Formula dependent on specific fork of library: possible?

[NJAldwin]

I've been using the tool kafkacat recently: https://github.com/edenhill/kafkacat

One particular feature in kafkacat (Avro decoding + JSON envelope) requires using edenhill's fork of yajl (a JSON library): https://github.com/edenhill/yajl

The current formula for kafkacat depends on yajl, but the regular library, not the fork:
https://github.com/Homebrew/homebrew-core/blob/aba5c04012b253654b28fa7aa0620ad165306416/Formula/kafkacat.rb#L19

If I wanted to change the formula to depend on edenhill's fork of yajl, is that possible? What would be required -- would I be able to (/should I) create a separate formula e.g. edenhill-yajl that installs the yajl fork, then depend on that?

fwiw, I am looking to contribute changes here back to the master formula (so kafkacat works out of the box for everybody), not on a custom tap.

[Rylan12]

I don't know if we allow forks of other formulae, so another maintainer should weigh in on that. If so, then yes, adding it and changing the depends_on "yakl" line to depend on the fork would probably do the job.

Instead of adding the yajl fork as a new formula, another option might be to add the yajl fork in a resource block.

If the change isn't accepted, you are encouraged to host your own tap which can contain a formula for the yajl fork and a version of kafkacat that depends on it. More information about creating and maintaining a tap can be found here.

---

Edit: I didn't read this close enough so here's a better answer:

Yes, adding a new formula with the yajl tap is probably the best option, but I'm not positive it will be accepted because it's a fork of another formula.

If it's not accepted, another option is to add the yajl fork as a resource block to the kafkacat formula.

[SMillerDev]

To add on this, why doesn't the fork just upstream it's changes so everyone can work on the same codebase?

[NJAldwin]

It seems that the upstream yajl library has been abandoned: edenhill/kcat#269 (comment)

Perhaps that will bolster the case for making this fork available as a separate formula?

[SMillerDev]

If it's abandoned and then forked that would actually be a reason to change the original formula