Add shell completion telemetry coverage

[tnaum-ms]

Adds telemetry instrumentation to the interactive shell's completion system, closing the gap where Tab completions, ghost text, and closing-bracket suggestions were untracked.

Changes

• completion.accepted — shell completions now flow into the existing accumulating event with src_shell, alongside playground and collection view. Includes trigger_tab / trigger_ghostText to distinguish acceptance method.
• shell.completionGhost — tracks shown vs accepted ratio for inline ghost text suggestions, broken down by category (field, operator, collectionName, etc.).
• shell.completionList — tracks when the multi-match picker is displayed (e.g., db.col.fin⇥ showing available methods), broken down by category.
• shell.closingBrackets — tracks shown vs accepted ratio for closing-bracket ghost text (}}) suggestions).
• Adds 'shell' as a CompletionSource in completionCategories.ts.

All events use callWithAccumulatingTelemetry for batched, low-overhead emission.

Telemetry correctness fixes (added during review)

• ShellGhostText.show() now returns boolean; shown telemetry is gated on actual render to prevent overcount when the same ghost text stays visible.
• completion.accepted is only emitted after a successful insert or replace — no-op Tab presses (fully typed word matching a single candidate) no longer inflate acceptance counts.
• candidateCount was removed from shell.completionList: callWithAccumulatingTelemetry sums measurements across the batch, making a per-display length value analytically misleading. Tracked as #636 for future min/max/median support.

[Copilot]

Pull request overview

Adds telemetry instrumentation to the interactive shell’s completion system so completion acceptance and suggestion surfaces (Tab, ghost text, completion list, closing-brackets) are tracked consistently alongside existing completion telemetry.

Changes:

• Adds 'shell' as a CompletionSource in the shared completion telemetry categories.
• Instruments shell completion acceptance into the accumulating completion.accepted event (with src_shell and trigger_tab / trigger_ghostText measurements).
• Adds new accumulating shell events for ghost text, completion list display, and closing-brackets ghost suggestions.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 3 comments.

File	Description
src/telemetry/completionCategories.ts	Extends shared completion telemetry source taxonomy to include the shell surface.
src/documentdb/shell/DocumentDBShellPty.ts	Adds accumulating telemetry emission for shell completion acceptance, ghost text shown/accepted, completion list shown, and closing-brackets suggestions.

Comments suppressed due to low confidence (1)

src/documentdb/shell/DocumentDBShellPty.ts:1139

• Same overcounting issue as completion ghost text: trackClosingBracketsShown() is called even when ShellGhostText.show(closing) doesn’t re-render because the same closing string is already visible. That will inflate shell.closingBrackets.shown and distort the accepted ratio. Gate the telemetry on an actual change in the ghost text (or have ShellGhostText.show() report whether it rendered).

                const closing = getClosingBrackets(buffer);
                if (closing.length > 0) {
                    this._ghostTextIsHint = false;
                    this._ghostTextIsClosingBrackets = true;
                    this._ghostCandidateKind = undefined;
                    this._ghostText.show(closing, (d) => this._writeEmitter.fire(d));
                    this.trackClosingBracketsShown();
                    return;

[tnaum-ms]

Independent finding fixed: Single-candidate Tab acceptance counted even when Tab inserts nothing.

applySingleCompletion() recorded completion.accepted telemetry before verifying the buffer actually changed. When the user fully types a word that matches a single candidate (e.g., typing use matching candidate use), the remaining text is empty — no insertion happens, but the acceptance counter still fired.

Fixed in 3f40929c by moving trackCompletionAccepted() to fire only after a successful insertText() or replaceText() call.

[tnaum-ms]

Review after fixes applied:

Three issues addressed in this round:

1. Ghost/closing-brackets shown overcount (Medium → Fixed in 33ecdafb) — ShellGhostText.show() now returns boolean; telemetry gated on actual render.
2. Tab acceptance fires on no-op (Medium → Fixed in 3f40929c) — trackCompletionAccepted() moved after insertText()/replaceText().
3. candidateCount removed from accumulating telemetry (Fixed in 3f40929c) — accumulating telemetry is for occurrence counts; summing candidates.length is not meaningful here. If per-display distribution is needed in the future, it should use non-accumulating telemetry or a min/max/median collector.

One remaining low-severity nit (file comment): inconsistent state reset across _ghostText.clear() call sites. Harmless but worth consolidating into a helper.

All 1937 tests pass. Build succeeds.

[github-actions]

✅ Code Quality Checks

Check	Status	How to fix
Localization (l10n)	✅ Passed
ESLint	✅ Passed
Prettier formatting	✅ Passed

This comment is updated automatically on each push.

[Copilot]

Pull request overview

Copilot reviewed 3 out of 3 changed files in this pull request and generated 1 comment.

[github-actions]

📦 Build Size Report

Metric	Base (next)	PR	Delta
VSIX (vscode-documentdb-0.8.0.vsix)	7.51 MB	7.51 MB	⬆️ +0 KB (+0.0%)
Webview bundle (views.js)	5.79 MB	5.79 MB	⬆️ +0 KB (+0.0%)

Download artifact · updated automatically on each push.