I have a question about aliases and referenced collections (JOINs): What are you recommendations on this topic? Is it ok to reference an aliased collection or only a "real" collection must be referenced? In an ideal world I would deal in production with only aliases, meaning I could have a

taxonomy_production

alias with different versions

taxonomy_1

,

taxonomy_2

and at a time

taxonomy_production

might be an alias of

taxonomy_2

. But in both

taxonomy_1

and

taxonomy_2

I would have a field referencing `ext_coll_production`which is an alias of `ext_coll_1`(or

ext_coll_2

).

You can reference by alias as well, it will point to the collection that the alias points to

I'm not sure to fully understand what you mean by

It will point to the first collection, not the aliased collections

Which first collection? Let's say I have

coll_1

collection referencing

taxo_alias

. and

taxo_alias

is an alias for

taxo_1

. Currently when I query on

taxo_alias

or

taxo_1

joining on `coll_1`I have the impression to get

"Failed to join on `coll_1`: No reference field found`

Sorry, by first I meant the collection the alias points to at the time of the referencing / child collection creation

so you mean the reference is static and not dynamic?

Exactly. If you have any specific use-case where it doesn't work, could you make a reproducible example through curl commands, so we can debug this further?

If it was fully dynamic I think it would be better but I can understand it may be complicated on you side to support it this way. Having this in mind I will rework my way of updating data. Currently I would say it's not a good idea to reference an alias because you don't know which collection is really referenced and having an explicit reference might be clearer. Let's say

taxo_alias

->

taxo_1

ext_coll_alias

->

ext_coll_1

references

taxo_1

taxo_alias

->

taxo_2

I need to duplicate

ext_coll_1

to

ext_coll_2

to explicitely reference

taxo_2

and then

ext_coll_alias

->

ext_coll_2

JOINs are actively being worked on, currently we're focusing on adding support to query by and facet by joined fields. Dynamic references are a woozy because what if the aliased collection does not include the reference field that the other collections point to? There's a lot of stuff that need to be cared for in situations like the one described

I'm still experimenting updates on collections with async reference and I have some weird behaviour: I deleted a collection

reaches

having an async ref on collection

taxo

and when I'm updating documents of

taxo

(with.

emplace

mode) I'm having 400s

Collection `reaches` with async_reference to the collection `taxo` not found.

Which is expected because I deleted that

reaches

collection. And I have also 400s on the new collection

reaches_2

wit async ref

taxo

,

Error while updating async reference field `segment_id` of collection `reaches_2`: Document `id: lr_1634368372_*` already has a reference to document `1783355` of `taxo` collection, having reference value `lr_1634368372

Does it mean I cannot update documents used in async ref ?

Cc: @Harpreet Sangar

I deleted a collection

reaches

having an async ref on collection

taxo

and when I'm updating documents of

taxo

(with.

emplace

mode) I'm having 400s

Copy code

Collection `reaches` with async_reference to the collection `taxo` not found.

This must be a bug. When referencing collection is deleted, we must remove all the reference info from the referenced collection. I'll try to reproduce it locally.

FYI I have really weird behavior like having 800 results in

facet_counts

but only 1 row in

hits

. Maybe it's related.

I was able to reproduce the issue https://github.com/typesense/typesense/pull/2229/files PR fixes it.