[API Proposal]: `ReadOnlySpan<char>` overloads for `StringNormalizationExtensions`

[udaken]

Edited by @tarekgh

Background and motivation

If we try to verify that the string represented by ReadOnlySpan<char> is Unicode normalized, we currently have to turn it into a string once.
Overloading for ReadOnlySpan<char> is needed to reduce allocation.

API Proposal

namespace System;

// Doesn't throw on the operation even when having wrong character sequence. 

public static class StringNormalizationExtensions {
+    public static bool IsNormalized(this ReadOnlySpan<char> source, NormalizationForm normalizationForm = NormalizationForm.FormC); // will return false in invalid sequence
+    public static OperationStatus TryNormalize(this ReadOnlySpan<char> source, Span<char> destination, out int charsWritten, NormalizationForm normalizationForm = NormalizationForm.FormC);
}

Original Proposal - Alternative Proposal

namespace System;

// throws on wrong sequence as String.Normalize does. 

public static class StringNormalizationExtensions {
+    public static bool IsNormalized(this ReadOnlySpan<char> strInput, NormalizationForm normalizationForm = NormalizationForm.FormC);
+    public static bool TryNormalize(this ReadOnlySpan<char> strInput, Span<char> destination, out int charsWritten, NormalizationForm normalizationForm = NormalizationForm.FormC);
}

API Usage

ReadOnlySpan<char> partOfhugeArray =  hugeArray.AsSpan(start, length);
if (!partOfhugeArray.IsNormalized())
    throw new Exception("Text must be Unicode normalized.");
/// ...

No response

Risks

No response

Tagging subscribers to this area: @dotnet/area-system-globalization
See info in area-owners.md if you want to be subscribed.

Issue Details

---

Background and motivation

If we try to verify that the string represented by ReadOnlySpan<char> is Unicode normalized, we currently have to turn it into a string once.
Overloading for ReadOnlySpan<char> is needed to reduce allocation.

API Proposal

namespace System;

public static class StringNormalizationExtensions {
+    public static bool IsNormalized(this ReadOnlySpan<char> strInput, NormalizationForm normalizationForm = NormalizationForm.FormC);
+    public static bool TryNormalize(this ReadOnlySpan<char> strInput, Span<char> destination, out int charsWritten, NormalizationForm normalizationForm = NormalizationForm.FormC);
}

API Usage

ReadOnlySpan<char> partOfhugeArray =  hugeArray.AsSpan(start, length);
if (!partOfhugeArray.IsNormalized())
    throw new Exception("Text must be Unicode normalized.");
/// ...

Alternative Designs

No response

Risks

No response

Author:	udaken
Assignees:	-
Labels:	api-suggestion, area-System.Globalization, untriaged
Milestone:	-

[tarekgh]

@udaken I made a few edits to the proposal and hope that works for you.

[bartonjs]

Video

• TryNormalize as proposed mixes semantics of OperationStatus and the Try pattern, and needs to pick one or the other.
  • We went with Try
• We added GetNormalizedLength for callers who are concerned about too many double-and-call-again for the Try-Write.

namespace System;

public static partial class StringNormalizationExtensions
{
    public static bool IsNormalized(this ReadOnlySpan<char> source, NormalizationForm normalizationForm = NormalizationForm.FormC);
    public static bool TryNormalize(this ReadOnlySpan<char> source, Span<char> destination, out int charsWritten, NormalizationForm normalizationForm = NormalizationForm.FormC);
    public static int GetNormalizedLength(this ReadOnlySpan<char> source, NormalizationForm normalizationForm = NormalizationForm.FormC);
}