Tag: comparison

A yield return method costs one 40-byte state machine object per call no matter how many items it produces; a List<T> costs the whole backing array up front. Measured on .NET 10.0.10: iterator wins on memory and early exit, List<T> wins when the caller enumerates more than once, needs Count, or calls ToList().

Both end in the same OS call, so the timestamp written is identical. The differences are around it: which cache gets invalidated, whether directories work on Windows, how a non-UTC DateTime is converted, and whether you can target an open handle. Use File.SetLastWriteTimeUtc, by handle when the file is open. Verified on .NET 10.0.10.

On .NET 10 a plain for loop over span[i] bounded by span.Length compiles to a bounds-check-free, strength-reduced, down-counting loop. A hand-written Unsafe.Add(ref r, i) loop with an int index keeps a sign extension and an up-counting compare, and measured 2-5% slower. Benchmarks and JIT disassembly for arm64 and x64 on .NET 10.0.10.

Use shared_preferences (SharedPreferencesWithCache) for a handful of flat flags and a JSON file in getApplicationSupportDirectory for a structured settings object. Covers the atomic-write pattern, the concurrent-save race, the Documents-folder trap on Windows and Linux, web support, and measured write costs on Flutter 3.44.8.

If a specialist agent is just a model loop wrapped around a few read-only calls into your own services, collapse it into a skill served over MCP: a SKILL.md plus typed tools that the parent model calls directly. Keep it an A2A agent when it needs its own model, private context, a long multi-step workflow or an independent lifecycle. Here is the decision matrix, the measured cost of each path, and the Agent Framework 1.17 wiring.

Run the orchestrator on claude-opus-5-5, give implementer and reviewer subagents claude-sonnet-5-5, and send search, triage and summarizing work to claude-haiku-5-5 at medium effort. Here is the role matrix, the cost math per subagent run, Haiku 5.5's 100k-token price cliff, and the Claude Code model-resolution rules that silently override your choice.

Turn on the local sandbox for everyday Copilot CLI work, move to a cloud sandbox when the code or the task should never touch your laptop, and use a dev container when the team needs the same reproducible toolchain. Compared on Copilot CLI 1.0.104: what each boundary isolates, what it leaks, what it costs, and how to stack them.

Use /fleet for a one-off parallel task in your terminal, a dynamic workflow when the same multi-agent process has to run the same way twice, and GitHub Agentic Workflows when that process runs unattended on repository events. Compared on Copilot CLI 1.0.103 and gh-aw v0.93.6: who plans the split, limits, resume, permissions, write safety and billing.

Use Expanded when a child must fill its share of a Row or Column, and Flexible when it may be smaller than its share. Measured layouts on Flutter 3.44 show where the leftover space goes, why loose children never donate it to siblings, and when only Flexible survives unbounded constraints.

Use Lazy<T> by default: it runs the factory once, caches exceptions, and its reads are too cheap to measure after initialization. Use LazyInitializer.EnsureInitialized when you lazily fill many fields or array slots and cannot afford an extra object per value. Hand-written double-checked locking is almost never worth it on .NET 10.