Add Go package github.com/theory/jsonpath #153
Conversation
theory
force-pushed
the
add-theory-jsonpath
branch
from
2664d7c to
c257097
Compare
|
Looks good! Manually merged in 66589d6 It should be clear that the consensus does not respect RFC9535, it only reflects the status quo of (often antiquated) implementations. Hopefully there is a huge overlap, eventually fully converging once more and more implementations follow the still newish spec. So for me it's less a question about the lack of consensus but rather about the behaviour of queries (syntax) not defined within the RFC. Here the consensus might penalise a lenient implementation if the others are strict (e.g. via compile errors) or vice versa. |
|
Ah, if the consensus field is missing (e.g. https://github.com/cburgmer/json-path-comparison/blob/master/regression_suite/regression_suite.yaml#L50), then no consensus was reached and only a certain tendency might be derived (https://cburgmer.github.io/json-path-comparison/#array_slice_with_large_number_for_end_and_negative_step). I believe doing nothing is the only option from that perspective. As for errors, the consensus does track agreement for not supported queries, e.g. https://github.com/cburgmer/json-path-comparison/blob/master/regression_suite/regression_suite.yaml#L50 so I wouldn't skip those in the test, I would check that your implementation doesn't return surprise answers unless intended. Mostly your implementation might actually be returning errors. |
Simplify `compare/compare_test.go` by skipping tests with no consensus, which were the source of the previous unknowns in the test. Based on advice from cburgmer/json-path-comparison#153#issuecomment-3374075044. While at it, upgrade `golangci-lint` to v2.5.0 and tweak its configuration to scan `compare/compare_test.go` and fix the long lines it complains about. Also have the `test-compare` target run the tests verbosely, and run the CI tests on Go 1.25.
Simplify `compare/compare_test.go` by skipping tests with no consensus, which were the source of the previous unknowns in the test. Based on advice from cburgmer/json-path-comparison#153#issuecomment-3374075044. While at it, upgrade `golangci-lint` to v2.5.0 and tweak its configuration to scan `compare/compare_test.go` and fix the long lines it complains about. Also have the `test-compare` target run the tests verbosely.
Simplify `compare/compare_test.go` by skipping tests with no consensus, which were the source of the previous unknowns in the test. Based on advice from cburgmer/json-path-comparison#153#issuecomment-3374075044. While at it, upgrade `golangci-lint` to v2.5.0 and tweak its configuration to scan `compare/compare_test.go` and fix the long lines it complains about. Also have the `test-compare` target run the tests verbosely, and run the CI tests on Go 1.25.
Simplify `compare/compare_test.go` by skipping tests with no consensus, which were the source of the previous unknowns in the test. Based on advice from cburgmer/json-path-comparison#153#issuecomment-3374075044. While at it, upgrade `golangci-lint` to v2.5.0 and tweak its configuration to scan `compare/compare_test.go` and fix the long lines it complains about. Also have the `test-compare` target run the tests verbosely, and run the CI tests on Go 1.25.
|
Great, that's what I needed to know. I updated my tests to skip cases with no consensus. That eliminates all the unknowns in the tests themselves (I was already tracking errors for those with consensus). |
2 participants
I could not figure out how to run the full suite, so started adding tests in theory/jsonpath#22. However, it's not clear to me what the behavior should be when there is no
consensusfield in a query object. Should it expect a compile error? Selection error? Empty result?