Hi, I'm making a sequencer patch that uses encoder and decoder that enables me to store numbers as many as 24bits to one [Float] object.
The reason I use them is because one step in my sequencer should hold several information and I thought this way would be more CPU efficient.

But, I am now having a serious problem when saving and loading data with message box(I use Maelstorm's state-saving abstractions)

because when I save huge numbers(larger than 19bits), which look like this (8.25344e+06) into the message box, and then try to load them, the resulting decoded numbers are quite inaccurate.

Funny thing is when I store data into the float not using message box, the resulting decoded numbers were 100% accurate.

Is there any difference between printed number and an Actual number that is stored inside [Float] object?

I guess it only happens when stored number exceeds 19bits and when it becomes to look like 8.25344e+06.

I thought Pd was accurate storing within 24bits numbers but am I wrong about this?
I spent a whole month developing this patch and I feel miserable now.
Please help.

This is one of the most annoying "features" in Pd. When loading a patch, the parser formats the numbers to be more readable, which causes very large, very small, or just numbers with lots digits to the right of the decimal to be changed. I believe it's more about the number of digits than number of bits used.

The only solution, unfortunately, is to use smaller numbers. Maybe split your encoded data into two floats instead of just one.

I've rebuilt the patch using smaller numbers and it works.
Thank you Maelstorm!