[geneus] treat uint8[] as string like rospy

[furushchev]

This is claimed at #14
Until merging this change, existing codes that treats messages containing uint8[] will fail.(e.g.: converting sensor_msgs/PointCloud2 to eus-pointcloud)

[garaemon]

Does it mean we need to modify existing code?

[garaemon]

https://github.com/jsk-ros-pkg/jsk_demos/blob/master/jsk_2015_06_hrp_drc/drc_com_common/msg/FC2OCSBasicInfoSmall.msg

[k-okada]

it depends on what kind of code you write, where is the code that uses FC2OCSBasicInfoSmall.msg?
Before i rewrite geneus, uint8[] is convert to string, and at here b157cd5, I have change them tointeger-vector but that may not correct. at least it fails at https://github.com/jsk-ros-pkg/jsk_roseus/blob/master/roseus/roseus_c_util.c#L50

[garaemon]

Hmm, I'm treating elements of uint8[] as integer.

This is another example:

• message https://github.com/jsk-ros-pkg/jsk_demos/blob/master/jsk_2015_06_hrp_drc/drc_com_common/msg/HumanoidJointAngles.msg
• code https://github.com/jsk-ros-pkg/jsk_common/blob/master/jsk_network_tools/euslisp/joint-state-decompressor.l#L48

[garaemon]

(elt <string> <index>) returns character constant and it may be no problem.

[k-okada]

humm, if we convert uint to integer-vector

• check integer-vector, not string in roseus_c_utils.c
• previous string implementation is very efficient because we do not run dolist, but memcpy to fill the data in deserialize phase, can we use that technique to integer-vector?

[garaemon]

Could you test roslaunch jsk_network_tools joint_state_compress_sample.launch Y_DIFF=1.0 before merging this?

If it work, I'm willing to accept this change.

[garaemon]

jsk-ros-pkg/jsk_common#852

This pull request may be required to run that.

[k-okada]

ok, please check this @furushchev

[furushchev]

I tried to test joint_state_compress_sample.launch, and looked rviz, but the poses of robots("no TF prefix" and with "decompressed") are quite different each other.(I use PR2 on local)
Is this correct? If correct, how can I test?

[furushchev]

Poses are different both on normal geneus environment and "this pullreq" geneus environment.

[garaemon]

It should be almost same

2015年4月9日木曜日、Furushchevnotifications@github.comさんは書きました:

Poses are different both on normal geneus environment and "this pullreq"
geneus environment.

—
Reply to this email directly or view it on GitHub
#15 (comment).

✉︎ from iPhone

[furushchev]

@garaemon I added test for angle vector compress/decompress jsk-ros-pkg/jsk_common#856.
I'll now check if occurs any problem using this test.
;; I could not check the diff of this pullreq using RViz, because poses are already different both on normal/"new" geneus environment.

[furushchev]

I checked error in both env:

• norm of difference between raw and decompressed angle-vector
  • 1.50817 (normal geneus env)
  • 1.50817 (new geneus env)
• norm of difference between raw and decompressed angle-vector

in existing geneus env:

head_tilt_joint: 0.000615
head_pan_joint: -0.003425
r_wrist_roll_joint: 0.002261
r_wrist_flex_joint: 0.005686
r_forearm_roll_joint: -0.003865
r_elbow_flex_joint: 6.28319
r_upper_arm_roll_joint: 0.003725
r_shoulder_lift_joint: 6.28319
r_shoulder_pan_joint: 0.003425
l_wrist_roll_joint: 0.002261
l_wrist_flex_joint: -0.005686
l_forearm_roll_joint: -0.003865
l_elbow_flex_joint: 6.28319
l_upper_arm_roll_joint: 0.003725
l_shoulder_lift_joint: 6.28319
l_shoulder_pan_joint: 0.011204
torso_lift_joint: 0.001526
(norm diff): 12.5664

• in new geneus env:

head_tilt_joint: -0.13585
head_pan_joint: -0.564602
r_wrist_roll_joint: -0.3536
r_wrist_flex_joint: -0.65
r_forearm_roll_joint: 0.003865
r_elbow_flex_joint: 1.30592
r_upper_arm_roll_joint: -0.540196
r_shoulder_lift_joint: 0.936318
r_shoulder_pan_joint: -0.790692
l_wrist_roll_joint: 1.2963
l_wrist_flex_joint: -1.54137
l_forearm_roll_joint: 0.98565
l_elbow_flex_joint: 4.97727
l_upper_arm_roll_joint: -0.540196
l_shoulder_lift_joint: 5.34687
l_shoulder_pan_joint: -0.011204
torso_lift_joint: 1.29626
(norm diff): 8.04649

the difference of joint_states are bigger in existing environment than new environment.

[furushchev]

I got it, lift_joint are rotational joint? (2 * pi different?)
Anyway there is the difference among enviromnents... What should I do...

[garaemon]

lift_joint problem should be resolved already.
https://github.com/jsk-ros-pkg/jsk_common/blob/master/jsk_network_tools/euslisp/angle-vector-compress.l#L9

[garaemon]

I will confirm this patch and please wait to merge this

2015年4月11日土曜日、Furushchevnotifications@github.comさんは書きました:

I got it, lift_joint are rotational joint? (2 * pi different?)
Anyway there is the difference among enviromnents... What should I do...

—
Reply to this email directly or view it on GitHub
#15 (comment).

✉︎ from iPhone

[furushchev]

The problem is two:

• lift_joint error is 2 * pi above when (de)compressed
• (de)serialization of uint8[n](fixed length) does not work by new geneus change

I fixed and updated new geneus pullreq and now I think it's ok (all geneus test passed and joint_state_compressor returns the same values as on existing geneus environment.)

[furushchev]

NOTE: Still please do not merge before the test ( jsk-ros-pkg/jsk_roseus#269 ) passes.

[furushchev]

Sorry I mistook comment.
At first please merge this for testing message.l generated by new geneus:

• [roseus] fix test for treating uint8[] as string jsk_roseus#269
• [jsk_network_tools] add test for angle-vector/JointStates compress/decompress jsk_common#856

After merging this PR, please re-run tests above to check