Use shared test data

[aslakhellesoy]

Description

Use shared test data

• Go
• Java
• JavaScript
• Perl
• Python
• Ruby

Motivation & context

So we have a single source of truth for all implementations

Type of change

• Refactoring/debt (improvement to code design or tooling without changing behaviour)

[aslakhellesoy]

@ehuelsmann @jenisys would you be able to push some more commits to this branch with similar changes for Perl and Python? That way we can ensure that all implementations behave consistently, and we don't have to maintain the acceptance tests in more than one place.

Thanks!
Aslak

[ehuelsmann]

@aslakhellesoy no problem. I'll have to look at it over the weekend. I love the idea of a single test suite, also because that ensures consistent behaviour across implementations. The perl "stringifier" can't match the output that's in the parser.yml tests at the moment: AND expressions always have exactly 2 terms (using a nested term-of-two-terms to model a 3-term AND). But I'll start with the behaviour that I can test and work my way through this from there.

[ehuelsmann]

@aslakhellesoy I spent some time working on this branch. I'm running into the assumption in the Perl code that tags start with an '@' sign, but in these tests, they don't. Should I remove the check for the '@'-sign to distinguish junk from tags?

[aslakhellesoy]

I don’t think the parser should require operands to start with an @. The library is a general-purpose parser/evaluator for boolean infix expressions.

[ehuelsmann]

Ok. I'll change the Perl version of the library.

One follow-up question: I noticed you don't have a test for the following expressions:

a and or

is the or on that line an operator or and operand?

[aslakhellesoy]

@ehuelsmann I added a test for that in ef521d9

The or token is an operator and should cause a parse error here.

[ehuelsmann]

Why does this expression not generate a syntax error? I'm reading these tokens in this line:

- 'x'
- or
- 'y or'
- 'z'

I'm also reading an escaping character at the end of the string, meaning there's a missing escaped character. Are the semantics of an escaped "end of line" just an end of line? (so, ignore the backslash?)

[aslakhellesoy]

I think a \ character that isn't followed by either \, ' ', ( or ) should result in a syntax error. If we agree on that, we need to change the test data and update the other parsers accordingly.

WDYT @yusuke-noda - you contributed some nice escaping fixes in cucumber/common#1778 (before we moved the library to this repo) and this seems related.

[yusuke-noda]

First, the current java, javascript, go, and ruby implementation differs from the perl implementation in the following points,
so I think that the java, javascript, go, and ruby implementation should be fixed.

• Perl implementation can escape spaces, but other implementations cannot.
• Perl implementation throws error when the end of the expression is an odd number of backslashes, but other implementations don't throw error.

And,

I think a \ character that isn't followed by either \, ' ', ( or ) should result in a syntax error. If we agree on that, we need to change the test data and update the other parsers accordingly.

WDYT @yusuke-noda - you contributed some nice escaping fixes in cucumber/common#1782 (before we moved the library to this repo) and this seems related.

I think its good idea.

[aslakhellesoy]

I merged #17 into this PR, and java/ruby/javascript/go now throw errors on illegal escapes. Test data has been updated too.

[aslakhellesoy]

@jenisys any update on this? The other implementations have some recent changes in how parenthesis and whitespace is escaped, and this is reflected in the shared test cases in this PR. The Python implementation has not been updated to reflect this.

We may need to disable the the build/release of the python implementation until this is fixed.

[jenisys]

@aslakhellesoy
I will look into it in the next few days.
First I fix the CI build problem related to PR #15.

[aslakhellesoy]

See #18 and #19 for Python / Perl.