All notable changes to this project will be documented in this file.
Release 0.0.5 - unreleased
Add page to document inverter bitfields and enumerations.
Add more fault codes and a warning that the added codes overflow the bitfield.
Update description for
battery.bat_statusin the App mapping after RCT added a proper name to it, also add a description to the field in the Registry.
Release 0.0.4 - 2023-10-05
Add a page showing the overview screens of the RCT Power app and which OIDs are used to display the values.
Mention that some behaviours and the entire content of the registry is an implementation detail of the vendor and are not mandated by the protocol itself and as such may not apply to other implementations.
Faultsfor interpreting the
fault[*].fltresponses in RCT devices.
Split documentation by moving the vendor-specific (aka RCT Inverter) parts to their own section (Registry, Faults and App mapping).
Add section about lack of protocol-level error handling.
Add note about
EXTENSIONcommands, their structure is unknown.
Document what is known about
READ_PERIODICALLYand how it is supposed to work with vendor devices.
Mention checksum algorithm.
List all known commands as well as reserved ones.
Describe plant communication, which has not been tested yet.
CLI invocation: Bash-completion activation changed with newer versions of
CLI: Document why there is now write support in the CLI.
Commandhas new functions
is_responseto help working with received frames.
READ_PERIODICALLY, but it has not been tested yet.
Commandlearned about the
PLANT_equivalents of the other commands.
CLI: Implement support for
Click 8.1caused by API changes related to custom completions (Issue #17, PR #18).
Make setuptools happy again by adhering to
7.0is the new minimum, it supported custom completion functions for the first time
8.1is the new maximum, to guard against API changes during unconditionally updating the dependencies.
Release 0.0.3 - 2021-05-22
Receiving of frames has been completely reworked
It now uses a streaming approach where parts are decoded (almost) as soon as they are received instead of waiting for the entire frame to be received. This was done in order to allow for more flexible handling. The correctness of the data still cannot be determined before the entire frame has been received and the CRC16 checksum indicating correct reception.
Except for invalid or unsupported (EXTENSION) commands, which will raise an exception and abort consumption, the
ReceiveFrame are now populated as soon as possible and no longer raise an exception when accessed
before the frame is
Rationale: The main use case is the detection and handling of frames with invalid length field. As the correctness
of the frame could previously only be determined after it the advertised amount of data was received, frames that
advertise an abnormal amount of data consumed tens or hundreds of valid frames with no way for the application to
determine what was wrong. With the change, the application can now check for command and length, and abort the frame if
it seems reasonable. For example when it detects that a frame that carries a
DataType.UINT8 field wants to consume
100 bytes, which is far larger than what is needed to transport such a small type, it can abort the frame and skip past
the beginning of the broken frame, as an alternative to keeping track of buffer contents over multiple TCP packets, in
order to loop back once it is clear that the current frame is broken.
FrameNotComplete was removed as it was not used any more, and
InvalidCommand was added in its place.
Furthermore, if the parser detects that it overshot (which hints at a programming error), it raises
FrameLengthExceeded, enabling calling code to abort the frame.
Time stamps in the output of tool
timeseries2csv.pyare off by one or two (during DST) hours.
Registry: Update with new OIDs from OpenWB.
read_pcap.pynow makes an attempt to decode frames that are complete but have an incorrect checksum to try to give a better insight into what's going on.
read_pcap.pyprints the time stamp encoded in the dump with each packet.
ReceiveFrame: Add a flag to allow decoding the content of a frame even if the CRC checksum does not match. This is intended as a debug measure and not to be used in normal operation.
Added type hints for
typing_extensionsfor Python version 3.7 and below.
Mention that tool
csv2influx.pyis written with InfluxDB version 1.x in mind (Issue #10).
ReceiveFramenow happens using the Python logging framework, using the
debug()method has been removed.
New CLI flag
--frame-debug, which enables debug output for frame parsing if
--debugis set as well.
timeseries2csv.pycan now output different header formats (none, the original header, and InfluxDB 2.x compatible headers). The command line switch
--no-headerswas replaced by
Disable Smartquotes (https://docutils.sourceforge.io/docs/user/smartquotes.html) which renders double-dash strings as a single hyphen character, and the CLI documentation can't be copy-pasted to a terminal any more without manually editing it before submitting. (Issue #5).
CLI: Fix incomplete example in
read-valuehelp output (Issue #5).
CLI: Change output for OIDs of type
UNKNOWNto a hexdump. This works around the problem of some of them being marked as being strings when instead they carry complex data that can't easily be represented as textual data.
Registry: Mark some OIDs that are known to contain complex data that hasn't been decoded yet as being of type
STRING. Most of them cannot be decoded to a valid string most of the time, and even then the content would not make sense. This change allows users to filter these out, e.g. when printing their content.
Simulator: If multiple requests were sent in the same TCP packet, the simulator returned the answer for the first frame that it got for all of the requests in the buffer.
read_pcap.pynow drops a frame if it ran over the segment boundary (next TCP packet) if the new segment looks like it starts with a new frame (
0x002b). This way invalid frames with very high length fields are caught earlier, only losing the rest of the segment instead of consuming potentially hundreds of frames only to error out on CRC-check.
csv2influx.pyhad a wrong
--resolutionparameter set. It has been adapted to the one used in
timeseries2csv.py. Note that the table name is made up from the parameters value and changes with it (Issue #8).
ReceiveFrameused to extract the address in plant frames at the wrong point in the buffer, effectively swapping address and OID (PR #11).
Release 0.0.2 - 2021-02-17
timeseries2csv.py: Reads time series data from the device and outputs CSV data for other tools to consume.
csv2influx.py: Takes a CSV generated from
timeseries2csv.pyand writes it to an InfluxDB database.
Refactored frame generation: The raw byte-stream generation for sending a frame was factored out of class
SendFrameand put into its own function
CLI: Implement simple handling of time series data. The data is returned as a CSV table and the start timestamp is always the current time.
CLI: Implement simple handling of the event table. The data is returned as a CSV table and the start timestamp is always the current time. The data is printed as hexadecimal strings, as the meaning of most of the types is now known yet.
Registry: Add handling of enum value mappings.
read_pcap.py: learned to output enum values as text (from mapping in
rctclientCLI is only installed if the
clidependencies are installed.
Tests: Some unit-tests were added for the encoding and decoding of frames.
Tests: Travis was set up to run the unit-tests.
Enum-mappings were added to the Registry documentation.
Event table: document recent findings.
Protocol: documentation of the basic protocol has been enhanced.
Added this changelog file and wired it into the documentation generation.
ENUMdata types are now correctly encoded/decoded the same as
Simulator: Fix mocking of
Release 0.0.1 - 2020-10-07