Enhance helper functions for register value conversions

[maslyankov]

• Added make_32bit to convert two 16-bit values into a 32-bit integer.
• Introduced split_to_16bit to split a 32-bit integer into two 16-bit values.
• Updated signed function to handle struct packing for signed integers.
• Modified NumberRWSensor to utilize split_to_16bit for handling two 16-bit registers.
• Refactored reg_to_value in Sensor class to use make_32bit for combining register values.

Related PR & issues that arised from it:

• Upstream/general improvements #391 (comment)
• Upstream/general improvements #391 (comment)
• Power values jump to 65535 kW and -65509 kW #398

Why?

Using struct for handling signed integers (especially when merging registers) is better for several key reasons:

1. Proper Two's Complement Handling:

   • Struct automatically handles two's complement representation for signed numbers
   • Direct bit manipulation might miss the sign bit handling, leading to incorrect negative values

2. Platform Independence:

   # Using struct (safe)
   import struct
   value = struct.unpack('>i', bytes([0xFF, 0xFF, 0x80, 0x00]))[0]  # Correctly handles -32768
   
   # Direct bit manipulation (risky)
   value = (register1 << 16) | register2  # Might not handle negative numbers correctly

3. Endianness Handling:

   • Struct allows explicit specification of byte order (big-endian '>' or little-endian '<')
   • Direct bit manipulation requires manual handling of endianness

4. Type Safety:

   • Struct enforces proper data type conversion
   • Helps prevent overflow issues and undefined behavior

5. Maintainability:

   • Code intent is clearer with struct
   • Less prone to bugs when dealing with different register sizes or signed/unsigned conversions

6. Standard Library Solution:

   • Well-tested and documented
   • Consistent behavior across Python versions and platforms

The small performance overhead of using struct is usually worth the reliability and clarity it provides when handling signed integers.

[codecov]

Codecov Report

Attention: Patch coverage is 92.30769% with 3 lines in your changes missing coverage. Please review.

Project coverage is 71.00%. Comparing base (f23619a) to head (cc92fef).
Report is 12 commits behind head on main.

Files with missing lines	Patch %	Lines
src/sunsynk/rwsensors.py	80.00%	2 Missing ⚠️
src/sunsynk/helpers.py	95.83%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #397      +/-   ##
==========================================
+ Coverage   70.97%   71.00%   +0.02%     
==========================================
  Files          26       26              
  Lines        1902     1914      +12     
==========================================
+ Hits         1350     1359       +9     
- Misses        552      555       +3     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[maslyankov]

I will work on tests and linting, but first @kellerza I'd like to hear your thoughts on this.

[JanKraslice]

How do I get back to the 2024 version? I'm keeping it permanently. We can't play here.
Accuracy has to be 99.9%. Anything less is wrong and we can't deploy it.
Thank you kellerza I trust you

[maslyankov]

How do I get back to the 2024 version? I'm keeping it permanently. We can't play here.

Accuracy has to be 99.9%. Anything less is wrong and we can't deploy it.

Thank you kellerza I trust you

Bro stop spamming around like that. This is an "edge" version - still in testing. Use the release (stable) version if it's that serious of an issue for you. Everyone here is working free of charge to make the project better.

P.S.: if you use the addon, simply do not use the "(dev)" one.

[kellerza]

No objections to using std lib struct

It does feel like it can be a bit more unified. I made a proposal.
As long as we have tests then it should be ok.

[kellerza]

How do I get back to the 2024 version? I'm keeping it permanently. We can't play here. Accuracy has to be 99.9%. Anything less is wrong and we can't deploy it. Thank you kellerza I trust you

Like @maslyankov - either use the multi version, or downgrade using the HA supervisor / point to another docker image

[maslyankov]

I have streamlined the implementation. @kellerza please take a look if I've missed anything. I will then work on passing linting and passing/updating tests.

[kellerza]

Thanks @maslyankov!