There are not many ShoutTests that could ensure that I'm no the right track...
And Shout code is hard to read for me (see below).
It's also strange to have to change such a central method after all these years of service.
Many reasons for a stage in inbox, if ever a few pairs of eyes are browsing there...
<rant>
I find this Shout Parser selector particularly misleading...
It's hard to guess that #isSelectorCharacter: is going to check exclusively for binary selectors before I browse the implementation details.
Should it be named #isBinarySelectorCharacter:, I could eventually omit such browsing.
In a lesser way, I also do not like self isName, self isBinary, self isKeyword,...
The intention is to check for the currentToken, not self.
They have the advantage of being short, but reading the code gives a strange taste.
If I wanted to avoid the heavy isCurrentTokenAName, isCurrenTokenABinarySelector, isCurrentTokenAKeyword, ... maybe I would just use has instead of is?
I've also ran the Shout Parser through a Debugger, and I encountered a lot of redundancy during the tokens scan (those isName isKeyword... are ran several consecutive times).
</rant>