Support for DBC float and double types?

[karlding]

Hi,

I was wondering if you were planning on adding support for float and double types?

dbcc currently treats float and double types as a signed value of the same size (ie. int{32,64}_t) when generating code, which results in incorrect values being encoded/decoded.

From a quick scan through the code, it seems like you already have an attribute on the struct allocated for floating-point types (is_floating), and the codegen code accounts for it, but it just isn't getting set anywhere when parsing the DBC.

This is indicated via the SIG_VALTYPE_ attribute in the DBC.

A float value is denoted via

SIG_VALTYPE_ <can-id> <signame> : 1;

And a double is denoted via

SIG_VALTYPE_ <can-id> <signame> : 2;

Thanks!

[howerj]

Hi!

I didn't have any plans to, because I didn't have an example of a DBC file that encoded floats, it's not something that I've really needed. It can be added fairly easily now that you've given an example of one.

Could you give a full example of a minimal file I can test this on? One that encodes a single and a double?

Also, in your example, is the CAN ID the numerical value or the message name?

Thanks,
Richard.

[karlding]

Sure!

Here's 2 DBC files that I've generated using Kvaser Database Editor.

The CAN ID is the numerical value (ie. the arbitration ID).

[howerj]

Cheers! I'll let you know I've gotten around to implementing this, it shouldn't take too long.

[howerj]

Ok, I've added floating point support, no doubt there are problems and I have not tested the functionality (unit tests are on my to do list).

The packing/unpacking code does not handle all case either (Inf and NaN are not decoded or encoded). I should also add an option to allow type punning for conversion instead of using the unpack754/pack754 routines.