Replace mustache_test with generated spec, and fix failures #43
Conversation
|
That looks really, really cool. Looking forward to seeing progress on this. Are there features this spec tests that mustache.go doesn't have yet? |
|
The failures are mostly deviations from the spec rather than missing features. After fixing the panic and the infinite loop, here are the failures, reformatted a bit for readability. Most of them are related to whitespace (which IMHO mustache makes way too complicated), but if we ignore those for now, here are 15 remaining failures. Mostly ampersand interpolation and dotted names. |
I thought it'd be clearer to embed the JSON straight from the spec, but with all the escaping, it was actually pretty opaque. This is simpler.
The TestPartialsRecursion test currently causes an infinite loop, because partials are being parsed eagerly, even when they will never be rendered. This defers parsing until the partial needs to be rendered.
This fixes a runtime panic in the TestInvertedContextMisses, which attempts to render an inverted block with an empty contextChain.
HTML entities inside {{&ersand}} mustaches should not be escaped.
The var name was previously not being stripped.
Missing partials should render an empty string, not an error message.
adammck
changed the title
|
Update: I've fixed everything except for 32 whitespace-related failures. Those will be a bit tricky, since Mustache's whitespace rules are pretty weird. I think it would be useful to skip those failing tests for now and get this merged, since the output is otherwise to spec. What do you think? |
This replaces the
mustache_testfile with a much more comprehensive test suite generated from the official Mustache spec. I wrote a little generator to do this rather than parsing the spec and generating the tests at runtime, because I think the output is surprisingly idiomatic and comprehensible.This improved test suite causes about 50 failures, a runtime panic and an infinite loop, so you might want to look into those before merging. (Thankfully most of them are related to whitespace and error messages, so should be simple to fix.) I've already started working on those in my fork, and can update this PR when they're all fixed.