static_transform_publisher more intuitive command-line arguments

[clalancette]

Bug report

The order of command-line arguments in the static_transform_publisher program is hard to remember and follow. This led directly to the problem described in #292. It would be better if the command-line arguments were much more specific so there is no ambiguity. We discussed this in #292 (comment) and #292 (comment) , and the consensus was to add the following command-line arguments:

Group 1:

• --roll - specify the roll in radians
• --pitch - specify the pitch in radians
• --yaw - specify the yaw in radians

Group 2:

• --roll_deg - specify the roll in degrees
• --pitch_deg - specify the pitch in degrees
• --yaw_deg - specify the yaw in degrees

Group 3:

• --qx - Specify the quaternion X
• --qy - Specify the quaternion Y
• --qz - Specify the quaternion Z
• --qw - Specify the quaternion W

Note that each of the groups is mutually exclusive; if the user specifies --roll, they cannot then specify --roll_deg or --qx (while we technically could support mixed --roll and --pitch_deg, we'll just specify that they are exclusive for simplicity).

In order to "tick-tock" this, we'll also continue to support the old command-line options up to and including Galactic, though with a deprecation warning. After Galactic, we'll remove the old command-line options. The old command-line options should also be mututally exclusive with any of the 3 groups above.

@allenh1 has graciously offered to take on implementing this. @Stapelzeiger FYI.

[allenh1]

@clalancette Starting on this now, but I'm wondering how you want the translation handled? I feel like it's appropriate to add --x [value] --y [value] --z [value] (certainly makes the implementation easier, since I can assume things are in pairs), but I also feel like it's kinda silly to do so.

Would you prefer to have a --translation x y z or --xyz x y z?

Also should --frame-id frame and --child-frame-id child_frame follow the same pattern? I don't really have any strong opinions here.

[clalancette]

@clalancette Starting on this now, but I'm wondering how you want the translation handled? I feel like it's appropriate to add --x [value] --y [value] --z [value] (certainly makes the implementation easier, since I can assume things are in pairs), but I also feel like it's kinda silly to do so.

I think we should make translation follow the same pattern as the rotation for consistency. Thus, I would go with --x, --y, --z here.

Also should --frame-id frame and --child-frame-id child_frame follow the same pattern? I don't really have any strong opinions here.

Yeah, I think so.

---

There's a couple of things I failed to think about in the original description.

1. What happens if the user specifies no rotation or no translation? In that case, I think we should assume the identity transformation.
2. What happens in the user specifies --x, but not --y or --z? We could either throw an error (meaning that if you specify one of a "group", you have to specify it all), or we could assume identity, but replace the axes that are specified with the user supplied ones. I'm on the fence with that one. I slightly lean towards throwing an error, but I could easily be persuaded otherwise.

[allenh1]

We could either throw an error (meaning that if you specify one of a "group", you have to specify it all), or we could assume identity, but replace the axes that are specified with the user supplied ones. I'm on the fence with that one. I slightly lean towards throwing an error, but I could easily be persuaded otherwise.

I would expect things to work with just --x [value] and no --y or --z specified, but I see where you're coming from. Maybe a warning should be printed?

[clalancette]

I would expect things to work with just --x [value] and no --y or --z specified, but I see where you're coming from.

Yeah, that's reasonable. Since this is going to be pretty verbose, I think we want to allow people to specify the minimum. I'm just wary of surprises, but in this case I think people will generally assume identity plus whatever they specified.

Maybe a warning should be printed?

I wouldn't suggest a warning, but maybe something informational. Like:

Applying translation of 0, 0, 0
Applying rotation of 0, 0, 0, 1

(you can probably come up with some better wording than that)

[allenh1]

@clalancette I'll be starting on this in the next day or so. Sorry for the delays.