All packages, except for `cmd` and `internal`, have been moved into a new `v1` root package.
Old packages are kept for backwards-compatibility reasons. All contained code is replaced with simple type aliases and proxy functions to `v1` implementations.
Old packages default to the Rego v0 syntax, new `v1` packages default to the Rego v1 syntax.
Signed-off-by: Johan Fylling <johan.dev@fylling.se>
With this change, `undefined` outcomes of complete rule evaluations
are now also cached. Previously, only defined results had been cached,
and empty partial sets/objects.
In the case of partial rules with string keys, the introduction of ref heads
changed how they had been evaluated: Before, they had been evaluated
as partial sets, and thus got cached when empty. After, they had been
evaluated as complete rules (with ref heads), and if they were undefined,
they had _not_ been cached. This caused a performance regression.
Fixes#593.
Signed-off-by: Edward Paget <edward.paget@chime.com>
This is porting some of the existing topdown tests --
but run with the Wasm SDK. I've factored out the data
generation bits into the util package; using it in both places.
I've come to believe that whatever memory usage data we're
collecting there is probably bogus: too many of my runs yield
1597 B/op 41 allocs/op
regardless of the input size. I'd think that the problem lies in
the boundary crossing to cgo-land, but I have yet to find a
source for that.
That aside, disabling the allocation reporting, and gathering
benchmark data for the run times is probably already useful,
so let's go with that.
Signed-off-by: Stephan Renatus <stephan.renatus@gmail.com>
Anders Eknert