team is much more effective and resilient. It’s just that you have to repeat your point and direction every week until other developers cave in and accept it :)

I guess it's harder now? Do you all work remotely?

Yeah everyone is remote now. Three-dev team, so not too large but still I’m having trouble breaking down tasks into something that’s both not too trivial or too daunting. I have the most context and information so I have to do a lot of writing and talking.

try to find the fun bits in people management

The most gratifying part is seeing other people sink their teeth in a hard problem (be that UI, ops, whatever), seeing them do a good job — and realizing that that’s one problem I don’t need to solve myself 🙂

that is the nice thing about it. That and helping the team and team members be better than they were before

@orestis I think the main responsibility for a tech-lead/product manager is to create an understanding of why something is being done, and possibly also why it needs to be delivered at some point.

If you can establish this understanding, ie the intent, and you've hired reasonable people, they will deliver the stuff you need, not necessarily the stuff you figured you should ask for.

IMNSHO, it makes for a much more interesting place to be a dev as well.

Definitely @slipset - I sought that behavior and created it in the current workplace. The context is a very small startup with a lot of baggage (operating for almost 10 years now with established product and clients). The trouble with communication is trying to figure out how to tune sender and receiver.

Also another issue is scoping and time boxing. As engineers we tend to fall into rabbit holes and it usually takes outside intervention to get us out - and some times in my experience it has to be an “order”, otherwise the calling of the unsolved problem is too strong, even if the solution isn’t relevant to the business any more.

Perfect is the enemy of good, as a bike tuner once told me in Belgium :)

@slipset I agree completely with that

https://twitter.com/JenMsft/status/1320436012405850113

Lol

I'm still of the opinion that mongodb is no worse than mysql used to be

That comic made me sad. There's too many people genuinely doing that.

So much complexity and "hacks" which we celebrate.

Go’ aften!

I'm sure a lot of people on http://dev.to would post that "ironically", while they actually do exactly that and are completely oblivious to it. That site really was an eye-opener for me to see what the dev world has turned into...

morning

you need to start getting a life @raymcdermott ...

This is at best UGT in your parts of the world

I start work later these days so it is officially morning for me too

Hmm, so morning is when work starts? or have you shifted your days around?

I'm doing life in the morning and work afternoon / evening ... for me it's a big lifestyle improvement tbh

also aligns with some of the US colleagues at work

I’m thinking of doing the same - moving back to Copenhagen where the daylight is precious. Not sure how I’m going to manage with baby-related sleep deprivation though.

Lots of vitamin D. I live close to Malmö, 2 kids under 4, I know the feeling.

Enjoying parental leave now, working 50% for a bit (afternoons)

first proper, non-trivial use of clojure.walk/prewalk today - quite the thing

it I mean, not my code

beware of https://clojure.atlassian.net/browse/CLJ-2568

Do you mean btw, that postwalk is more common in your code than prewalk?

I think I found a use for walk recently to optimize something, but I'm using postwalk in a weird way instead :)

I only used prewalk - yeah I saw that issue but it's doesn't concern me. I'm very direct 🙂

I usually use postwalk as a search and replace in nested data structures. I also use it quite a lot in clj-kondo and sci, but more in the spirit of, not the built-in implementations.

yeah it was for finding paths after a search ... for making what I thought was a simple game (checkers / draughts) with my son, not work

The metadata thing has only tripped me on a handful of uses of walk.

Hmm, I think they're all library uses as well. So probably something in that.

sci leans quite heavily on metadata so that issue became apparent to me quite early in that project. I've been bitten by it before

the metadata design in clojure is quite fantastic, there's always an escape hatch. I've used this in clj-kondo lots of times

"crap, this map is symbol to symbol, but we need locations as well.."

btw please upvote CLJ-2568 :)

@raymcdermott It's great that you're teaching your son Clojure. He can take your morning shift

I haven't figured out what is and isn't a good use for metadata yet. I bet Alex would have something pithy to say

hehe have you ever met a gamer?

@dominicm please don't ask, I might get regrets

@borkdude have you looked at how zippers use metadata?

I think I remember the stock answer: use metadata only when you don't care about equality. (Or you want it to not be considered)

yes, I think that's the greatest thing about metadata: I can compare two symbols for equality even if they have different location metadata. <3

you can even serialize data structures while preserving metadata: EDN supports it. Although transit might not

https://nextjournal.com/mk/transit-metadata-keys

@raymcdermott Do you drink coffee pre or post dogwalk in your morning life?

hehe post, I like to enjoy it

We use metadata for, well metadata 🙂

btw datafy also uses metadata to keep track of the real thing that was datafied

One example is that we import data from excel. Customer data might not start at row 0, so there is a row offset going on. I attach the real row-num to the vector that contains a row using metadata.

So given something like:

[[col1 col2 col3] ;; this might be row 3 in the excel sheet it came from, so (:rownum (meta row)) will be 3
 [col1 col2 col3] 
 [col1 col2 col3]]

But I think it’s hard to define where to use meta and where to create another datastructure. We could have solved this as:

[{:rownum 3
  :columns [col1 col2 col3]}
 {:rownum 4
  :columns [col1 col2 col3]}
 {:rownum 5
  :columns [col1 col2 col3]}]

I couldn’t use meta for column-numbers as cell values could be primitives…

Using meta feels a bit like using dynamic vars.

@slipset why not use an excel lib that makes nice maps out of this?

e.g. docjure

consider it an example 🙂

aye

yeah, metadata offers flexibility and prevents pre-mature refactorings also in cases like this

We do use poi at the edges, but then make it into a matrix at some point and continue from there.

There is stuff around the excel format that doesn’t suit us, or at least didn’t suit us at the time of the implementation of this thing.

@slipset wow, this got me thinking. what if poi was properly datafied, that could maybe a a nice way of dealing with it

One thing about the excel format is that it’s sparse, so you get matrices with ragged edges.

hmm, I just needed interleave but I forgot it existed, so I wrote (mapcat list xs ys)

almost makes you wonder why core hasn't implemented it like that

probably performance: interleave is much faster