Support for "special words" or dictionary lookups

[aaronsteers]

Have you given any thought to special words lookups?

For instance, "AWSInstance" and "EC2Box" (to my knowledge) require a dictionary lookup on special words "AWS" and "EC2" in order to know that that aws_instance and ec2_box are both correct and all of the below are incorrect:

• a_w_s_box
• a_ws_box
• ec_2_box
• e_c_2_box

In a past project, I was able to resolve these via a hardcoded "special words" lookup, which would prioritize translations that favored keeping special words together.

Is this something you've considered or perhaps are there other solutions I'm not aware of?

Background: I'm interested in potentially using this library for data pipeline name conversations and wanted to inquire how to solve this problem which I've had some difficulty with before.

Thanks!

[blakeembrey]

You could use the "split regexp" option to add these words manually as split options. It's a bit clunky though. That said, both these examples work as-is today since that's within the definition of how we accept "words": https://runkit.com/blakeembrey/6176190cc0655900096c5104. Open to API improvements via a PR that doesn't add too much extra bytes to the module if you'd want to officially support something like this.

[aaronsteers]

I'll look at this. Thanks, @blakeembrey!