Read hypertable max time value with an ordered scan

[svenklemm]

The max value of a hypertable's open dimension was read with a max()
aggregate. For UUID time columns the aggregate argument is an expression
that extracts the timestamp from the UUID. Because the argument is an
expression and not a plain column, the planner cannot rewrite the
aggregate into a min/max aggregate, so it reads every row instead.

Also lock down search_path for the SPI execution and fix a sql
injection vulnerability in the UUID path.

[github-actions]

@akuzm, @antekresic: please review this pull request.

Powered by pull-review

[codecov]

Codecov Report

❌ Patch coverage is 83.33333% with 3 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
src/hypertable.c	83.33%	0 Missing and 3 partials ⚠️

📢 Thoughts on this report? Let us know!

[akuzm]

I think this change is good by itself, because the plan for min/max is going to be like this in any case, only here we skip an extra rewrite that would be done in build_minmax_path. But does this mean that sort pushdown doesn't work for this extract function? We're going to run into other performance problems then, and it makes sense to double-check this.

[svenklemm]

I think this change is good by itself, because the plan for min/max is going to be like this in any case, only here we skip an extra rewrite that would be done in build_minmax_path. But does this mean that sort pushdown doesn't work for this extract function? We're going to run into other performance problems then, and it makes sense to double-check this.

For ordering you dont need the extract function. Dont think we optimize extract function by itself atm. But outside scope of this PR