Fix Graph batch request contracts for public delegation

[ashwinikarke]

Summary

Fix request handling in the private Invoke-ZtGraphBatchRequest helper,
addressing the prerequisite issues raised in the review of #1584.

This PR updates the private helper only. Public delegation and parameter
forwarding remain part of #1584. The API-version fix is handled separately in #1591. This PR does not add
private GraphBaseUri support.

Changes

• Support OutputType, mapping PSCustomObject to the SDK's PSObject representation.
• Correct collection output, matched results, and NoPaging behavior.
• Preserve zero and string request IDs and reject case-insensitive duplicates.
• Validate response IDs and statuses before emitting results.
• Fail incomplete or malformed batches without partial output or resubmission.
• Use URL-bearing request descriptors as malformed-response error targets.
• Restore outer retries for GET/HEAD-only batches using Invoke-ZtRetry.
• Isolate intermediate retry errors from caller error variables after successful recovery.
• Avoid replaying batches containing writes after outer failures.
• Normalize and validate dependsOn IDs; require dependency-bearing workloads to fit within one batch.
• Remove satisfied dependency references from retry descriptors without replaying successful prerequisites.

[Copilot]

🟡 Changes recommended

Dependency handling breaks for zero-valued IDs and when batch splitting separates dependent requests.

2 open findings

• Preserve zero-valued dependsOn IDs · New
• Keep dependent requests together across batches · New

What changed in this PR

Updates the private Graph batching helper to enforce request/response contracts and configurable routing.

Changes:

• Adds Graph URI, API version, batch size, and output controls.
• Improves paging, IDs, response validation, and retry behavior.
• Adds comprehensive contract tests.

File	Description
src/​powershell/​private/​graph/​Invoke-ZtGraphBatchRequest.ps1	Implements corrected batching contracts and retry handling.
code-tests/​commands/​Invoke-ZtGraphBatchRequest.Tests.ps1	Tests routing, output, validation, paging, IDs, and retries.

🧠 Review effort: Balanced

---

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

[Copilot]

🟡 Changes recommended

A dependent 424 response is discarded when its prerequisite is throttled, preventing the dependent request from being retried.

1 open finding

• Retry 424 dependents after temporarily throttled prerequisites · New

2 resolved since last review

• Keep dependent requests together across batches
• Preserve zero-valued dependsOn IDs

🧠 Review effort: Balanced

[Copilot]

🟡 Changes recommended

The advertised caller-selected BatchSize remains unimplemented because batching is hard-coded to 20.

2 open findings

• Implement caller-selected BatchSize for member batching · New
• Retry 424 dependents after temporarily throttled prerequisites

🧠 Review effort: Balanced

[Aleksandar Nikolić (alexandair)]

ashwinikarke Please, address my feedback.

[Copilot]

🔵 Needs a closer look

Interacting retry, timeout, paging, and dependency behavior requires runtime validation and final human review.

0 open findings

2 resolved since last review

• Implement caller-selected BatchSize for member batching
• Retry 424 dependents after temporarily throttled prerequisites

🧠 Review effort: Balanced

[Copilot]

🟡 Changes recommended

Unresolved ID-handling defects can bypass completeness validation or create invalid dependencies, and explicit-ID validation scales quadratically.

1 open finding

• Request ID Count shadows hashtable Count property · New

Previously missed (2)

In code that hasn't changed since last review

Explicit ID validation scales quadratically

src/​powershell/​private/​graph/​Invoke-ZtGraphBatchRequest.ps1:456

This enumerates every previously created task's ID for each explicit ID, making request preparation quadratic before transport starts. The public caller assigns explicit IDs to every request (public/Invoke-ZtGraphRequest.ps1:287–293), so 10,000 requests cause roughly 50 million prior IDs to be enumerated, with repeated array allocation. Track seen explicit IDs in a case-insensitive hashtable instead; generated IDs already avoid explicit IDs through ReservedIDs.

Generated numeric IDs collide with explicit string IDs

src/​powershell/​private/​graph/​Invoke-ZtGraphBatchRequest.ps1:462

Generated request IDs remain integers, so these membership checks can compare explicit string IDs numerically. For example, -Request @('users', @{ id = '01'; url = 'groups' }) rejects '01' as a duplicate of generated ID '1', although they are distinct correlation IDs. Dependency validation also accepts '01' as a reference to generated ID '1', leaving a dangling dependency in the submitted batch. Store generated request IDs as strings and add regression tests for both cases.

🧠 Review effort: Balanced

[Copilot]

🔵 Needs a closer look

Mixed numeric and string task IDs can incorrectly reject distinct correlation IDs.

0 open findings

1 resolved since last review

• Request ID Count shadows hashtable Count property

Previously missed (1)

In code that hasn't changed since last review

Store generated task IDs as strings for consistent ID comparisons

src/​powershell/​private/​graph/​Invoke-ZtGraphBatchRequest.ps1:462

Generated task IDs remain integers, so this containment check can compare IDs numerically. For example, -Request @('users', @{ id = '01'; url = 'groups' }) rejects '01' as a duplicate of generated ID 1, although the correlation strings '1' and '01' are distinct. Store generated task IDs as strings so duplicate and dependency checks consistently use string equality, and add a regression test for this mixed-ID workload.

🧠 Review effort: Balanced

[Aleksandar Nikolić (alexandair)]

LGTM