Set publisher components at instantiation

[domdfcoding]

Docutils has deprecated the set_components method, but it looks like these could always be provided when creating the Publisher so I don't think this change breaks backwards compatibility.

[twolfson]

Thanks for the PR and inspecting backwards compatibility!

I'm currently tight on bandwidth to run some checks before landing/releasing.

Is this time sensitive? Would it be okay if I revisit this in a month? (obv preferable to do sooner, but unsure how long it'll be until I get bandwidth back)

[domdfcoding]

Yes no rush. It's only a deprecation warning at the moment (and I can hide the warning at my end). I couldn't find a planned release date for docutils 2.0, so I assume it's a fair way off to changed like this to be made in good time.

…

On Thu, 14 Aug 2025, 17:18 Todd Wolfson, ***@***.***> wrote: *twolfson* left a comment (twolfson/restructuredtext-lint#64) <#64 (comment)> Thanks for the PR and inspecting backwards compatibility! I'm currently tight on bandwidth to run some checks before landing/releasing. Is this time sensitive? Would it be okay if I revisit this in a month? (obv preferable to do sooner, but unsure how long it'll be until I get bandwidth back) — Reply to this email directly, view it on GitHub <#64 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AB5NRJPX6XEAYXDZYYGTFUT3NSZDRAVCNFSM6AAAAACD4TXWV6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTCOBZGA2DKMBVGA> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[domdfcoding]

On closer examination I realise this only works with docutils 0.22. The parameters for Publisher existed on earlier versions, but took a class not a string. The code could be gated with docutils.__version_info__, but it depends how far off docutils 2.0 is and when you plan to drop support for older docutils and Python versions (0.22 requires 3.9+). I'll mark this as draft in the meantime.

[twolfson]

👍 Goooood catch! I'll prob stare at release dates and make a call based on that (eg past 1, 2, or 3 years. Could go 5 but kind of silly) If we do drop support, then it'd be documented in the README as a breaking change and a matching semver major, no biggie =) People can always pin to the older version for their needs =D

…

On Tue, Aug 19, 2025, 12:54 PM Dominic Davis-Foster < ***@***.***> wrote: *domdfcoding* left a comment (twolfson/restructuredtext-lint#64) <#64 (comment)> On closer examination I realise this only works with docutils 0.22. The parameters for Publisher existed on earlier versions, but took a class not a string. The code could be gated with docutils.__version_info__, but it depends how far off docutils 2.0 is and when you plan to drop support for older docutils and Python versions (0.22 requires 3.9+). I'll mark this as draft in the meantime. — Reply to this email directly, view it on GitHub <#64 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAG4KWF2ZRUONSIDV5JH4733ON6HRAVCNFSM6AAAAACD4TXWV6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTEMBSGAZDSMBQHA> . You are receiving this because you commented.Message ID: ***@***.***>

[twolfson]

I'm still quite under water regarding my queue, but I just glanced at the PyPI release dates and it seems 0.22 was only Jul 29 2025? https://pypi.org/project/docutils/#history

Can you share more about when the deprecation error comes up / what it looks like? Is it at import time or when some rst-lint method is called?

[twolfson]

I took a deeper look just now since I'd realized I didn't even read your code yet.

I usually don't like modifying code behavior off versions, since it can break in interesting ways -- and instead like modifying off any errors instead (hence deprecation error ask)

Buuuut it looks like in this case, there's no way to sniff for that -- which makes backwards compatibility a form of gymnastics.

Maybe we just release this as a breaking change (i.e. major semver) + document it accordingly. What are your thoughts?