Documentation often states "slave" parameter but "unit" has to be used

[TheChatty]

Hi... a few pages claim you have to use slave to specify the unit but instead it has to be unit to be working.

For instance this page for function write_register (tip: search for slave in entire documentation).

[janiversen]

Indeed on dev there are some confusion at the moment. Client only use "slave=" and Server continue for now to use "unit=".

Also if you use the advanced client classes (call .execute manually) the request classes still uses "unit=".

We are moving from "unit=" to "slave=" but it is a process that takes a while. All Client documentation are updated, and the server documentation are not touched so are in theory also correct.

[janiversen]

If you find a place where doc says "slave=" and "unit=" is needed in practise, please let us know, your general search for "slave=" seems correct.

[TheChatty]

I already listed an example. You probably meant further examples?

[janiversen]

but your example is not wrong, that is a client function and uses slave= as documented.

[TheChatty]

I just analyzed the bytes with Wireshark... if I use slave=247 0x00 will be used. If I use unit=247 0xf7 will be used. Thus doc and code differ.

[janiversen]

Hmmm I will control write_register, sadly enough we need to catch this function by function.

[TheChatty]

Btw, read_holding_register is the same (needs unit= but slave= is documented). Is there a reason for the general exchange of unit with slave?

[janiversen]

yes, unit caused quite a lot of confusion because on serial lines the same thing is called slave. It dates back ton an error in the modbus documentation for tcp.

[janiversen]

please see "client_sync_basic_calls.py"

rq = client.write_register(1, 10, slave=0x25)

generates a request for unit/slave 0x25.

rq = client.write_register(1, 10, unit=0x25)

fails with

WriteSingleRegisterRequest.__init__() got multiple values for argument 'unit'

So it seems you are not using the latest dev, or have another problem.

[TheChatty]

# pip list|grep pymodbus
pymodbus            3.0.0.dev4

The following code produces the below Wireshark captures:

#!/usr/bin/env python3
from pymodbus.client.sync import ModbusTcpClient as ModbusClient

client = ModbusClient("tasmota-0bbcbe-7358", port=5020, timeout=10)
client.connect()

rr = client.read_holding_registers(0x90, 1, unit=247)
# tcp dump: 00 01 00 00 00 06 f7 03  00 90 00 01

rr = client.write_register(0x90, 0x1234, unit=247)
# tcp dump: 00 02 00 00 00 06 f7 06  00 90 12 34

rr = client.read_holding_registers(0x90, 1, slave=247)
# tcp dump: 00 01 00 00 00 06 00 03  00 90 00 01

rr = client.write_register(0x90, 0x1234, slave=247)
# tcp dump: 00 01 00 00 00 06 00 06  00 90 12 34

client.close()

There were no errors.

[janiversen]

You are comparing apples and oranges !!!

You have selected "dev" in readthedocs, which is not "3.0.0.dev4" (that is quite old)

In 3.0.0.dev4 everything is still "unit="

If you want to test the newest please clone the repo and use "dev", which correspond to the "dev" documentation.

[TheChatty]

I wasn't aware of that. I just upgrade to latest git commit but the above example fails with:

Traceback (most recent call last):
  File "Trovis_WriteEnable_min.py", line 2, in <module>
    from pymodbus.client.sync import ModbusTcpClient as ModbusClient
      ModuleNotFoundError: No module named 'pymodbus.client.sync'

[janiversen]

Because there are no "Pymodbus.client.sync" it is now "Pymodbus.client" (see the documentation).

[TheChatty]

Pardon me. Now it's solved (slave= working, unit= showing error).