Improvements to run list and run build

[agirault]

Context

While preparing support for generating packages from the run and dev_container scripts, having a good list mechanism seemed like it would impact existing infrastructure. Instead of a large MR that mixes infra improvements and packaging support, this MR is scoped for infra improvements that the packaging support will then be able to leverage.

Changes

List improvements

• list applications, operators, and packages
• uses simpler but more robust heuristic based on required cmake functions needed to configure these targets (while the metadata.json should be present but might not be)

Build improvements

• support building operators in addition to apps
• verify input validity using the improve listing from the last commit
• allow passing a path to the app/operator dir, for improve UX by leveraging
  terminal autocompletion

Results

List improvements

• enhanced human-readable list of targets to build or package

$ ./run list

== APPLICATIONS =================

adv_networking_bench
basic_networking_ping
body_pose_estimation
colonoscopy_segmentation
cvcuda_basic
dds_video
deltacast_transmitter
endoscopy_depth_estimation
endoscopy_out_of_body_detection
endoscopy_tool_tracking
evt_cam_calibration
grpc_endoscopy_tool_tracking
grpc_h264_endoscopy_tool_tracking
h264_endoscopy_tool_tracking
h264_video_decode
high_speed_endoscopy
holoviz_hdr
holoviz_srgb
holoviz_ui
holoviz_vsync
holoviz_yuv
hyperspectral_segmentation
imaging_ai_segmentator
laser_detection
model_benchmarking
multiai_endoscopy
multiai_ultrasound
network_radar_pipeline
nvidia_nim_chat
nvidia_nim_imaging
nvidia_nim_nvclip
object_detection_torch
openigtlink_3dslicer
orsi_in_out_body
orsi_multi_ai_ar
orsi_segmentation_ar
prohawk_video_replayer
pva_video_filter
qt_video_replayer
realsense_visualizer
simple_pdw_pipeline
simple_radar_pipeline
tao_peoplenet
ucx_endoscopy_tool_tracking
ucx_h264_endoscopy_tool_tracking
ultrasound_segmentation
usb_cam_calibration
velodyne_lidar_app
video_deidentification
volume_rendering
volume_rendering_xr
webrtc_video_server
xr_hello_holoscan

== OPERATORS ====================

advanced_network
apriltag_detector
basic_network
cvcuda_holoscan_interop
dds_shapes_subscriber
dds_video_publisher
dds_video_subscriber
deltacast_videomaster
emergent_source
grpc_operators
lstm_tensor_rt_inference
npp_filter
openigtlink
orsi_format_converter
orsi_segmentation_postprocessor
orsi_segmentation_preprocessor
orsi_visualizer
prohawk_video_processing
qt_video
realsense_camera
tensor_to_video_buffer
tool_tracking_postprocessor
velodyne_lidar
video_encoder
volume_loader
volume_renderer
vtk_renderer
xr_basic_render
XrFrameOp
XrTransformOp
yuan_qcap

== PACKAGES =====================

holoscan-networking

=================================

• specialized listing that can more easily be parsed by code

$ ./run list_apps

adv_networking_bench
basic_networking_ping
body_pose_estimation
colonoscopy_segmentation
cvcuda_basic
dds_video
deltacast_transmitter
endoscopy_depth_estimation
endoscopy_out_of_body_detection
endoscopy_tool_tracking
evt_cam_calibration
grpc_endoscopy_tool_tracking
grpc_h264_endoscopy_tool_tracking
h264_endoscopy_tool_tracking
h264_video_decode
high_speed_endoscopy
holoviz_hdr
holoviz_srgb
holoviz_ui
holoviz_vsync
holoviz_yuv
hyperspectral_segmentation
imaging_ai_segmentator
laser_detection
model_benchmarking
multiai_endoscopy
multiai_ultrasound
network_radar_pipeline
nvidia_nim_chat
nvidia_nim_imaging
nvidia_nim_nvclip
object_detection_torch
openigtlink_3dslicer
orsi_in_out_body
orsi_multi_ai_ar
orsi_segmentation_ar
prohawk_video_replayer
pva_video_filter
qt_video_replayer
realsense_visualizer
simple_pdw_pipeline
simple_radar_pipeline
tao_peoplenet
ucx_endoscopy_tool_tracking
ucx_h264_endoscopy_tool_tracking
ultrasound_segmentation
usb_cam_calibration
velodyne_lidar_app
video_deidentification
volume_rendering
volume_rendering_xr
webrtc_video_server
xr_hello_holoscan

Build improvements

• Building operators now works also

$ ./run build dds_shapes_subscriber

Building dds_shapes_subscriber
[command] cmake -S . -B build/dds_shapes_subscriber --no-warn-unused-cli -D Python3_EXECUTABLE=/usr/bin/python3 -D Python3_ROOT_DIR=/usr/lib/python3 -D HOLOHUB_DATA_DIR:PATH=/home/agirault/projects/holoscan/holohub/data -D CMAKE_BUILD_TYPE=release -D OP_dds_shapes_subscriber:BOOL=1
Not searching for unused variables given on the command line.
-- The CXX compiler identification is GNU 11.4.0
...

• Can pass a path to the app or operator, improving UX by leveraging fast terminal autocomplete

$ ./run build applications/h264/h264_endoscopy_tool_tracking/

Building h264_endoscopy_tool_tracking
[command] cmake -S . -B build/h264_endoscopy_tool_tracking --no-warn-unused-cli -D Python3_EXECUTABLE=/usr/bin/python3 -D Python3_ROOT_DIR=/usr/lib/python3 -D HOLOHUB_DATA_DIR:PATH=/workspace/holohub/data -D CMAKE_BUILD_TYPE=release -D APP_h264_endoscopy_tool_tracking:BOOL=1
Not searching for unused variables given on the command line.
-- The CXX compiler identification is GNU 11.4.0
...

• Passing invalid value gets an early failure

$ ./run build nein

ERROR: Unknown application or operator: nein
You can run ./run list for a full list of options.

[agirault]

looks like the autocompletion_list below could also be updated to use some of the logic changed here... though it might need more love than that, since I was confused why apps where even listed, or what some of the options in there related to (cpp, python).

Happy to update it if I get some confirmation on the expected behavior and on the parts that don't currently seem correct in there.

[tbirdso]

Going to defer to @jjomier as I haven't touched the autocomplete implementation before. I'm OK with moving forward without in this PR and updating as necessary.

[tbirdso]

Update to include pkg looks good. LGTM, merge away