Document escaping rules for \ in `{string}`

[mpkorstanje]

👓 What did you see?

When using a Cucumber expression the implementation details of that expression leak into the message representation. Each capturing group in the tree-regexp a group is created as seen in cucumber/compatibility-kit#175.

With the exception of {string} the built in parameter only capture a single string and transform that. So the additional groups do not add any useful detail.

✅ What did you expect to see?

Removing the unused capturing groups would make the output more concise

Note: For {string} the transformation looks like:

                String arg = args[0] != null ? args[0] : args[1];
                return (String) internalParameterTransformer.transform(arg
                                .replaceAll("\\\\\"", "\"")
                                .replaceAll("\\\\'", "'"),
                        String.class);

But that should probably be

                String arg = args[0] != null ? args[0].replaceAll("\\\\\"", "\"")
 : args[1].replaceAll("\\\\'", "'");
                return (String) internalParameterTransformer.transform(arg,
                        String.class);

📦 Which tool/library version are you using?

No response

🔬 How could we reproduce it?

No response

📚 Any additional context?

No response

[mpkorstanje]

One reason for the current regular expressions appears to be that the RegularExpression determines its parameter types by using the ParameterTypeRegistry.lookupByRegexp. So regular expressions and parameter types have a relationship this way. This means that changing the regular expression may have unintended consequences.

But it also look like Java and Javascript have a different regex for String. Javascript defines a single regex for both " and ' delimited string while Java has two separate ones.

[mpkorstanje]

But it also look like Java and Javascript have a different regex for String. Javascript defines a single regex for both " and ' delimited string while Java has two separate ones.

In the end the difference is pretty immaterial. The default for a regular expression is string.

[mpkorstanje]

But that should probably be

               String arg = args[0] != null ? args[0].replaceAll("\\\\\"", "\"")
: args[1].replaceAll("\\\\'", "'");
               return (String) internalParameterTransformer.transform(arg,
                       String.class);

When using " I would not expect that \' to be unescaped. So this test should pass:

---
expression: three {string} mice
text: three "bl\'nd" mice
expected_args:
- bl\'nd

But this test will pass instead.

---
expression: three {string} mice
text: three "bl\\'nd" mice
expected_args:
- bl\'nd

So it looks like the \\ must be escaped consistently. Worth documenting.