evaluate test cases

[pkra]

Anything we can do to improve things?

[pkra]

Coverage is lacking for:

• book-title.js
• meta-name
• meta-value
• subj-group
• subject

[pkra]

Two conflicting thoughts:

• should we refactor our tests to be more unit-test-like? That might ease adding new elements.
• Should our test sample be generated from TeX via texml? That might more easily catch unexpected changes in texml.

[pkra]

Not using texml to generate our test files risks diverging unintentionally. However, this hasn't been a major problem in the past.

I'm leaning more towards unit-testing because the tests would be easier to understand (and thus maintain). But that would move us even further from a texml-generated approach because texml is too slow to pre-process each snippet.

I suppose we could keep the snippets for the unit tests separately but the XML would just be full of texml-generated boilerplate.

What I'm hoping will happen: if we refactor to unit tests, we'll find which tests are too interconnected. That'll give us a few more complex test cases which we can then generate with texml.

[pkra]

447 makes me want to find time for this.

[pkra]

Random note: perhaps texml's test files could serve as integration/snapshot tests.