Tag: comparison

Pick by the return type of build(): T means Notifier, FutureOr<T> means AsyncNotifier, Stream<T> means StreamNotifier. Here is the decision matrix, the type hierarchy that explains why, and the == filtering and state-clobbering gotchas that bite each one. Verified against flutter_riverpod 3.4.2 on Flutter 3.44.2.

Scalar ships 1.02 MiB of gzipped JavaScript and a far better request builder. Swagger UI ships 514 KiB and renders OpenAPI 3.2, which is what .NET 11 now emits by default. Measured payloads, the 3.2 gap, endpoint routing on both sides, and the auth details that decide it.

Zstandard is the right default for dynamic API responses in .NET 11, but not at the quality the ASP.NET Core provider ships with. Benchmarks on real JSON payloads showing why quality 1 beats the default quality 3 on both size and CPU, when Brotli still wins, and why Gzip survives only as a fallback.

Aspire 13.4.6 wins the .NET inner loop because it runs your projects as host processes you can debug, while Docker Compose wins when the compose file is also your CI and deployment contract. Measured startup and edit-to-running timings on both, the configuration each one injects for you, and the six gotchas that decide it.

They are not alternatives. WebApplicationFactory boots your app, Testcontainers boots your dependencies. Measured on .NET SDK 10.0.201: a container fixture costs 1.7 s per class against 10 ms for SQLite, and a HasMaxLength(16) violation that Postgres rejects with 22001 is silently accepted by SQLite.

Pick xUnit v3 for greenfield .NET projects, NUnit 4.6 if you live in its constraint model, MSTest 4 if you already ship it. A measured comparison on .NET SDK 10.0.201 covering parallelism defaults, test class lifecycle, assertion failure output, and the Microsoft.Testing.Platform version conflict that breaks the NUnit runner.

For internal service-to-service calls in .NET 11, default to gRPC when you own both ends of the contract and the call is point-to-point. Use REST with JSON when anything you do not control has to call the service. SignalR is not an RPC transport between services: reach for it only when one producer has to fan a message out to many long-lived consumers.

A named subagent starts cold, a fork inherits your whole conversation and shares your prompt cache, and a background agent is a second Claude Code session entirely. The trap is the naming: /fork stopped being the in-session fork at v2.1.212 and /subtask took over.

Use declarative YAML when the graph is a sequential routing decision that non-developers change often. Use code-first WorkflowBuilder the moment you need parallelism, custom executors, or a non-Foundry agent. Here is the decision, the action catalog limits, and the code both ways.

Manual gates every tool call on you. Auto mode gates them on a classifier with 32 built-in block rules. The mode that surprises people is neither: acceptEdits auto-approves rm, mv, and sed, not just file edits. Measured on Claude Code 2.1.123.