Add configuration option to skip search paths

[lunaluxie]

Summary

Add a [skip_paths] configuration option to allow the user to extend the skipped paths from the search.

The documentation and test-suite has been updated to reflect the changes. And I have checked for regressions. The PR has been rebased onto the latest commit on main.

I am not sure about the level of granularity wanted for the commits. Let me know if you need me to squash or reword the commits to better reflect the conventions of this project.

Benefits

• The user can run Asimov in the home directory while excluding quarantined directories such as Photos, Documents, etc. This avoids permission errors, or having to give access to those directories reducing the attack surface.
• Excluding large sub-directories which are irrelevant to Asimov can modestly speed up execution time.

Considerations

• If code repositories are all located in one subfolder eg. ~/git, or ~/Developer, then it may be more appropriate to run Asimov only in that folder. This has been explained in the readme.
• While Asimov internally uses skip_paths to refer to paths skipped during search, it may be more appropriate to call the user-facing configuration option skip_dirs for consistency with the existing fixed_dirs configuration option.

Test plan

• make check passes

• [?] Shellcheck clean (ran make lint which passes)

• CHANGELOG.md updated (if user-facing)

• Add test to verify behavior (skips directories given in [skip_paths] config behavior test)

• Update readme/changelog to reflect changes.

[django23]

@lunaluxie thanks for this, and sorry for the wait — it's still wanted. 🙏

Heads-up that v0.10.0 changed the ground under this PR: there's a real config parser now (~/.config/asimov/config, with [sentinels] and [fixed_dirs] sections), so skip paths should slot in as another key rather than a separate mechanism:

[skip_paths]
extra = ~/Applications

If you're up for rebasing onto main in that shape, I'd be glad to review and land it. This would also close out #14, and is the natural counterpart to the just-shipped [scan] (#108).

Totally fine if you'd rather not — just say so and I'll pick it up. 🙏

[lunaluxie]

@django23 Thank you for taking the time to review my pull request.

I have attempted to rebase onto upstream/main, and I have manually verified that my changes still work, and I have rerun the tests with make check.

However, I am not quite sure what you mean by reworking the config parser. It appears that the main branch still uses the switch case mechanism which skip_paths was already slotted into.

case "${section}:${key}" in
                fixed_dirs:enabled)
                    ASIMOV_CONFIG_FIXED_DIRS_ENABLED="$value"
                    ;;
                fixed_dirs:extra)
                    value="${value/#\~/$HOME}"
                    ASIMOV_CONFIG_EXTRA_FIXED_DIRS+=("$value")
                    ;;
                scan:extra)
                    value="${value/#\~/$HOME}"
                    ASIMOV_CONFIG_SCAN_DIRS+=("$value")
                    ;;
                sentinels:extra)
                    ASIMOV_CONFIG_EXTRA_SENTINELS+=("$value")
                    ;;
                sentinels:disabled)
                    ASIMOV_CONFIG_DISABLED_SENTINELS+=("$value")
                    ;;
                skip_paths:extra)
                    value="${value/#\~/$HOME}"
                    ASIMOV_CONFIG_EXTRA_SKIP_PATHS+=("$value")
                    ;;
            esac

Finally, I am a bit worried that I have made a mistake in rebasing. I can see that my pull request now includes all the new changes from main as well which I doubt is the proper way to do it. But I haven't used git before in a collaborative environment, so I am not sure what I have done wrong.

[django23]

@lunaluxie you rebased it correctly - nothing to fix. The commit list shows main's commits because of the merge commit, but the actual diff (gh pr diff 107) is still just your 4 files. That's the thing that gets reviewed and merged, so you're good. 🙏

And you're right about the parser: [skip_paths] slots straight into the existing case "${section}:${key}", which is exactly what you did. My earlier comment described work you'd already done - my mistake for not re-reading the branch first.

Two things I've pushed on top of yours:

• asimov doctor didn't know about the new section. It validates the config against an allowlist and was reporting unknown section [skip_paths], exit 1.
• The behaviour test passed for the wrong reason. create_project prefixes $HOME, so an absolute first argument built the project at $HOME/$HOME/.skipme/... - outside the skipped path. Now it uses a relative base, with a control case that fails when the config is absent.

I also merged the configured paths into ASIMOV_SKIP_PATHS at the point it's defined rather than in each consumer, so find and the Spotlight pass can't drift apart:

readonly ASIMOV_SKIP_PATHS=(
    "${ASIMOV_ROOT}/.Trash"
    "${ASIMOV_ROOT}/Library"
    ${ASIMOV_CONFIG_EXTRA_SKIP_PATHS[@]+"${ASIMOV_CONFIG_EXTRA_SKIP_PATHS[@]}"}
)

make check is green (225 tests + shellcheck). This closes #14 too.

Thanks for sticking with it. 🎉
If you agree with the changes too, I'll merge.