The busy pin from the e-paper display is only used to read, not to
write, and thus there's no need for the PinIO interface. PinIn is
enough.
In addition the WaitForEdge function can be used to shorten the delays
when waiting for the display to become idle. Spurious or missed edges
are acceptable: the delay isn't very long.
This is an API-breaking change, albeit a simple one: Previously the
Dev.Clear function received a byte named "color". It was sent directly
to the display, thus acting more like a pattern to fill the display
with. In practice only 0 and 255 are likely to be used.
With this change the function is changed to receive a value of type
color.Color which is then converted to image1bit.Bit before filling the
display. The latter is also given a unittest.
Signed-off-by: Michael Hanselmann <public@hansmi.ch>
When the destination rectangle didn't start at (0,0) or cover the whole
display the Dev.Draw function would continue to update the display in
full. Obviously that didn't result in a good output.
This change updates the drawing logic to only update the affected area
(aligned to whole bytes on the horizontal axis) and limits the amount of
transferred data to the minimum needed to cover the destination
rectangle.
The calculation of the various offsets as well as the function sending
the image data are written to support unittesting.
Signed-off-by: Michael Hanselmann <public@hansmi.ch>
The structure definition and the functions were already split up. Moving
to a separate file simplifies the code structure.
Signed-off-by: Michael Hanselmann <public@hansmi.ch>
Set the start/end positions as well as the number of bytes to write by
computing them from an `image.Rectangle`. This is a preparatory step for
implementing partial image updates.
Also change the data entry mode setting to 0x03, the value applied at
power-on reset. It's not actually used given the way memory updates are
written (one large block instead of chunks).
Signed-off-by: Michael Hanselmann <public@hansmi.ch>
Introduce named constants where possible (based on the data sheet) and
compute values instead of using hardcoded ones.
Signed-off-by: Michael Hanselmann <public@hansmi.ch>
Dev.Clear: Avoid writing one beyond the end of the line. Also build the
row data only once before sending it for each line.
Dev.Draw, Dev.DrawPartial: Deduplicate logic for iterating over pixels
and simplify it such that a whole row of bits is built before sending
them in one go.
Tested using a Waveshare e-Paper 2.13in V2 display.
Signed-off-by: Michael Hanselmann <public@hansmi.ch>
Expose the quality settings for the JPEG encoder and the compression
level of the PNG encoder via the Options structure.
Signed-off-by: Michael Hanselmann <public@hansmi.ch>
The videosink package provides a display driver implementing an HTTP
request handler. Client requests get an initial snapshot of the graphics
buffer and are updated further on every change.
Signed-off-by: Michael Hanselmann <public@hansmi.ch>
Go 1.13 was released in September 2019, more than two years ago.
Updating to one later release should be fine.
The primary reason for doing this is the introduction of
"testing".T.Cleanup in Go 1.14 (https://pkg.go.dev/testing#T.Cleanup).
Signed-off-by: Michael Hanselmann <public@hansmi.ch>
Avoid using strconv.Itoa
fmt.Sprintf() can format numbers directly using the "%d" format
specifier.
Call Dev.Init in example code
Without initializing the hardware nothing is shown on the display.
Add constants for commands
Translate the command names from the datasheet into constant names,
similar to `epd/epd.go`. These names make it easier to understand what
the driver does.
One command, 0x37, is not documented in any of the datasheets available
(Waveshare 2.13inch e-Paper, Good Display GDEH0213B72).
Signed-off-by: Michael Hanselmann <public@hansmi.ch>
- Use go1.17 for github actions.
- Use go mod edit instead of go get for testing other packages. There's two
benefits:
- This removes the round trip to pkg.go.dev and use the already present local
checkout.
- The go get periph.io/x/conn/v3@${GITHUB_SHA} command worked if a maintainer
push a branch, but fails for a PR. This is because the go get command does a
clone with all the branches, so it finds the commit even if it's not in main.
For a PR, one has to pull a specific refs/pull/ID/head ref that is not a
branch, thus go get cannot find it because git clone doesn't know about it.
The only way to work around is to git clone manually.
- Remove 'go version' and 'go env' since they are already run as part of
setup-go@v2.
- Fix a new check in go vet in example code.
This does result in a few awkward cases but in general it's worth
enforcing and most importantly will make my life as a code reviewer
easier.
It's mostly deletion of dead code.
* Added support for Inky pHAT v2
* Added support for detecting Inky board. Detect Inky board opts from eeprom, and detect new inky dev with sane default from opts
Calling LowLevel.Init() in selectCard() clears the device interrupt
as a side effect, but another interrupt will occur due to the
subsquent operations. This additional host interrupt must be cleared
before the next call to selectCard.
Fixes https://github.com/google/periph/issues/425
Michael Hanselmann