Array of float to byte for storage in patchfile?

stoermelder · June 16, 2019, 5:50pm

One of my modules stores an array of float to the patchfile. Currently the size of the array is limited to 96000 but this is quite a random number and I want to loose it eventually.
The module saves only used array cells to JSON and I’ve implemented a super stupid compression but the patchfile can get quite big, especially when using more instances of my module.

The question: Is it a good idea to convert the array of floats to an array of byte to avoid the overhead involved with JSON? I’ve seen trivial solutions using an union:

union Float {
    float    m_float;
    uint8_t  m_bytes[sizeof(float)];
};

I don’t mind implementing it, I’m just worried about any platform specific problems. Will the patchfile still be useable on all supported platforms of Rack?
I have only little knowledge about platform-independent development but I know these things can be quite a pain sometimes.

Vortico · June 16, 2019, 6:03pm

What does your module do?

stoermelder · June 16, 2019, 6:06pm

It’s a recorder for “automation” of knobs/sliders/switches.

github.com

stoermelder/vcvrack-packone/blob/v1/docs/ReMove.md

# stoermelder ReMOVE Lite

ReMove Lite is a utility module for recording and replaying movements of any parameter on any module in the rack. It supports up to 8 recorded sequences, various  sampling rates, different recording-modes, a phase-input for directly controlling the playback and some more settings.

![ReMove Intro](./ReMove-intro.gif)

### Mapping of parameters

To map the module to a parameter activate the mapping mode by mouse press inside the display on the top. While showing "Mapping..." click on any parameter of any module in your rack to bind the module. You can unbind the parameter using the context menu of the display. Also, you can "locate" the module and mapped parameter if you got lost inside you rack.

### Precision and number of sequences

The module has storage of 96000 samples. At full sampling rate this corresponds to 2 seconds of recording at audio samplerate of 48kHz. Such high precision is not needed for parameter automation, so ReMOVE Lite allows at most an 8th of the audio sampling rate for recording. The lowest setting at 2048th sample gives you ~23 samples per second at audio samplerate of 48kHz and could still be ok for slowly changing parameters or low accuracy.
Be careful using higher rates: The recorded sequences are stored to the patch-file and these can get quite huge if several modules are used.

ReMOVE Lite can be configured to store 1 to 8 sequences. The maximum length for each sequence is evenly diveded, so you get 1/8 of the available recording time when using 8 sequences. The available time is shown in the context menu-option and in the display as soon as a recording starts. Changing the number of sequences resets all recordings.

Both precision setting and number of sequences are found on the context menu.

### Recording-Modes

This file has been truncated. show original

Vortico · June 16, 2019, 6:12pm

There is no good way to store 100k values in a patch. So you should avoid the problem and use a much smaller sample rate to store your signal. 30-60 Hz would be enough.

stoermelder · June 16, 2019, 6:19pm

Currently its default setting is to store each 128th sample, so 375Hz at 48kHz. I guess I will leave it this way and adjust the default setting as most users won’t care anyway.

Besides that, I just noticed I can’t do it this way because it’s samplerate dependent. Bugfixing-time…

Thank you anyway for thinking about it!