Note that pyparsing 3.0 requires Python 3.6 or later.

I also discovered that there are several non-PEP8 attributes on ParserElement, specifically resultsName - which has a couple of referenes in matplotlib. As I did not set up the PEP8 synonyms for that attribute (or several others), it will unfortunately remain. (Since it has no leading underscore, it is fair game for users to assume it is part of the public API - an important lesson learned...)

I'll implement the synonyms for resultsName and the other non-PEP8 public attributes in 3.3.0 or possibly even 3.2.2, but they will not emit the DeprecationWarnings until at least 4.0.

Note that pyparsing 3.0 requires Python 3.6 or later.

Thanks for the hint. I think it's not a problem in practice. Per our dependency policy we are requiring
3.10 in the current version of Matplotlib. The latest version not requiring Python 3.6+ is matplotlib 3.0 (released in 2018). That is the latest version one can possibly break. But even there, you'd have to explicitly pin python to < 3.6; and if you really want to run such an old version and pin python as well, I think it's bearable for these users that they have to pin pyparsing as well - while there is no upward pinning that could prevent installing the latest pyparsing, you cannot expect that a mixture of pinned very old packages necessarly works with the latest versions of other dependencies.

At the library level we should provide functional floors to our version pinning (using caps only if we know it will break us) and leave retro-actively adjusting pins to packagers / integrators.

Do you want to keep commits or should this be squashed? - Either way, you can merge yourself.

I'm about to release pyparsing 3.3.0 next month, but the latest 3.10.6 release of matplotlib still emits DeprecationWarnings. I thought this work got merged in - I guess "merged" doesn't necessarily mean "goes out in the next release." How can I coordinate my 3.3.0 release with matplotlib to head off a storm of "I'm getting DeprecationWarnings now!" reports?

Thanks for the heads-up. We are a bit later than originally planned with the 3.11 release, but should possibly have put the changes in the 3.10.x bugfix releases. I'll do that now, but I'm unsure whether we'll still do a 3.10.7. Eithe way, we need to coordinate, because we indeed don't want the deprecation warnings to show up with users.

we need to coordinate, because we indeed don't want the deprecation warnings to show up with users.

Will definitely do so. I have a set of matplotlib tests as part of my unit tests, so that I can see this before releasing.