Filter test inputs before feeding them to functions under test #3
Conversation
|
Can you also commit the generated I tried to run |
0.6 and 0.8 is not supported anymore.
|
I could commit data.json but it is not a source file, therefore it shouldn't be committed in the source control system. I will make travis CI to put it in a build artifact but that needs some work. Until then data.json is available here. |
sorry, I don't get that. It is a very static file. If we don't include it here we have to have nodejs with all the dependencies present in all test environments. |
|
It is a very big, nearly static file. I see it is easier to include it in source control, but I think we should consider that topic. I open a new issue but commit data.json for now as you requested. |
Please note that this is generated using a modified clear method which deletes the nullified properties. This ensures that noise is filtered also from the output besides the input is clean. There is an open PR for that on the secret-handshake project. This is not a source file as it is generated by other source files. Inclusion of data.json in source control should be reconsidered. An issue will be opened for this.
|
The problem of including |
| args[0] = saved_filtered_input_state // revert the changes fn made on its input | ||
| output(clone({name: name, args: args, result: result})) | ||
| args[0] = saved_input_state | ||
| return fn.apply(null, args) |
There was a problem hiding this comment.
Calling fn twice causes duplicate call of verifyChallenge() on client side because clientVerifyChallenge() calls verifyChallenge()
|
It's way simpler to just check it in. What we should do, though, is make the tests fully deterministic, (i.e. pass in random seeds for every test), then it will always be the same data.json. |
|
I agree, tests should be deterministic. Every time I run generate.js it generates a different data.json. This behavior causes a lot of noise in the diff. |
This PR filters out noise from test input mentioned in #2