A block with a single or expression needs to be treated as a comprehension instead of a
set/array with 1 item. e.g.: {1 | 1 }, [2 | foo]
Allow successfully parsing object comprehensions as rule body
x if { 1:2 | 1 }
fixes#306, fixes#307
Signed-off-by: Anand Krishnamoorthi <anakrish@microsoft.com>
Handle undefined values correctly in ordered-else. Previously an undefined value
in one of the blocks could cause the entire rule to evaluate to undefined.
Handle undefined values correctly in generic rule refs to prevent them from
propagating to output.
Signed-off-by: Anand Krishnamoorthi <anakrish@microsoft.com>
In case of empty delimiter, Rust's split returns leading and trailing
empty strings whereas Golang's doesn't.
Change behavior to match Golang/OPA.
fixes#291
Signed-off-by: Anand Krishnamoorthi <anakrish@microsoft.com>
Earlier scheduler only recognized rules and would raise an
`unsafe var` error on alias.
Register alias var names to fix this.
fixes#284
Also fix clippy warning treated as error
Signed-off-by: Anand Krishnamoorthi <anakrish@microsoft.com>
If a rule is written to produce a constant value, then not all iterations of loops
within it need to be executed. Execution can stop via early return once the first iteration
that produces a value has been executed.
This brings forth the question : What if one of the subsequent iterations would have resulted
in an error?
e.g:
x {
[1, "hello"][_] + 1
}
Such errors are not raised; consistent with OPA.
Signed-off-by: Anand Krishnamoorthi <anakrish@microsoft.com>
Second lookup of an object rule without fully qualified path, resulted
in returning the object instead of the requested field.
Signed-off-by: Anand Krishnamoorthi <anakrish@microsoft.com>
Previously we used to rely on whether there was a value in the
data document for a given rule path. This approach cannot handle
the case of evaluating a.b when a.b.c has been evaluated but
a.b.d has not been evaluated. Upon evaluating a.b.c, the data document
will already have a value of a.b even though a.b.d has not yet
been evaluated.
Hence we need to keep track of evaluated rules separately.
Signed-off-by: Anand Krishnamoorthi <anakrish@microsoft.com>
Allow `import input` instead of erroring out.
This import is redundant and has no effect.
Emit `print` messages to stderr onstead of stdout.
Signed-off-by: Anand Krishnamoorthi <anakrish@microsoft.com>
Implement `import rego.v1`
https://www.openpolicyagent.org/docs/latest/policy-language/#the-regov1-import
- `if` required before rule body
- import rego.v1 automatically imports future.keywords
- handle import shadowing
- data, input cannot be shadowed
- deprecated functions as disallowed
- rules must have assignment or body
- `contains` required for parital set
Signed-off-by: Anand Krishnamoorthi <anakrish@microsoft.com>
- ignore worktrees
- feature guard time module
- Apply with modifiers before evaluating loop expressions
- Support value modifier for functions
- stubs for http.send and io.jwt.decode_verify
- Initialize with-document after initializing init data
- In case of conflict, with modifier override init-data values.
- In case of conflict, subsequent with modifier overrides earlier ones.
- Ensure that zero parameter functions are evaluated and added to document
- opa.runtime builtin
returns:
- git commit hash
- environment vars
- regorus features enabled
- builtins available
- deprecated builtins available
- If `sort_bindings` is specified, sort the bindings in OPA tests
- gather inputs, used vars and comprehensions in with modifiers
- For refs starting with `data`, ensure that modules are evaluated before looking up
value of the expression. Thie ensures that modules that have only been partly populated
(E.g via with mods) are completely evaluated before the value is looked up
- Mark rules overridden using with modifiers are evaluated.
- Exclude env vars in opa.runtime.
- Include regorus version in OPA runtime
- update to opa v0.60.0
- scheduler: Handle function refs in with modifers. Error out only if
a truly undefined ref.
- Handle undefined params, parameter expression evaluation errors before
applying with modifiers.
- When applying with modifiers, first determine whether the target is a
function. If so, handle cleanly.
- concat: raise error only in strict mode
- In strict mode, propagate errors raised by function rule execution
in case of multiple function definitions for same rule
- skip "withkeyword/builtin-builtin: arity 0" test which can never pass.
- When a mock has is being applied, clear with_function so that
other mocks won't be applied during the evaluation of the mock.
- Ability to specify strictness in tests
Signed-off-by: Anand Krishnamoorthi <anakrish@microsoft.com>
- object.union
- object.union_n
- treat negative integers as two separate tokens (Sub and Number)
when seen in arithmetic expressions
- Ensure that fully query string is parsed
- Evaluate queries in a separate module instead of the last read module.
This correctly handles queries of the form `x = data.test.y` where x is
already a ref in `data.test`
- Handle queries producing multiple outputs in test infrastructure
- Add tests for engine
- Add tests locking down valid queries
- Update opa.passing
Signed-off-by: Anand Krishnamoorthi <anakrish@microsoft.com>
- Remove unnecessary memory allocations
- Add --non-strict flag
- Ensure that only empty modules (ones without rules) are initialzed prior to evaluating rules.
- Record rule as entry for each of its prefixes.
For example, for a rule a.b.c =... in package test, record it in
rules["data.test.a"], rules["data.test.a.b"] and rules["data.test.a.b.c"]
This allows evaluating the correct list of rules based on expessions
a.b.c, a.b, a, data.test.a.b.c, data.test.a.b, data.test.a
Closes#69Closes#70
Signed-off-by: Anand Krishnamoorthi <anakrish@microsoft.com>
Number is implemented using rust_decimal::Decimal which uses a 96 bit mantissa.
TODO:
a) Support u64, i64 variants
b) Determine desired semantics for floating-point
c) Determine desired big integer length
d) Explore other big int/big float crates
Signed-off-by: Anand Krishnamoorthi <anakrish@microsoft.com>
- units.parse, units.parse_bytes
- json.is_valid, json.marshal, json.unmarshal
- yaml.is_valid, yaml.marshal, yaml.unmarshal
- object.subset
- set_diff
* Also print number of errors due to each missing function
* Also lock down fully passing OPA suites
Signed-off-by: Anand Krishnamoorthi <anakrish@microsoft.com>
1. Numeric indices will be converted to strings for refs beginning with `data`
if there is not valid numeric key
2. Error out if input document already contains value for a ref
Signed-off-by: Anand Krishnamoorthi <anakrish@microsoft.com>
When evaluating rules, upon error the last pushed scope wasn't being
popped from the stack of scopes. This causes incorrect behavior
when there are multiple definitions for the same rule name.
The fix is to make sure that the scopes are popped manually upon encountering errors.
Once the interpreter logic is locked down, then we need to clean up scope management
using Drop functions so that the cleanup happens even during short circuited return.
Signed-off-by: Anand Krishnamoorthi <anakrish@microsoft.com>
