memoize-one: `5.2.0` introduces a breaking change

Okay, based on feedback I think I have acted correctly and followed SemVer. I will close this issue for now.

This whole issue was super stressful! The last thing I want to do is cause anybody friction in their consumption.

Thanks everybody for your input

Yep, my stance on this is very similar to @jesstelford and @ehmicky . Stuff like this happens from time to time - we need to be pragmatic here, the current versions are OK - and whoever has adjusted their code already with the named export in mind… can just revert their changes (there shouldn’t be that many of such people anyway). Semver is just a human-based guideline anyway - somebody’s bug fix is always somebody’s breaking chance if we consider software at scale so it’s not worth stressing things like this too much.

I say you leave it as-is right now:

• 5.1.1 no named export
• 5.2.0 named export. Deprecated
• 5.2.1 no named export
• ~6.0.0~ doesn’t exist. Won’t be released any time soon

Edit to add:

40hrs isn’t long for folks to have upgraded their imports. Those who have can easily revert their commits. Sucks for them, but that’s the nature of software development. If they can upgrade to 5.2.0 that quickly, they can also upgrade to 5.2.1 quickly.

For those with dependabot setup, they’ll get the 5.2.1 notification just like the one for 5.2.0.

I would definitely classify this as a bug because it broke behaviour in a minor release. But you fixed it in a patch release.

@alexreardon you semver’d correctly 👍

@JGAntunes I created a PR to netlify/cli to upgrade memoize-one and move back to the default import https://github.com/netlify/cli/pull/2218

I’m now pretty convinced doing a 6.0.0 is the right course of action (it wont have named imports) and deprecating 2.5.1 so that anybody on 2.5.0 doesn’t upgrade to 2.5.1

My thoughts…

Do this:

release 5.2.1 and deprecate the new (broken) 5.2.0. It felt like this option was the best of the available options

Then make the change which alters the export. Then do this:

release 6.0.0

If the current common situation is the ideal situation, but we think it is a breaking change: I am happy to deprecate 5.2.0 and release a 6.0.0