complete.py is a python script that takes as input a GN label and emits a list of suggested completions based on the GN build targets available in the build. For interactive performance, the list of targets must be generated ahead of time by the fx set command and pre-processed by prepare.py.
prepare.py is a python script that takes as input a JSON project file generated by fx set and creates a copy with nothing but the names of the GN build targets. This is then read by complete.py during shell completion.
The completion scripts require some extra metadata to be generated by the fx set command. When invoking your usual fx set command, include the following arguments:
fx set ... --ide json --json-ide-script //scripts/gn_complete/prepare.py
This will generate two files: out/default/project.json and out/default/project_lite.json. The latter file is a stripped down version of the former and is the input to complete.py.
Make sure that scripts/fx-env.sh is sourced in your shell. See fx_workflow.
The default completion configuration for BASH treats : characters as word breaks, which breaks completion for GN targets (foo/bar:test).
To remove the : character from the set of completion word break characters, add the following to your .bashrc:
COMP_WORDBREAKS=${COMP_WORDBREAKS//:}
ZSH currently has its own implementation of GN build target shell completion, which lives at //scripts/zsh-completion/_fx_build. Unfortunately, since it queries ninja to assemble the target names, the latency is too high for interactive use-cases.
Tests can be run from the Fuchsia project root with:
python3 -m unittest discover -s scripts/gn_complete