Specify encoding when decoding registers to string

[Breina]

When calling convert_from_registers with data_type=DATATYPE.STRING, the encoding utf-8 is used.

Please parameterize this encoding so that it works with other text encoding formats (such as latin-1 in my case).

[janiversen]

Pull requests are welcome, we use utf-8 because it is the most generic one, and please be aware that it is actually not really used, since the byte value is used directly.

What is your problem exactly ?

[janiversen]

Please be aware you are not forced to use the convert functions, you can easily do it in the app.

Btw saw the app is actually a custom component of HA, which means you are using a pretty old version of pymodbus (due to the component modbus) where among others the convert functions work quite differently.

[Breina]

Thanks for the quick response!

I used BinaryPayloadDecoder.decode_string before and parsed that with python's bytes.decode. The warning message on BinaryPayloadDecoder advised me to use convert_from_registers, which did feel pretty clean.

My user has a starting byte of 0x80, which isn't ASCII nor UTF-8, so I need latin-1 extended ASCII specifically. I can create a PR for this.

Btw saw the app is actually a custom component of HA, which means you are using a pretty old version of pymodbus

Not exactly sure how it works, but I can specify a required version of pymodbus and it will get that for me. I'm running on the latest version currently.

[janiversen]

Interesting, if you can use multiple versions of the same lib. Did you actually check either pip what is installed ?

The python builtin decode function needs a character type like utf-8, if there is a "raw" or similar I quest that would be the best change.....we do our best to avoid extra parameters.

[janiversen]

Just checked someone updated ha pymodbus version to 3.8.3, which is why you do not have problems.

[Breina]

The python builtin decode function needs a character type like utf-8, if there is a "raw" or similar I quest that would be the best change.....we do our best to avoid extra parameters.

I basically need to have a latin-1 instead of utf-8, would an optional parameter that defaults to utf-8 be alright? That's still perfectly backwards compatible then.

[janiversen]

A better solution would be to add "errors='ignore'" to the decode call, which should solve all your problems.

[Breina]

I don't think this would solve my problem. I need the 0x80 byte to be decoded to Ç.

If you don't like the parameter, I can parse the bytes outside of your lib as you said, but I felt it would be cleaner to provide a way to configure it on your end.

[janiversen]

A parameter with a correct default, would be accepted, please remember to add test cases (e.g. what happens if parameter="raw").