Fast F#: Read-Only Struct Wrappers

Подписаться 2,2 тыс.

50% 1

There are many situations where we want to take advantage of the performance of a mutable collection to build a result, but then we want to provide a read-only wrapper of the data. The facilities built into .NET are slower than using the raw, mutable collection, so let's look at how we can implement our own read-only wrappers that match the performance of the underlying collection.
Writing a Dictionary Video: • Fast F#: Writing a Dic...
=== Contact ===
Email: hi@fastfsharp.com
Mastadon: mastodon.sdf.org/@fastfsharp
Twitter: / fastfsharp
=== Tags ===
Tags: #fsharp, #dotnet

Наука

Опубликовано:

9 апр 2023

Ссылка:

Скачать:

Готовим ссылку...

Добавить в:

Мой плейлист

Посмотреть позже

Комментарии : 12

@tw1s73d Год назад

Thanks for the video! Keep it up!

@alexk6745 8 месяцев назад

Great! I wonder if that static struct will be consuming sort of twice memory for taht type from the stack? Am I right? I have a feeling that that static type already uses on a heap pointer to the dictionary and we're adding one more. Am I right or not?

@ilyamurashov Год назад

You could hide _values with EditorBrowsableAttribute

@FastFSharp Год назад

Oh nice! I didn't know that!

@robertlenders8755 Год назад

Thanks for the great video as always. I'm curious if you've done any work with the MemoryMarshal class and the ability to "project" a Span as arbitrary struct data types?

@FastFSharp Год назад

I have. It's super dangerous and rarely necessary. I'll be going deeper and deeper into the darkness with this series so we'll get there 😊

@robertlenders8755 Год назад

@@FastFSharp I am already down the rabbit hole working with a complex C ABI with unions and variable-length structures. Looking into this technique as a way to work with stack-allocated bytes or byte arrays for interop with minimal copying.

@WerewolfAll Год назад

Why not to use :> IReadOnlyDictionary interface implemented by Dictionary?

@FastFSharp Год назад

Performance. There is significant performance degradation when you hide concrete types behind interfaces.

@jeremybobbin Год назад

Interesting that readOnlyDict is suboptimal, and that inlined-functions are subject to member accessibility. Especially from an almost 20-year-old language. What’s going on here?

@FastFSharp Год назад

The F# compiler is actually inlining that code but the CLR then tries to compile the IL and it's trying to access a private member, therefore it fails.

@bentrasmussen7631 Год назад

What about [] on the indexed property getter - do you think that would have worse performance than using F# compile-time inlining? I guess the F# inline keyword would guarantee inlining whereas the attribute would make inlining likely but not guarantee it - but guarantee no mutation because the internal state would not need to be made public.