Add map argument autocomplete

[dheeraj12347]

Summary

Add autocomplete support for map API arguments in CloudMonkey.

Changes

• Add support for map argument autocomplete.
• Extract indexed map fields from API argument descriptions.
• Support generic key/value map descriptions.
• Suggest indexed fields such as:
  • tags[0].key=
  • tags[0].value=
• Avoid assuming that every map argument follows the key/value structure.
• Add regression tests covering:
  • Indexed map fields
  • Multiple map fields
  • Multiple indexes
  • Generic key/value maps
  • Non-key/value map arguments
  • Autocomplete option generation

Testing

• gofmt -w cli/completer_test.go
• go test ./...
• git diff --check

All tests are passing.

[dheeraj12347]

Hi @Pearl1594 and @DaanHoogland , I’ve raised the map argument autocomplete work as a separate PR: #229.

It adds autocomplete support for map API arguments, including indexed fields like tags[0].key= and tags[0].value=, while avoiding assumptions about maps that don’t follow the key/value structure. I’ve also added regression tests covering the different cases.

go test ./... and git diff --check are passing.

Whenever you’re available, could you please have a look and let me know if the approach looks good or if anything should be changed?

PR: #229

[github-actions]

✅ Build complete for PR #229.

📦 Binary artifacts are available in the workflow run (expires on October 23, 2026).

Note: Download artifacts by clicking on the workflow run link above, then scroll to the "Artifacts" section.
Artifacts from PR builds are for testing only and may contain unreviewed, malicious code.

[Copilot]

Copilot review overview

🟡 Changes recommended

Argument routing prevents indexed map suggestions from appearing with real cached argument names.

Review effort: Balanced
Findings: 1

Open (1)

• Match map-field candidates before argument prefix guard · New

What changed in this PR

Adds description-based autocomplete for CloudMonkey map arguments without assuming every map uses key/value fields.

Changes:

• Extracts indexed fields from argument descriptions.
• Adds generic key/value suggestions.
• Adds regression tests for field extraction and suggestion generation.

File	Description
cli/​completer.go	Adds map-field extraction and completion handling.
cli/​completer_test.go	Tests indexed, generic, and non-key/value maps.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[Damans227]

the new map suggestions never show up when you press tab. typing tags gives only "=", and tags= or tags[0].k give nothing, because the new part only runs after the full tags= is typed and none of its suggestions start with tags=. copilot's comment above is right. can we add a test that goes through the normal tab completion with tags= so this gets caught?

[dheeraj12347]

the new map suggestions never show up when you press tab. typing tags gives only "=", and tags= or tags[0].k give nothing, because the new part only runs after the full tags= is typed and none of its suggestions start with tags=. copilot's comment above is right. can we add a test that goes through the normal tab completion with tags= so this gets caught?

Thanks for the feedback! I've pushed a follow-up commit, c79972f, that moves map-field matching before the argument-prefix check and adds a regression test through the normal autoCompleter.Do path using tags=. The focused test and full go test ./... suite pass locally. Could you please review the updated implementation when you get a chance?

[Damans227]

I've pushed a follow-up commit, c79972f, that moves map-field matching before the argument-prefix check

@dheeraj12347 tags and tags[0].k work now, but tags= still breaks in the real shell. pressing tab there gives tags=[0]. because the shell adds the text after the = instead of replacing it. can the test check the full line after tab, not just the suggestions?

[dheeraj12347]

I've pushed a follow-up commit, c79972f, that moves map-field matching before the argument-prefix check

@dheeraj12347 tags and tags[0].k work now, but tags= still breaks in the real shell. pressing tab there gives tags=[0]. because the shell adds the text after the = instead of replacing it. can the test check the full line after tab, not just the suggestions?

@Damans227 , I traced the tags= issue to readline's completion behavior. AutoCompleter.Do() returns candidate suffixes, and readline appends them to the existing buffer, so the trailing = cannot be removed by the CloudMonkey completer alone. A proper fix appears to require changing readline's replacement behavior. Would you be comfortable with a narrowly scoped change to the vendored readline code, or would you prefer an approach that avoids modifying the dependency?

[Damans227]

Would you be comfortable with a narrowly scoped change to the vendored readline code, or would you prefer an approach that avoids modifying the dependency?

@dheeraj12347 the vendored readline already has cloudmonkey changes from #199 that rewrite the text after the last =, so a small change there looks fine to me. can it build on that instead of adding a separate path?

[dheeraj12347]

Thanks @Damans227 for the suggestion!

I've pushed bc7d4b0, which extends the existing vendored readline completion helper to handle the trailing = case while preserving its previous behavior.

I've also added a regression test that exercises OnComplete() and verifies the full command line after completion. The focused readline test, go test ./..., and git diff --check all pass locally. I also verified the behavior manually in the interactive shell.

Could you please take another look when you get a chance?

[Damans227]

I've pushed bc7d4b0, which extends the existing vendored readline completion helper to handle the trailing = case

@dheeraj12347 that covers it. tags, tags= and tags[0].k all complete right in the shell now.

[Copilot]

Warning

Copilot couldn't run its full agentic review because it didn't start before the timeout. Make sure your repository has a runner available, or add a copilot-code-review.yml file specifying one with the runs-on attribute. See the docs for more details.

Copilot review overview

3 open findings

• Compiling the regexp on every call to mapFieldsFromDescription adds unnecessary overhead (this… · New
• Match map-field candidates before argument prefix guard
• For a pure set, map[string]struct{} is the idiomatic choice in Go and avoids storing an extra… · New

🧠 Review effort: Lite