fix: pending tests, single/double quotes, logger #38
Conversation
|
|
||
| if ok then | ||
| file:write(result) | ||
| file:close() | ||
| else | ||
| io_write("Failed to encode test results to json: " .. result .. "\n") | ||
| io_flush() | ||
| os.exit(1) |
There was a problem hiding this comment.
Why is this necessary? Is it because the process cannot continue in a sane way if json encoding failed?
There was a problem hiding this comment.
Yes, that and also busted is often run as a separate process wrapped in status, error = pcall(..) for example. Setting the exit code, allows to assert the error. Neotest also runs the tree with the spec that your adapter provides wrapped in xpcall and asserts the exit status.
I have actually updated the busted json formatter here https://github.com/lunarmodules/busted/pull/748/files
There was a problem hiding this comment.
I see but couldn't you assert on the error if you returned false as the second return value? Or called error? I don’t believe os.exit is necessary for pcall/xpcall to register an error.
Nice that you contributed to busted (and that it seems maintained). We can wait and see what the outcome of that PR is since it would probably be a good idea to align with what busted’s builtin handlers do.
| ;; pending blocks | ||
| ((function_call | ||
| name: (identifier) @func_name | ||
| arguments: (arguments (_) @test.name (function_definition)) | ||
| ) (#match? @func_name "^pending$")) @test.definition | ||
|
|
There was a problem hiding this comment.
I’ve had a chance to play around with this query and there are a few corner cases.
- In the example you gave here Help with busted, local-lua config 🙏 #17 (comment), the call to
pendingin the first test should not be matched, only theittest itself should. I think it needs match calls topendingwhose parent is adescribe. Runningbusted —list <spec>on the file shows what busted considers a test. - The query won’t match calls to
pendingwithout a second function argument.
I’ve got some working queries locally for these cases and I’ll continue to test it out a bit to make sure I haven’t missed something.
There was a problem hiding this comment.
Totally agree. That was a quick copy/paste from it block.
There was a problem hiding this comment.
Fair enough, I figured it was. Here’s what I got so far. Haven’t found a better way yet to encode the parent-child relationship.
;; pending blocks constrained to have a describe as a parent
((function_call
name: (identifier) @func_name1 (#match? @func_name1 "^describe$")
arguments: (arguments
(_) @namespace.name
(function_definition
parameters: (parameters)
body: (block
((function_call
name: (identifier) @func_name2
arguments: (arguments (_) @test.name (function_definition)?)
) (#match? @func_name2 "^pending$")
) @test.definition
)
)
)
)) @namespace.definitionThere was a problem hiding this comment.
I am afraid this does not work properly with Neotest summary panel.
it('this test does not run..) should not come up in the panel, as it is inside the pending block'
I would also argue, that top level pending should be parsed as it does come up with busted --list and it is quite convenient to mark a whole buch of tests to pending by just renaming describe to pending.
There was a problem hiding this comment.
;; pending blocks
((function_call
name: (identifier) @func_name
arguments: (arguments (string) @test.name)
) (#match? @func_name "^pending$")) @test.definition
this fixes it.
I would not bother trying to encode the nested cases:
- in the case
pending..it- all nesteditare skipped anyway - in the case of
it..pending, busted correctly marks the parentitas pending
This is just to show the changes I have made locally. Still investigating the json encoding of non-string objects.