TableCell with unescaped backslash is escaped in the AST value about common HOT 5 CLOSED

Behat currently uses its own parser which isn't 100% compatible with the Cucumber one. We do attempt to test against the same testdata but exclude some of the example features . The data table examples don't seem to be in that exclusion list, which indicates the Behat parser produces the same result as Cucumber for those testdata.

Therefore a good step would be to add an example with this problematic case to the testdata in this repo and canonicalise the correct parsing. This would help us ensure that other implementations are parsing the source in the same way too.

The good news is that Behat should use the new parser at some point which will mean the parsing will be much more compatible, and maintained in this organisation

from common.

Unfortunately this looks like a feature to me:

Though since the escape is invalid, I think you might be okay mapping every \ to \\. I don't immediately see any clashes.

So reformatting:

| \o\no\ |

Should yield

| \\o\no\\ |

from common.

OK thank you for your fast response.

One think I find weird then (but that possibly because I do not know the implementation details here) is the fact that there is a difference between \| and the rest : \| are converted to |. But after reading my phrase several times, it seems logical as you want the cell value here.

About mapping \ to \\, that's what I did, but I think that Behat does not escape the same way as cucumber handles it, so changing from \ to \\ in gherkin files does crash on the Behat side (I might open an issue about this when I do investigate a little further on their side).

For the moment I'm still stucked on the prettier plugin, and the hack I found is not pretty 😰

https://github.com/mapado/prettier-plugin-gherkin/blob/main/src/GherkinAST/index.ts#L46-L59

https://github.com/mapado/prettier-plugin-gherkin#escapebackslashes

But it does work for the moment, so that's fine 👍

As an evolution, what do you think about adding the initial data in the AST ? or at least the a endColumn information in the Location of every nodes ? This way we might get the original data when we want (I can not actually do that because the size of the TableCell is lost as the value is transformed).

Thanks again

from common.

About mapping \ to \, that's what I did, but I think that Behat does not escape the same way as cucumber handles it, so changing from \ to \ in gherkin files does crash on the Behat side (I might open an issue about this when I do investigate a little further on their side).

I think that would be the way forward. I know that @ciaranmcnulty contributes to both Cucumber Gherkin and Behat Gherkin. Perhaps he can help out.

from common.

@ciaranmcnulty thank you for your response.
The file I did work with is escaped_pipes.feature that is in your list. I did not remember all other files.

I did add a file with lots of \ in my test stack and the actual output of the prettier plugin is visible here (but with backslashes escaped, which is nearly unreadable 😄 ).

I think the fix for the prettier plugin should work for now. I will watch carefully the fact that behat will use the cucumber parser and test it once released.

I'm closing this as the option in the prettier plugin should handle most cases.

Thanks @mpkorstanje too for your time.

from common.