Runtime behavior tests #135
Conversation
stasm
force-pushed
the
runtime-behavior-tests
branch
from
a51498e to
2e64d46
Compare
| @@ -24,11 +24,13 @@ ifneq (,$(wildcard ./test/__setup.js)) | |||
| @mocha --recursive --ui tdd \ | |||
| --require babel-register \ | |||
| --require babel-polyfill \ | |||
| --require ./test/__setup | |||
| --require ./test/__setup \ | |||
| test/ | |||
There was a problem hiding this comment.
I was getting some strange unknown option --recursive errors from mocha without this.
| val += ch; | ||
| } | ||
|
|
||
| if (ps.currentIs('\n')) { | ||
| throw new ParseError('E0020'); |
There was a problem hiding this comment.
I submitted projectfluent/fluent#83 to reflect this change in the spec.
| while (this._index < this._length) { | ||
| this.skipInlineWS(); | ||
|
|
||
| if (this._source[this._index] === ')') { | ||
| return args; | ||
| } |
There was a problem hiding this comment.
Moving this here allows the last argument to be followed by a comma. I filed projectfluent/fluent#84 to allow this in the spec.
| const [entries] = parse(source); | ||
| const expectedEntries = JSON.parse(json); | ||
|
|
||
| for (const [id, expected] of Object.entries(expectedEntries)) { |
There was a problem hiding this comment.
Iterating over expectedEntries here means that the runtime parser could possibly parse more entries and we wouldn't learn about it from these tests. I'd like to fix this in a follow-up later on.
| test(ftlfilename, function() { | ||
| return Promise.all( | ||
| test(ftlfilename, async function() { | ||
| const [ftl, expected] = await Promise.all( |
There was a problem hiding this comment.
No functional changes here; just switching to async.
|
This is PR consists of 3 commits: the first one adds the tests and the other two fix the differences in parsing between the runtime and the tooling parser. |
Partial port of projectfluent/fluent.js#135 affecting the tooling parser.
Partial port of projectfluent/fluent.js#135 affecting the tooling parser.
Partial port of projectfluent/fluent.js#135 affecting the tooling parser.
This doesn't quite add behavior tests to the runtime parser because the runtime parser doesn't produce the same errors as the tooling one. What we can do, however, is parse all fixtures in
fluent-syntaxand check if the runtime parser parsed the same messages (check if keys are present) and if so, if their interface is the same (is value present or missing? which attributes are defined?).