YoursFunny 0a9ff58a69 refactor: cover the edit/answer surface in MediaSender, test the button flows ...

docs/architecture-refactor.md §3 sketched the trait with "按需扩展：
edit_message_caption / delete_message / answer_callback_query …", but only the
five send methods landed, so `callback.rs` and the edit-before-forward caption
swap were stuck on the concrete `Bot` and remained untested (AGENTS.md still
lists callback.rs as untestable).

- `MediaSender` gains `answer_callback_query`, `edit_message_caption` (HTML
  parse mode baked in, every caller uses it) and `delete_message`; the mock
  records call order plus the texts, captions and answer toasts, so tests can
  assert what the user saw.
- `send_message` now returns the sent message id instead of the whole
  `Message`: the only consumer of the value is the edit-before-forward prompt
  (which keys its record by it), and returning a `Message` forced every mock
  to build a teloxide type. `reply`/`reply_html` follow.
- `callback.rs`: the dptree entry only unpacks the update; `handle_callback`
  takes plain values + `&AppContext`. `handlers/mod.rs::edit_message_handler`
  likewise takes the values the reply carries. Admin/setup APIs
  (`get_chat`, `get_chat_administrators`, `get_me`, `set_my_commands`) stay on
  the concrete `Bot`: they are not user flows worth a trait.
- The scripted mock moves to `parking_lot::Mutex` (no poisoning unwraps).

Tests: +11 (template button, forward ok/no-channel/retryable, expired+unknown
prompt, caption swap via template, escaping of user text into the caption,
failed swap still consuming the reply, prompt record written by post_send).
fmt/clippy clean, 70 + 69 tests pass.

2026-09-17 01:31:30 +08:00

..

src

refactor: cover the edit/answer surface in MediaSender, test the button flows

2026-09-17 01:31:30 +08:00

Cargo.toml

chore: bump version to 1.5.0

2026-09-07 21:26:24 +08:00