@marcgravell

C# fan; ex-MVP, ex-Microsoft, ex-Stack Overflow; author of Dapper, protobuf-net, SE.Redis, and a few other things. Doing Redis things at Redis.

 UK
blog.marcgravell.com
Joined March 2009
  • Tweets19.7K
  • Following776
  • Followers24.1K
872Media
Woohoo, .NET Conf 2026, I'll be talking about SE.Redis v4 in the community days - probably active:active, client-side-caching, with that as a segue into custom commands and the new high-perf reader/writer API (no session links yet, but: dotnetconf.net/)

.NET Conf 2026

Join the .NET Conf 2026 free virtual event November 10 - 13 2026 to learn about the newest developments across the .NET platform, open source, and dev tools. Mark your calendar!

dotnetconf.net
2
4
81
3,343
Today's job: psychic debugging PR to review: "reverts #...." - that's it, the entire body. I super want to give them the benefit of the doubt and assume there is a reason or some context, but they're not replying to me. Very frustrating. Closing it is easy, but... dammit!
1
4
1,649
RC2 didn't fix it, so logged: github.com/dotnet/runtime/is… - this is a hard "oof" for SE.Redis, otherwise (as in: performance collapses)

.NET 11 thread pool: multiplexed clients leave cores idle, 30–45% lower throughput than .NET 10...

Description A client where many callers share one socket and each does a few µs of its own CPU work per operation loses 30–45% of its throughput on .NET 11 compared with .NET 10. The thread pool ke...

github.com
Marc Gravell@marcgravellOct 13
I thought I'd re-run my SE.Redis "async runtime" (.NET 11) checks; instead, found a catastrophic perf collapse in RC1, that I *think* (testing now against "daily") is already fixed for RC2 (github.com/dotnet/runtime/pu…). Fun!
3
6
49
6,289
I thought I'd re-run my SE.Redis "async runtime" (.NET 11) checks; instead, found a catastrophic perf collapse in RC1, that I *think* (testing now against "daily") is already fixed for RC2 (github.com/dotnet/runtime/pu…). Fun!
4
1
1
75
12,691
Hmmm.... still seeing -25% on "daily"....
1
3
1,027
Marc Gravell retweeted
Jeff Needles
@jsneedlesSep 29
Sure, OpenAI may have released yet another bot assistant thing. But I just shipped a convenient package for C# devs to source generate enum metadata. htmlcsstoimage.com/blog/fanc… Which do you think is a bigger deal?

Fancy Enum Generator: Convenience and Performance for C# Enums

How we replaced hand-written enum switches across HCTI with an attribute-driven source generator, why we open-sourced it and how we made it fast.

htmlcsstoimage.com
3
5
11
977
Marc Gravell retweeted
protobuf-net@protobuf_netSep 15
Need the "Connect protocol" (gRPC semantics over plain http)? see: protobuf-net.Connect connect.protobuf-net.dev - works with both protobuf-net and Google.Protobuf .NET types, fully AOT

protobuf-net.Connect

A .NET implementation of the Connect protocol - AOT-first, gRPC semantics over ordinary HTTP

connect.protobuf-net.dev
1
423
Having fun exploring interpolated string handlers: github.com/StackExchange/Sta…

V4 core spike: RESP write path, client-side cache, and the context surface by mgravell · Pull...

Draft / spike. Public but gated behind SER010, so nothing here is a commitment. Design notes — what was measured, what was only reasoned about, and a decision log of what was tried and rejected — a...

github.com
1
1
13
2,581
Versioning is hard... spotted in the wild: ~/.nuget/packages/xunit.v3.core.mtp-v2/4.0.0/buildTransitive/xunit.v3.core.mtp-v2.targets so... which version is that?
11
1
1
60
6,443
I love this kind of use of interceptors - kudos.
Lucas Trzesniewski@Lucas_TrzJun 15
I've released v1.5.0 of PCRE​.NET with this feature. Now, an anti-pattern becomes a pattern! 🥳 Here's a quick benchmark between instance calls (`_regex.IsMatch`) and static calls (`PcreRegex.IsMatch`) over 10 million iterations:
1
1
19
6,000
More quality-of-life things in SE.Redis, so you don't find problems at runtime - Roslyn is a heck of a drug:
5
4
56
4,921
Marc Gravell retweeted
protobuf-net@protobuf_netAug 20
Protobuf "editions" support in protobuf-net: docs.protobuf-net.dev/editio… (including DELIMITED in .proto, which has been in protobuf-net code-first since 2008)

Editions

Protocol Buffers library for idiomatic .NET

docs.protobuf-net.dev
1
463
Marc Gravell retweeted
protobuf-net@protobuf_netAug 20
Hot on the heels of protobuf-net, protobuf-net.Grpc goes AOT: grpc.protobuf-net.dev/aot

Native AOT and trimming

gRPC bindings for protobuf-net and grpc-dotnet

grpc.protobuf-net.dev
protobuf-net@protobuf_netAug 11
protobuf-net finally goes AOT in v3.3 with build-time generator support! protobuf-net.github.io/proto…

Compile-time serializers, for native AOT and trimming

Protocol Buffers library for idiomatic .NET

docs.protobuf-net.dev
2
1
574
Marc Gravell retweeted
protobuf-net@protobuf_netAug 15
protobuf-net's online schema and payload tools are resurrected and improved, finally, and with a new home: protobuf-net.dev/
1
3
841
Marc Gravell retweeted
protobuf-net@protobuf_netAug 11
protobuf-net finally goes AOT in v3.3 with build-time generator support! protobuf-net.github.io/proto…

Compile-time serializers, for native AOT and trimming

Protocol Buffers library for idiomatic .NET

docs.protobuf-net.dev
2
1
5
1,234
LLMs in code; about 18 months ago I was firmly on the "artisanal hand-written code" side of things; I'm very much a convert now - the code was never the goal: solving tech problems was. Most of my time now goes into designing features and correctness tests, which is far more (1)
22
20
4
403
43,732
valuable than any actual code, which is ultimately transient. I'm about to merge a feature into protobuf-net that I've wanted *for years* (build-time "AOT" emit), but which I'd failed repeatedly to land, as it is tedious iterative slog; LLMs *excel* at that, especially when (2)
2
1
58
4,974
there are robust validity tests; a 60kloc feature that would never have arrived without claude (claude.com/contact-sales/cla…). You need to think of it as an intern and provide all the safety fences, but: it is alarmingly competent.

Claude for Open Source

Apply to the Claude for Open Source program. Eligible OSS maintainers and contributors get Claude Max for 6 months on us.

claude.com
7
53
4,752
I confess to feeling inordinately happy about how this shower-thought idea turned out...
Image shows a C# IDE with two methods that perform Redis transactions; one StringSetAsync with a Condition.KeyNotExists, the other a StringSetAsync with Condition.StringEqual; the condition lines have warnings in the IDE and a console "dotnet build" advising the user that these operations can be expressed as StringSet[Async](key, value, When.NotExists) and StringSet[Async](key, value, ValueCondition.Equal(expected)), respectively

ALT Image shows a C# IDE with two methods that perform Redis transactions; one StringSetAsync with a Condition.KeyNotExists, the other a StringSetAsync with Condition.StringEqual; the condition lines have warnings in the IDE and a console "dotnet build" advising the user that these operations can be expressed as StringSet[Async](key, value, When.NotExists) and StringSet[Async](key, value, ValueCondition.Equal(expected)), respectively

1
38
7,019
daysWithoutForcePushBecauseWrongBranchCommandUsed = 0;
2
1
16
2,461
Load more