Indented scenarios in stories was possible in JB2 but after upgrade failed in JB3

I believe that this had to do with the change to support i18n'd keywords that in some languages are smaller (y in Spanish and e in Italian) and a shift to look for those at the beginning of the line (if I remember correctly). Just want to mention that so that we don't break the work that was done before.

Brian is quite right - this is not a regression, rather a feature (JBEHAVE-325).

It is also documented in http://jbehave.org/reference/stable/migrating-from-2.x.html

The usecase of indented scenarios is quite rare and is outweighted by the advantage of allowing keyword in comments, titles, tables etc ...

If one wants to keep at all costs tabs/spaces before keywords, a possible workaround would be writing a pre-parser that strips these tabs/spaces before passing it to the RegexStoryParser.

If there' a reason for it - fine. The HTML output is indented appropriately

Could we perhaps throw an exception to make more obvious to users not to have leftmost whitespace ?

The notification to users is the fact that the step is not matched as expected and it marked as pending.

We can't do more than that, even if we wanted. That's the nature of regex parsing.

Au contraire, we could regex for whitespace at the start of a line, *before* we do any other regex for G/W/T (etc):

For (multiline):

^[\x20\x09]+[^\s]

On:

It will find the space-space-h sequence and nothing else.

Yes, it's essentially the pre-parsing concept proposed above.

But it's probably best done in a decorator class as it's a rather rare corner case to justify it being done by default (would slow down parsing that is already not lighting fast). Anybody wanting to use indentation could simply configure the decorator around the regex story parser.

we've already got the configuration around greedy vs non-greedy qualifiers - it might interesting to wrap that together with this in a nice tidy design.