Projections as Readmodels

I’ve been watching “In the brain of Greg Young: EventStore as a read model.”
I see the benefits of the Projections API for temporal queries.

But I think it could build my “document-based” read models too.

Although I have been reading on this mailing list that it’s not the best idea. Why?

Greg already said that Document Databases are better in handling this kind of stuff.

If I have a scenario in which I just want to be able to lookup a person’s medical record, or bank account, …
I don’t need complex querying, I just want to fetch it by a person’s (~stream) identifier.

Why should I write any client code that starts reading a stream and keeps updating my read model, if I can do it in EventStore?

It’s just an extra layer of infrastructure that I will need to write…

I’ll be probably ending up with json as well and I’m comfortable with writing the code in javascript as well (it also TDD’able nowadays…)

Any thoughts?


You can do if key only the problem is no indexing, we have been discussing adding lucene

What do you mean by 20000 projections? You mean 20000 states with say a foreach query? That’s no problem at all