Keyboard: Add Movable interface to be implemented by page objects #695
Conversation
By taking a snapshot of the DUT state in the constructor. This is analogous to calling `get_frame()`.
This can be useful for transparent backgrounds, or to implement other means of interaction. I've not made it public for now, so we have more freedom to play, but we can make it more public later if it turns out to be useful, even without a new stb-tester release. I've used a base class here rather than duck typing to make the behaviour more explicitly opt-in.
wmanley
force-pushed
the
keyboard-interaction
branch
from
6f6c7f1
to
d2d8e74
Compare
| stbt.press_and_wait("KEY_OK", mask=self.mask, stable_secs=0.5, # pylint:disable=stbt-unused-return-value | ||
| timeout_secs=1) | ||
| page = page.refresh() | ||
| page = self._move_one(page, "KEY_OK") |
There was a problem hiding this comment.
This is a change in behaviour: We are now asserting the press_and_wait result, and we increased timeout_secs to 10.
If you have specified a mask that excludes the text box where the typed-text goes —a common thing to exclude, because it usually has a blinking cursor that breaks press_and_wait when used for navigation— then this will fail.
It seems that move_one isn't the right abstraction for this operation, which isn't about moving (maybe it "moves" —i.e. changes to a different key— if you're in shift mode or if you're pressing OK on a mode-change key, but most of the time it doesn't). Because of the short timeout and lack of assert, this use of press_and_wait should be OK even for UIs with transparency or some other reason why press_and_wait won't work for navigation.
There was a problem hiding this comment.
Note: This used to be a plain stbt.press("KEY_OK") (sometimes with interpress_delay_secs=1) before I added support for shift modes (see commit 07555d0). I wonder if it would make sense to check whether the node we're on has an outgoing edge labelled with key="KEY_OK" and use that information to choose press vs. press_and_wait. It would make enter_text a tiny bit faster, by not having to call page.refresh().
So that callers can override it if they need to provide a custom implementation (for example to handle live video behind a translucent background). I won't document this for now, but it gives us some flexibility without having to make a new Stb-tester release. This is an interim measure until we iron out the "Movable" API in <#695>.
So that callers can override it if they need to provide a custom implementation (for example to handle live video behind a translucent background). I won't document this for now, but it gives us some flexibility without having to make a new Stb-tester release. This is an interim measure until we iron out the "Movable" API in <#695>.
This allows page objects to override the mechanism for doing
presses.Builds on top of #685.