My local SOCKS connection is suddenly not connecting. None of the AWS setup changed. Any tips on how to troubleshoot? How would I confirm that the primary stack is running?

With Datomic analytics, does cardinality-many additional tables, work with column joins? In the example, in the product_x_tags table and tags was a ref, how can I get the the attributes of tags, instead of it’s eid?

@danie what is your schema and what is your metaschema? I am not sure I entirely follow, are you trying to have another column join from a card many on the tags side?

Hi, In my app i need to "watch" for changes of particular type and react to it

i found this blog post re: tx report queue https://blog.datomic.com/2013/10/the-transaction-report-queue.html

but link 404's .

Is that kind of usecase still supported?

link works for me

seems like it's getting read in the preview above too

I think he means the javadoc link

ohh

It is supported, it’s an on-prem feature only though

real javadoc link: https://docs.datomic.com/on-prem/javadoc/index.html

https://docs.datomic.com/on-prem/clojure/index.html#datomic.api/tx-report-queue

I’m curious, are you actually writing java against datomic?

no its clojure.

maybe I am looking for wrong thing. eg use case: send an email to customer for every order placed.

I was hoping I could watch for new order creation tx and send email.

you can, it’s a perfectly fine way to do it

I am using cloud offering

oh, yeah nevermind

cloud doesn’t offer this

you will need to poll the tx-range repeatedly

unless there’s some ion way to do it?

ok. sounds like I'd have to have to poll tx manually and keep track of my position somewhere https://docs.datomic.com/cloud/time/log.html

do people write applications without 'tx functions' with datomic

my current thinking is that is not possible because when you transact data you have no way to know if they data moved underneath you since you read it. There is cas but thats only for a single attribute.

so by extension one cannot really use datomic cloud without ions

am i missing something here?

it depends on what you need. datomic transactions are really set operations, so they can compose. it’s only if you really need a specific computed write from a read that you reach for CAS or a transaction function

for example, lets say you have a cardinality-many string attr of document tags

when the user adds or removes a tag, do you a) read all the tags, remove the tag, and re-assert the entire set b) only add the tag?

sometimes you want A, in which case yes you need a tx function because you want the set’s final value to be exactly what you stated. (It’s a further question whether you want to fail and recompute if someone else wrote in the meantime, or if you want the last writer to win always)

but sometimes you want B, in which case it doesn’t matter what changes happened to the set in the meantime, the end result of adding your tag will be the same

gotcha. I was thinking of super typical ecommerce 101 buy a product flow.

1 . check invenotry  count x
2. create order record
3.  sen inventory count to x-1

Often you can split things up so the final unsafe atomic operation happens by itself

e.g. create order record but mark inactive, then activate+change inventory count using two CAS

what happens if activate goes through and change inventory fails

they’re in one tx, they succeed or fail together

oh two CAS in one tx. got it

you can also use CAS to assert no-change, e.g. [:db/cas ent attr 1 1] to ensure the value is still 1

I wish you could :db/cas to nil as an atomic retraction

thank you @favila

all that said, do make and use tx functions! but maybe you only need a smaller set of primitive functions which do precondition checking or db/ensure for postcondition checking; instead of a function for every single business operation

My team doesn't seem too comfortable with adopting both datomic and ions in one go. So I was researching if its even possible to write apps without tx functions.

Anyone have a library for turning a full map into a Datomic transaction, including retractions and additions from the current entity in the DB?

@kenny I recently just rolled my own. A bit of a pain but I couldn't find a clean solution although perhaps I missed one

Can anyone point me in the right direction for how to give my Datomic Cloud s3 access to a bucket I've created? I'm getting this error when trying to fetch an object from my bucket which makes me think it has to do with permissions {:Error {:Code \"AccessDenied\", :CodeAttrs {}, :Message \"Access Denied\", :MessageAttrs {}, :RequestId \"<reqId>\", :RequestIdAttrs {}, :HostId \"<hostId>\", :HostIdAttrs {}}, :ErrorAttrs {}, :cognitect.anomalies/category :cognitect.anomalies/forbidden}

I figured I’d end up needing to write it myself. Kinda surprised a lib out there doesn’t have a fn for it already. I also searched and couldn’t find one.

@kenny @brian.rogers that use-case is inherently racey unless done within a tx function

@brian.rogers permission for Ions?

https://docs.datomic.com/cloud/operation/access-control.html#authorize-ions

https://docs.datomic.com/cloud/operation/access-control.html#add-policy-to-nodes

That’s fine. Seems like a general function for doing that is possible.

@ghadi I think this is exactly what I was looking for thank you!

np

The idea of a “full-map” transaction seems like it could be extended to support automatic ordering of cardinality many attributes.

you can do whatever your heart desires in a tx-function

but if you don't account for the basis of the tx-data you generate being old, it will be a nasty transaction

Sure. Just a lot of interesting things you can do with a full-map transaction. Curious if there’s a reason no one has written a library for this sort of thing. Seems super useful, unless I’m missing something.

I think there are better ones out there now @kenny but this was mine: https://gist.github.com/favila/8ce31de4b2cb04cf202687c6a8fa4c94

you are talking about “reset-attributes” (the second one)

there has been discussion about a “make it look like this” tx function off and on here, maybe the archives can help

Yeah, going for exactly that.

you can also use cas if your schema is designed for it

might have to retry though

@ghadi can I private message chat you about setting up these ion policies?

it's best to chat in here if possible -- there are more people

basically if you update your stack(s) with a reference to the policy you created, then the Datomic compute nodes (whether the primary or query groups) will have augmented capabilities

Kk. Well I'm not quite sure where my code is falling down permissions wise. The docs list 4 things that I need to do: 1. use default credentials provider chain. I am making a client this way https://github.com/cognitect-labs/aws-api/blob/master/examples/s3_examples.clj#L12 and my understanding is that without specifying credentials, it is using the default ones. Am I right so far on that?

that's right, it will use the EC2 Instance Role's credentials

which is what that extra policy will augment

@brian.rogers DM me your policy if you want -- it's probably missing some sort of S3 permission if it's 403'ing

OH MY GOD IT WORKED

Thank you @ghadi for pointing me in the direction go ion permissions

my pleasure

I didn't even realize that was my problem for the longest of times

Does Datomic support "union" pull queries as described here? https://edn-query-language.org/eql/1.0.0/specification.html#_unions