32-bit float data from hero6.raw

[bcardarella]

Towards the end of the hero6.raw sample file there are a few chunks of data that identify of type 'f' for a 32-bit float but the size is 12 with a repeat of 30. For a 32-bit float the size should be 4. Or is there something I'm missing here?

[bcardarella]

I'm actually, seeing this for other types as well. The s type which should be a 16-bit signed integer, so 2 bytes has a size of 6 instead with a repeat of 204.

[dnewman-gpsw]

This is correct, as it describes the array size. So float [][3] would be type 'f' with a size 12. short [][3] is type 's' with size 6.

[bcardarella]

@dnewman-gpsw ah, so an array of arrays?

[dnewman-gpsw]

Yes. The repeat is the number on time entries. So GPMF is always in form type[repeat][typesize/sizeof(type)] e.g. float[repeat][typesize/sizeof(float)]

[bcardarella]

What about char values? Do those remain as two-dimensional array?

[dnewman-gpsw]

No 1D, generally you will typesize = strlen("the string"), and a repeat of 1. On the earliest cameras this was reverse, but still 1D. The only common acception, but still following or the rules. SUIN/UNIT fields can have a 2D array of chars, which is 1D arrary of fixed length strings.

[bcardarella]

@dnewman-gpsw I'm seeing in the hero6.raw file that I get ['c', 1, 0, 50] for the length segment. In this case I was reading this as a <type><size><repeat> but your comment is saying the order should be reversed? Or is the hero6 of that older format?

[ghost]

@bcardarella See this thread for ideas:
#143

I "convert" all 'c' streams as if it had repeats 1 before parsing as a string (switching place of size and repeat if necessary).