better handling of "pinned" modules?

Andrew Jaffe a.h.jaffe at gmail.com
Fri Jan 8 16:55:59 EST 2021


On 08/01/2021 18:21, Chris Angelico wrote:
> On Sat, Jan 9, 2021 at 5:18 AM Andrew Jaffe <a.h.jaffe at gmail.com> wrote:
>>
>> Hi,
>>
>> I don't know if this makes more sense here or on "python-ideas" (or
>> elsewhere?) but I'll try this first:
>>
>> I am starting to encounter more and more instances of packages requiring
>> older, pinned, versions of modules, and this is occasionally actually
>> starting to cause conflicts.
>>
> 
> The first thing to do is to see if those packages ACTUALLY need older
> versions of those dependencies. There's a good chance they don't.
> 
> To avoid creating this sort of problem, don't depend on a highly
> specific version of things; just depend on a minimum version, and if
> there's a problem (ONLY if there's a problem), a maximum version.

Well, sure. But there's still the "aesthetic" problem that `pip[3] 
check` reports a problem in such a case, and the real (albeit 
correctable) problem that `pip[3] install --upgrade` will occasionally 
automatically downgrade required packages.

So perhaps my original query about whether there could be a way to 
actually solve this problem is still potentially interesting/useful.

AndrewJ

> ChrisA
> 




More information about the Python-list mailing list