Missing compatible strategy.save in docs

[wgordon17]

• I have searched the issue tracker and believe that this is not a duplicate.

This is more of a docs bug than anything at the moment. When trying to understand the "save" strategy, I found on the configuration page

---

strategy.save	Specify how to save versions when a package is added	minimum(can be: exact, wildcard, minimum)	Yes

---

2 issues with this

1. I don't see compatible listed, but it looks like it was added over 2 years ago Change the compatible strategy to use ~= operator #230
2. I see wildcard listed (also on the CLI page), but I don't see wildcard as any option https://github.com/pdm-project/pdm/blob/main/src/pdm/cli/utils.py#L468, so I'm confused as the behavior of wildcard.

Finally, this one might be a bit controversial, but I was wondering if it's worth opening an Enhancement issue to suggest moving from minimum to compatible as the default save strategy. This could be considered a breaking change, but IMO, it's the correct behavior, to default to compatible package versions instead of just specifying a minimum and letting packages upgrade to breaking semvar versions. I feel like that behavior is surprising, to say the least (I was surprised by it). I don't want to hash out this discussion here, since I want to focus on the docs bug here. Mostly just wondering if it's worth opening the Enhancement or let me idea just die here if there's no interest

[frostming]

but IMO, it's the correct behavior, to default to compatible package versions instead of just specifying a minimum and letting packages upgrade to breaking semvar versions.

Not really, you may be from the application perspective, but for a library, version specifiers with upper bound may be harmful, and this effect is harder to fix than a minimum strategy used on an application, so we decide to make it the default.

2. but I don't see wildcard as any option https://github.com/pdm-project/pdm/blob/main/src/pdm/cli/utils.py#L468,

wildcard mean to not change the specifiers, a => a, a>=1 -> a>=1

[wgordon17]

Ahh, ok...that definitely makes sense then. Yeah, coming at this from an application perspective...at least it's configurable 🚀