Support russian language about natural HOT 12 OPEN

E.g. the parseTableData test case currently fails because the Cucumber.xtext grammar definines that a step description must start with either (WORD | NUMBER | STRING) . The test case processes a content with Given <foo> . <foo> is recognized as a PLACEHOLDER, so extending the grammar definition by (WORD | NUMBER | STRING | PLACEHOLDER) makes the parseTableData test case pass.

from natural.

Thanks for the assist on this one! We were able to incorporate the idea in the upcoming release. Cheers!

from natural.

from natural.

See: #43 (comment)

from natural.

See also a similar discussion in the Xtext forum.

from natural.

I think the most appropriate solution would be to allow any IDs for the keywords in the grammar and implement the validation / content assist in a way that it evaluates the first line of the *.feature file to determine the used language and validate / offer proposals accordingly.

From the Spoken Languages section of the Cucumber Gherkin documention:

A # language: header on the first line of a feature file specifies what spoken language to use.
For example # language: fr for French. If you omit this header, Cucumber will default to English (en).

When implementing like that, a change in the language specification would result in a lot of validation errors (since the used keywords are not valid any more). The editor should offer multi quickfixes available from Xtext 2.13 to fix all these problems in a single action.

from natural.

This issue has not been updated for a while: marking it as stale.

from natural.

@miklossy I just wanted to let you know I've been slowly plugging away at this in the background. My current challenge has been getting this to work with existing SL_COMMENT terminal rule which makes it hidden. I suppose I can work around this by grabbing the text from the parser rather than the AST, but I was hoping for a less finicky solution. Also step keywords need to be terminals or they clash with TEXT_LITERAL. I believe this last one may just be a mistake by mypart. I will keep you posted with any major developments on this front.

Edit: I just realized I can get rid of STEP_KEYWORD terminal entirely and simply match the first word, but how then to distinguish the end of the Narrative and beginning of the steps. Maybe we should simply treat this all as a generic bag of text lines at the parser level, then add a post-processing step to distinguish which lines are actually steps by referencing a table localized keywords (cucumber provides one in json form).

Thoughts?

from natural.

@cdietrich @szarnekow @kthoms Do you have any hints on this?

from natural.

One other idea I had is to implement a step in the generator to preprocess the json into a set of optional feature bundles for each locale. The goal wouldn;t be to reduce the bundle size but to keep the terminal rules to a minimum. I am making the assumption that performance scales poorly with n, where n is the number of terminal rules defined. I could be mistaken about this.

from natural.

@cdietrich @szarnekow @kthoms @miklossy

FYI the underlying work for this can be found in drkstr101/next

The current mission is to update the grammar to remove all ambiguities (generator warnings) so that it can be loaded and compiled in ANTLRWorks. The grammar in next does achieve this, but there is some parser regression in CucumberExamplesTest

@Test def void pathological_01() {
    assertExampleParses('''
        # language: en
        Feature: ASCII punctuation
        ,./;'[]\-=
        <>?:"{}|_+
        !@#$%^&*()`~
        
        Scenario:
        Given a step
    ''')
}

Specifically, It doesn't seem to like any symbols in the Narrative that are also used in keywords and/or terminals (@, :, ", and |). If anyone could offer some suggestions so that we can make some progress on this front, I believe all the rest will fall into place.

from natural.

Hello Aaron !

I have just checked out the current state of your branch and can execute the pathological_01() test case without any problem.

However, there are some other test cases failing on my machine:

How is the current state on your machine?

from natural.