Spanify AntiFormat function in AvTrace, decrease allocations, improve perf

[h3xds1nz]

Description

Improves performance of the AntiFormat function in AvTrace, reducing allocations and improving the overall performance up to 2 times in some cases. Also changes FormatChars from char[] to ReadOnlySpan<char>, occurrence I've missed in #9230.

Second commit features added documentation for the function. I've included some basic asserts at the Testing section.

Small strings within StringBuilder default capacity

Method	Mean [ns]	Error [ns]	StdDev [ns]	Gen0	Allocated [B]
PR__Edit	236.1 ns	4.75 ns	5.84 ns	0.0730	1224 B
Original	400.7 ns	8.04 ns	15.29 ns	0.1040	1744 B

Benchmark code

AntiFormat2("{aaaa}");
AntiFormat2("{aaaa}}");
AntiFormat2("{{aaaa}}");
AntiFormat2("{aa{{aa}}");
AntiFormat2("{aa{{aa}}ab}c}d");
AntiFormat2("aaaa}");
AntiFormat2("aaaa");
AntiFormat2("a{{{aaa");
AntiFormat2("a{{{aaa}}}");

Large strings, mostly without any replacements

Method	Mean [ns]	Error [ns]	StdDev [ns]	Gen0	Allocated [B]
PR__Edit	127.5 ns	1.29 ns	1.08 ns	0.0880	1472 B
Original	242.9 ns	4.87 ns	5.00 ns	0.1094	1832 B

Benchmark code

AntiFormat("aaaaxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx");
AntiFormat("aaaaxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx");
AntiFormat("aaaxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxa");
AntiFormat("aaaaxxxc ccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccc");
AntiFormat("aaaaabcd}yyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyy");
AntiFormat("aaaa}xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx");
AntiFormat("aaaa}6666666666666666666666666666666666666666666666666666666666666");
AntiFormat("aaaa7777777777777777777777777777777777777777777777777777777777777");
AntiFormat("aaaaggggggggggggggggggggggggggggggggggggggggggggggggggggggggggggggg");

Large strings, with replacements (up to two times the perf)

Method	Mean [ns]	Error [ns]	StdDev [ns]	Gen0	Allocated [B]
PR__Edit	394.7 ns	7.73 ns	10.05 ns	0.3271	5472 B
Original	798.9 ns	11.30 ns	10.57 ns	0.4425	7408 B

Benchmark code

AntiFormat("aaaaxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx}");
AntiFormat("aaaaxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx}");
AntiFormat("aaaxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxa}");
AntiFormat("aaaaxxxc ccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccc}");
AntiFormat("aaaaabcd}yyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyy");
AntiFormat("aaaa}xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx");
AntiFormat("aaaa}6666666666666666666666666666666666666666666666666666666666666");
AntiFormat("aaaa7777777777777777777777777777777777777777777777777777777777777{");
AntiFormat("aaaaggggggggggggggggggggggggggggggggggggggggggggggggggggggggggggggg}");
AntiFormat("aaaa77777777777777777777777777{}777777777777777777777777777777777{");
AntiFormat("aaaaggggggggggggggggggggggggggggggggggggggggggggggggg{gggggggggggggg}");

Customer Impact

Improved performance, decreased allocations.

Regression

No.

Testing

Local build, base testing between the two functions.

Assert.AreEqual(AntiFormat("{aaaa}}"), AntiFormat2("{aaaa}}"));
Assert.AreEqual(AntiFormat("{aa{{aa}}"), AntiFormat2("{aa{{aa}}"));
Assert.AreEqual(AntiFormat("{aa{{aa}}ab}c}d"), AntiFormat2("{aa{{aa}}ab}c}d"));
Assert.AreEqual(AntiFormat("aaaa}"), AntiFormat2("aaaa}"));
Assert.AreEqual(AntiFormat("aaaa"), AntiFormat2("aaaa"));
Assert.AreEqual(AntiFormat("a{{{aaa"), AntiFormat2("a{{{aaa"));
Assert.AreEqual(AntiFormat("a{{{aaa}}}"), AntiFormat2("a{{{aaa}}}"));
Assert.AreEqual(AntiFormat("{aaaa}"), AntiFormat2("{aaaa}"));
Assert.AreEqual(AntiFormat("{{aaaa}}"), AntiFormat2("{{aaaa}}"));
Assert.AreEqual(AntiFormat("{{{}aaaa}}"), AntiFormat2("{{{}aaaa}}"));
Assert.AreEqual(AntiFormat(string.Empty), AntiFormat2(string.Empty));
Assert.AreEqual(AntiFormat("a{{{}aaaa}}"), AntiFormat2("a{{{}aaaa}}"));
Assert.AreEqual(AntiFormat("a}aaaa}"), AntiFormat2("a}aaaa}"));
Assert.AreEqual(AntiFormat("}aaaa}"), AntiFormat2("}aaaa}"));

Risk

Low.

Microsoft Reviewers: Open in CodeFlow