Improve tests

[antaldaniel]

Editor review:

Should the line https://github.com/dataobservatory-eu/dataset/blob/7bf85ac7abe9477d02b429b4d335179d94993a77/tests/testthat/test-agent.R#L9 be removed as it is not used in the test?

I don't understand why the iris object needs to be duplicated in lines like https://github.com/dataobservatory-eu/dataset/blob/7bf85ac7abe9477d02b429b4d335179d94993a77/tests/testthat/test-creator.R#L23

expect_true(is.dataset_df might become a custom expectation and/or rely on expect_s3_class() instead. Same comment for expect_true(is.subject.

https://github.com/dataobservatory-eu/dataset/blob/7bf85ac7abe9477d02b429b4d335179d94993a77/tests/testthat/test-datacite.R#L18

should be

expect_type(as_datacite(iris_dataset, "list"), "list")
What is https://github.com/dataobservatory-eu/dataset/blob/master/tests/testthat/test-dataset_prov.bak?

When using expect_error() as in https://github.com/dataobservatory-eu/dataset/blob/7bf85ac7abe9477d02b429b4d335179d94993a77/tests/testthat/test-dataset_title.R#L3 maybe add a pattern for the error message, just in case another error happens?

[antaldaniel]

Most of the expect_error() and expect_message() and expect_warning() conditions are now explicitly tested for the message. When no clear error message is written, the code will be gradually updated to give better exception messages (and test against the messages, too.) Thank you for pointing this out, indeed this gives a lot of room for further improvements.

Testing for classes:

• expect_true(is.subject(default_subject))

• expect_s3_class(default_subject, "subject")

• expect_true(is.list(as_datacite(iris_dataset, "list"))) is now changed to expect_type(as_datacite(iris_dataset, "list"), "list") , though I thihnk that the latter semantically correctly conveys the meaning of the text and actually tests more lines of code.

Both added as strictly speaking they are testing something else.

The iris_dataset_3, iris_dataset_2, etc have a different name from the built in iris_dataset because the test are changig the built-in datasets, each test in a different way. Of course, they could be called iris_dataset_alterred, but this way the tests could be taken out of the test_that flow and the changes can be inspected side-by-side.