check typing with mypy by terencehonles · Pull Request #458 · anymail/django-anymail

terencehonles · 2026-03-09T09:55:11Z

This picks up stalled PR #394. I'm not sure if squashed PRs are the default, but it makes sense to squash all this work to get rid of the earlier churn, but keep the co-authorship.

terencehonles

Some comments about the changes I've made

terencehonles · 2026-03-09T09:59:26Z

        super().init_payload()
        # late-bind in finalize_payload:
-        self.recipients = {"to": [], "cc": [], "bcc": []}
+        self.recipients: dict[str, list[Any]] = {"to": [], "cc": [], "bcc": []}


I'm leaving narrowing Any to the appropriate type as a follow up PR, since I wanted mypy to pass without disabling too many strictness flags, but I didn't want to figure out what was the most narrow type definition would be.

terencehonles · 2026-03-09T10:01:12Z

            esp=self.esp_name.lower(),
            version=__version__,
-            orig=session.headers.get("User-Agent", ""),
+            orig=cast(str, session.headers.get("User-Agent", "")),


I assume the UA is generally a string, the alternative would be to use a byte string and never decode the existing UA and just encode all the other parameters since bytes doesn't have a .format but it can interpolate.

terencehonles · 2026-03-09T10:02:47Z

        http_headers["Content-Type"] = "application/json"

-        super().__init__(
+        super().__init__(  # type: ignore[misc]


I just squelched this error, but technically since *args is unbounded headers can come as a positional argument or a keyword argument. Moving RequestsPayload to use keyword only arguments will fix this issue, but is a breaking change.

All of the payloads can be considered internal types.

fwiw, in other projects I'm starting to drop *args and **kwargs where they're not absolutely necessary. But maybe start with this change, then improve the API as a follow-on commit.

(btw, this is likely to be part of a breaking change release anyway. We might just add a blanket "internals have been reworked" changelog item.)

terencehonles · 2026-03-09T10:09:38Z

        return HttpResponse()

-    def parse_json_body(self, request: HttpRequest) -> dict | list | None:
+    def parse_json_body(self, request: HttpRequest) -> dict:


This looks like it was a bad type definition because below it was only used as a dict and would break if it were a list or None.

terencehonles · 2026-03-09T10:11:04Z

        content_type = self.get_content_type()
        content = self.get_content_bytes()
-        return SimpleUploadedFile(name, content, content_type)
+        return SimpleUploadedFile(name, content, content_type)  # type: ignore[arg-type]


name may be None here, but I don't know if the Django stubs type definition is wrong by requiring non None, if get_filename() might actually not be none, or if this code needs to change.

terencehonles · 2026-03-09T10:11:52Z

            )
            # headersonly forces an empty string payload, which breaks things later:
-            msg.set_payload(None)
+            msg.set_payload(None)  # type: ignore[call-overload]


The Python typeshed seems to be incorrect if this is valid, but I didn't look into the source to see what happens here, but I'm going to assume this is valid.

There were definitely some significant errors in the Python typeshed around mail last year (when I was looking at this for Django), particularly in legacy APIs like set_payload. I'm not sure the current status.

It looks like this is threading a needle which does work according to the source, but it's unlikely a PR will be accepted to change this. Looking into this more should this be calling the clear_content method instead? That will strip some headers, but I assume any content-* headers aren't valid without content.

terencehonles · 2026-03-09T10:14:26Z

 BASIC_NUMERIC_TYPES = (int, float)


-UNSET = type("UNSET", (object,), {})  # Used as non-None default value


This is basically a funny way to write class UNSET: pass, which the latter mypy can actually use since it's a concrete type, but this is used as a singleton which mypy doesn't like for is checks so I've replaced this implementation with an implementation I use for the same functionality including the type that mypy should use.

It looks like this needs to be pickle safe, so I've used the TYPE_CHECKING definition which is more like the original definition where there is a class defined, but this still creates a singleton of that class.

I mean, the right thing to do here is a Sentinel, but that keeps getting deferred.

terencehonles · 2026-03-09T10:17:58Z

        content = self.content
        if isinstance(content, str):
-            content = content.encode(self.charset)
+            content = content.encode(self.charset or settings.DEFAULT_CHARSET)


charset seems like it's only guaranteed to be non None if it's a "text/*" mime type, I assume this is coming from the user, so if for some reason the charset isn't defined the default makes sense, right?

terencehonles · 2026-03-09T10:18:37Z



 class CaseInsensitiveCasePreservingDict(CaseInsensitiveDict):
+    _store: OrderedDict


typeshed doesn't define _store (trying to keep it private)

terencehonles · 2026-03-11T12:48:50Z

@medmunds friendly ping (just making sure you also saw this PR)

Co-authored-by: Terence D. Honles <terence@honles.com>

medmunds · 2026-03-11T21:17:15Z

Thanks for picking this up! (And of course thanks to @YPCrumble for the earlier work.)

I won't have time to get into details before mid-next-week at the earliest. In the meantime:

I've created a "next" branch that drops older Python and Django versions, from your other PR, so typing can start with Python 3.10. Suggest rebasing this onto that. (I've already updated the target branch for this PR from main to next.)
Yes, please squash the earlier commits so we're starting from a simpler diff. You can use Co-Authored-By to preserve YPCrumble's credit.
I'll glance through your comments and see if I have any quick answers.

terencehonles · 2026-03-20T02:58:40Z

I figured we might just target main with this and use GitHub's squash feature, but I can squash and retarget the PR

This change adds tox configuration to run mypy, and updates the type information so test passes. --- Co-Authored-By: Ian Campbell <ipcampbell@gmail.com>

terencehonles force-pushed the add-mypy-typing branch from 114836d to 0578047 Compare March 9, 2026 10:44

terencehonles commented Mar 9, 2026

View reviewed changes

terencehonles force-pushed the add-mypy-typing branch 4 times, most recently from 4773003 to 1127487 Compare March 9, 2026 11:28

terencehonles mentioned this pull request Mar 10, 2026

drop Django 4.x and EoL Python versions it supported & run pyugrade #459

Merged

medmunds mentioned this pull request Mar 11, 2026

Feature/implement mypy type checking #394

Closed

medmunds and others added 3 commits March 11, 2026 14:04

Enable test workflow on next branch

76ab128

Drop Django 4.0--4.2, Python 3.8--3.9 support

b93116a

Co-authored-by: Terence D. Honles <terence@honles.com>

Run pyupgrade

f460b58

Co-authored-by: Terence D. Honles <terence@honles.com>

medmunds changed the base branch from main to next March 11, 2026 21:11

check typing with mypy

ef330c6

This change adds tox configuration to run mypy, and updates the type information so test passes. --- Co-Authored-By: Ian Campbell <ipcampbell@gmail.com>

terencehonles force-pushed the add-mypy-typing branch from 1127487 to ef330c6 Compare March 20, 2026 03:15

medmunds changed the base branch from next to main April 17, 2026 22:19

		BASIC_NUMERIC_TYPES = (int, float)


		UNSET = type("UNSET", (object,), {}) # Used as non-None default value



		class CaseInsensitiveCasePreservingDict(CaseInsensitiveDict):
		_store: OrderedDict

Uh oh!

Conversation

terencehonles commented Mar 9, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

terencehonles left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

terencehonles commented Mar 11, 2026

Uh oh!

medmunds commented Mar 11, 2026

Uh oh!

terencehonles commented Mar 20, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

terencehonles commented Mar 9, 2026 •

edited

Loading