Optimize indexing by letting view specify a "type" value

[snej]

It's very common for a view to index only a particular type of document, as indicated by its type property. So the map block will often start with if (![doc[@"type] isEqual: @"sometype"]) return;.

If we allowed a view to state that it only applies to docs with a specific type value, then the indexer could use that information to avoid a lot of wasted time fetching and parsing irrelevant docs:

• In SQLite we can add a type column to rev and index it, then restrict the indexer query to only the matching type(s).
• In ForestDB we can put the type into the document metadata and check it before reading the document body.

[snej]

Note 1: CBLQueryBuilder can make use of this by checking whether the map predicate tests doc.type.

Note 2: If we want developers not using CBLQueryBuilder to use this, we'll need an API addition — something like a CBLView.documentType property.

[snej]

There's a tech note in the wiki now, describing what this is and how to use it.

[jeanregisser]

Great addition!

Would it be possible to make the "type" key configurable?

When my team migrated to couchbase we had a set of existing documents to migrate and couldn't use the "type" key because it was already used in them for a different purpose.
We ended up using a key name "doctype" instead.

[snej]

I knew someone was going to ask that… 😬 I'll have to think about it. It's probably not too hard, but I'm not sure where to store the magic property name and how to set it.

[skela]

I have a somewhat related question in that we don't have "type", but use a document id prefix as a type descriptor. Would it be possible to make use of this ?

[snej]

Same answer, I'm afraid! If we make it configurable we can do something like letting you specify a pattern in the docID.

There wasn't any issue covering making this flexible, so I've just filed one.