ROS2 message format should support enums

[dirk-thomas]

From @calvertdw on January 8, 2018 16:53

You guys know what this means.

Constants aren't good enough. They don't constrain you to the range of values and you are able to set the field to any number you want.

// Current way

int16 FOO=0
int16 BAR=1
int16 value

// Proposed way

enum foo_bar { FOO, BAR }

Copied from original issue: ros2/ros2#447

[dirk-thomas]

We are aware of this. Currently the .msg doesn't provide the necessary expressability to implement this. Therefore on the roadmap we have an items to consider other formats (see https://github.com/ros2/ros2/wiki/Roadmap#design--concept):

Revise IDL file format, consider desired features (grouping, comments, etc.), suitability of existing formats like IDL 4.2

[dirk-thomas]

With the IDL changes merged this becomes more feasible. Potential follow up of #346.

[allenh1]

@dirk-thomas / @calvertdw is anyone working on this, or could I start attempting something on this one?

On that note, I had one concern I'd like to address before I dive in on this (assuming I can). Suppose that one host has an outdated copy of the .msg file where the enum has contents enum foo_bar { FOO, BAZ }. Further, suppose that the copy of the .msg file on the second host is up to date and has this field as enum foo_bar { FOO, BAR, BAZ }. The standard dictates that FOO = 0, BAZ = 1 on host 1, but also that FOO = 0 and BAZ = 2 on the second host.

While this is definitely not good practice to do, such a mistake occurs quite easily and is not so simple to track down. So I'm not quite sure how to store such a type internally to detect such an incompatibility. @dirk-thomas maybe you have some idea?

[dirk-thomas]

is anyone working on this, or could I start attempting something on this one?

Nobody is currently working on this or planning to in the near future.

First, any change should happen for .idl files. The format of .msg files is not intended to be extended with additional features anymore.

Adding an enum value in the middle is only compatible if you explicitly specify the enum values (so that they don't change). Also I think the IDL spec already defines enums - so that part would likely be covered by the spec already.

[davidhodo]

@allenh1 Have you started work on this? This is a feature we need as well and would be willing to contribute towards. I'm not able to open the design concept link Dirk posted above, but is there any initial thoughts on how this should be approached?

[mjcarroll]

@davidhodo This the corrected link: https://github.com/ros2/ros2_documentation/blob/master/source/Roadmap.rst#design--concept

[oysstu]

Is the addition of bit masks also being considered along with this (IDL v4.2 section 7.4.13.4.3.3)?

[dirk-thomas]

Is the addition of bit masks also being considered along with this

No, bit masks are orthogonal to enums. Also nobody is currently working on enum support either.

[RFRIEDM-Trimble]

Is this on the roadmap?

If not, can anyone give some recommendations on how to deal with checking enums are valid before publishing or after receiving, handling conversions from the ROS integer back to enums, and how to avoid maintaining a list of constants in the message file in addition to an enum in other source code?

[clalancette]

Is this on the roadmap?

The roadmap is published here. In short, nobody is currently working on adding enum support. That said, if you are interested in working on it, we'd be happy to review pull requests.