Use ReadOnlySpan<char> to replace the 'new string(...)' to reduce memory allocation

[xuzhg]

Use ExpressionLexer to do ReadOnlySpan investigation and seek feebacks to move forward to all other ODL parts.

Be noted,

1. the ReadOnlySpan doesn't override the operator == and != for string. Remember to use 'Equals(...)' to replace them.

ReadOnlySpan<char> s1 = "nav/$ref".AsSpan().Slice(4,4);

Console.WriteLine(s1.ToString());
Console.WriteLine(s1 == "$ref");
Console.WriteLine(s1 != "$ref");
Console.WriteLine(s1.Equals("$ref", StringComparison.Ordinal));

It outputs

$ref
False
True
True

2. Anytime to call 'ToString()' on ReadOnlySpan will do the 'new string(...)', so be noted, not to call it multiple times until you need the string and remember to cache it.

Benchmark:

See the allocated, using 'ReadOnlySpan' is 1/3 memory allocation comparing to the 'new string(...)'.

Codes

   int NumberOfItems = 100000;
   public string expression = "abc(2014-09-19T12:13:14+00:00)/3258.678765765489753678965390/SRID=1234(POINT(10 20))/Function(foo=@x,bar=1,baz=@y)";

    [Benchmark]
    public void ExpressionLexerUsingNewStringSingleRun()
    {
        ExpressionLexer2 lexer = new Microsoft.OData.UriParser.ExpressionLexer2(expression, true, false);
        while (lexer.CurrentToken.Kind != ExpressionTokenKind.End)
        {
            lexer.NextToken();
        }
    }

    [Benchmark]
    public void ExpressionLexerUsingReadOnlySingleRun()
    {

        ExpressionLexer lexer = new Microsoft.OData.UriParser.ExpressionLexer(expression, true, false);
        while (lexer.CurrentToken.Kind != ExpressionTokenKind.End)
        {
            lexer.NextToken();
        }
    }

    [Benchmark]
    public void ExpressionLexerUsingNewString()
    {
        for (int i = 0; i < NumberOfItems; i++)
        {
            ExpressionLexer2 lexer = new Microsoft.OData.UriParser.ExpressionLexer2(expression, true, false);
            while (lexer.CurrentToken.Kind != ExpressionTokenKind.End)
            {
                lexer.NextToken();
            }
        }
    }

    [Benchmark]
    public void ExpressionLexerUsingReadOnly()
    {
        for (int i = 0; i < NumberOfItems; i++)
        {
            ExpressionLexer lexer = new Microsoft.OData.UriParser.ExpressionLexer(expression, true, false);
            while (lexer.CurrentToken.Kind != ExpressionTokenKind.End)
            {
                lexer.NextToken();
            }
        }
    }

Issues

This pull request fixes #xxx.

Description

Briefly describe the changes of this pull request.

Checklist (Uncheck if it is not completed)

• Test cases added
• Build and test with one-click build and test script passed

Additional work necessary

If documentation update is needed, please add "Docs Needed" label to the issue and provide details about the required document change in the issue.

[habbes]

@xuzhg could you share the benchmarks that you used for the perf tests, did you also try to run the existing UriParser benchmarks before and after these changes?

[xuzhg]

@xuzhg could you share the benchmarks that you used for the perf tests, did you also try to run the existing UriParser benchmarks before and after these changes?

I create a branch to hold the perf test at: https://github.com/OData/odata.net/tree/ReadOnlySpanBenchMark

This PR is just a starting point. We will move to UriParser and others in the comming step.

[xuzhg]

@habbes Resolved the comment. Please take a look again. I will keep moving to next step. It could be 'UriParser'.