Indexing information derived from the row can be powerful. For example, using the hypothetical extract_date function,
create table ticks (
PRIMARY KEY (symbol, ticked_at)
CREATE INDEX ticks_by_day ON ticks(extract_date(ticked_at));
SELECT * FROM ticks_by_day WHERE extract_date(ticked_at) = '2014-5-13';
Support for pure user-defined functions (UDF)
operator functionality in CQL
A couple of questions about the scope of this ticket:
1. Let's start with just indexing a single column. We can worry about the other when we start adding compound indexes in the non-functional case.
2. I'm okay with just accepting a single index per column if that simplifies implementation a great deal, but that's a limitation we'll want to lift sooner than later.
I'm okay with just accepting a single index per column if that simplifies implementation a great deal, but that's a limitation we'll want to lift sooner than later
We have CASSANDRA-7771 for that. We can and should definitively deal with that separatly.
Mikhail, is this still something you have time to work on? If not, we can move it elsewhere.
Jonathan Ellis I've almost completed the initial implementation, but had no time to finish that. I hope I should be able to complete that somewhen in January.
Bumping to fix version 4.x, as 3.11.0 is a bug-fix only release.