Minimize System.Version formatting buffer's length

[Chasmical]

Change the length of the char buffer used in Version.ToString() from 47 to 43 chars (4 non-negative Int32s + 3 periods). 2147483647.2147483647.2147483647.2147483647. Even if a version has invalid internal state (any component less than -1), the TryFormat method just casts them to UInt32 anyway.

[Chasmical]

Oh, I didn't realize that Number.UInt32NumberBufferLength is the same as Int32NumberBufferLength - 11, for some reason.

[EgorBo]

To be fair, this doesn't reduce the actual stack allocation size - JIT is going to round up both to % 16 anyway, e.g. see sharplab

[Chasmical]

I see. 😄 Then, I guess it could still improve readability of the decompiled code. It was confusing to see 47 characters allocated when the actual maximum length is 43.

On another note, looks like Number.Int32NumberBufferLength isn't used anywhere else, besides the Number class and here, and the extra 1 to length is actually for the terminating nulls. Would make more sense not to use it here, in case Number.NumberBuffer changes.

runtime/src/libraries/Common/src/System/Number.NumberBuffer.cs

Lines 13 to 16 in 153a94b

	// We need 1 additional byte, per length, for the terminating null
	internalconstintDecimalNumberBufferLength=29+1+1;// 29 for the longest input + 1 for rounding
	internalconstintDoubleNumberBufferLength=767+1+1;// 767 for the longest input + 1 for rounding: 4.9406564584124654E-324
	internalconstintInt32NumberBufferLength=10+1;// 10 for the longest input: 2,147,483,647

[stephentoub]

Thank you, but I'd rather keep it how it is today. I think it's more maintainable as it is currently and this doesn't actually improve perf.

[Chasmical]

Change the length of the char buffer used in Version.ToString() from 47 to 43 chars (4 non-negative Int32s + 3 periods). 2147483647.2147483647.2147483647.2147483647. Even if a version has invalid internal state (any component less than -1), the TryFormat method just casts them to UInt32 anyway.

[Chasmical]

Oh, I didn't realize that Number.UInt32NumberBufferLength is the same as Int32NumberBufferLength - 11, for some reason.

[EgorBo]

To be fair, this doesn't reduce the actual stack allocation size - JIT is going to round up both to % 16 anyway, e.g. see sharplab

[Chasmical]

I see. 😄 Then, I guess it could still improve readability of the decompiled code. It was confusing to see 47 characters allocated when the actual maximum length is 43.

On another note, looks like Number.Int32NumberBufferLength isn't used anywhere else, besides the Number class and here, and the extra 1 to length is actually for the terminating nulls. Would make more sense not to use it here, in case Number.NumberBuffer changes.

runtime/src/libraries/Common/src/System/Number.NumberBuffer.cs

Lines 13 to 16 in 153a94b

	// We need 1 additional byte, per length, for the terminating null
	internalconstintDecimalNumberBufferLength=29+1+1;// 29 for the longest input + 1 for rounding
	internalconstintDoubleNumberBufferLength=767+1+1;// 767 for the longest input + 1 for rounding: 4.9406564584124654E-324
	internalconstintInt32NumberBufferLength=10+1;// 10 for the longest input: 2,147,483,647

[stephentoub]

Thank you, but I'd rather keep it how it is today. I think it's more maintainable as it is currently and this doesn't actually improve perf.

[Chasmical]

Change the length of the char buffer used in Version.ToString() from 47 to 43 chars (4 non-negative Int32s + 3 periods). 2147483647.2147483647.2147483647.2147483647. Even if a version has invalid internal state (any component less than -1), the TryFormat method just casts them to UInt32 anyway.

[Chasmical]

Oh, I didn't realize that Number.UInt32NumberBufferLength is the same as Int32NumberBufferLength - 11, for some reason.

[EgorBo]

To be fair, this doesn't reduce the actual stack allocation size - JIT is going to round up both to % 16 anyway, e.g. see sharplab

[Chasmical]

I see. 😄 Then, I guess it could still improve readability of the decompiled code. It was confusing to see 47 characters allocated when the actual maximum length is 43.

On another note, looks like Number.Int32NumberBufferLength isn't used anywhere else, besides the Number class and here, and the extra 1 to length is actually for the terminating nulls. Would make more sense not to use it here, in case Number.NumberBuffer changes.

runtime/src/libraries/Common/src/System/Number.NumberBuffer.cs

Lines 13 to 16 in 153a94b

	// We need 1 additional byte, per length, for the terminating null
	internalconstintDecimalNumberBufferLength=29+1+1;// 29 for the longest input + 1 for rounding
	internalconstintDoubleNumberBufferLength=767+1+1;// 767 for the longest input + 1 for rounding: 4.9406564584124654E-324
	internalconstintInt32NumberBufferLength=10+1;// 10 for the longest input: 2,147,483,647

[stephentoub]

Thank you, but I'd rather keep it how it is today. I think it's more maintainable as it is currently and this doesn't actually improve perf.

[Chasmical]

Change the length of the char buffer used in Version.ToString() from 47 to 43 chars (4 non-negative Int32s + 3 periods). 2147483647.2147483647.2147483647.2147483647. Even if a version has invalid internal state (any component less than -1), the TryFormat method just casts them to UInt32 anyway.

[Chasmical]

Oh, I didn't realize that Number.UInt32NumberBufferLength is the same as Int32NumberBufferLength - 11, for some reason.

[EgorBo]

To be fair, this doesn't reduce the actual stack allocation size - JIT is going to round up both to % 16 anyway, e.g. see sharplab

[Chasmical]

I see. 😄 Then, I guess it could still improve readability of the decompiled code. It was confusing to see 47 characters allocated when the actual maximum length is 43.

On another note, looks like Number.Int32NumberBufferLength isn't used anywhere else, besides the Number class and here, and the extra 1 to length is actually for the terminating nulls. Would make more sense not to use it here, in case Number.NumberBuffer changes.

runtime/src/libraries/Common/src/System/Number.NumberBuffer.cs

Lines 13 to 16 in 153a94b

	// We need 1 additional byte, per length, for the terminating null
	internalconstintDecimalNumberBufferLength=29+1+1;// 29 for the longest input + 1 for rounding
	internalconstintDoubleNumberBufferLength=767+1+1;// 767 for the longest input + 1 for rounding: 4.9406564584124654E-324
	internalconstintInt32NumberBufferLength=10+1;// 10 for the longest input: 2,147,483,647

[stephentoub]

Thank you, but I'd rather keep it how it is today. I think it's more maintainable as it is currently and this doesn't actually improve perf.

[Chasmical]

Change the length of the char buffer used in Version.ToString() from 47 to 43 chars (4 non-negative Int32s + 3 periods). 2147483647.2147483647.2147483647.2147483647. Even if a version has invalid internal state (any component less than -1), the TryFormat method just casts them to UInt32 anyway.

[Chasmical]

Oh, I didn't realize that Number.UInt32NumberBufferLength is the same as Int32NumberBufferLength - 11, for some reason.

[EgorBo]

To be fair, this doesn't reduce the actual stack allocation size - JIT is going to round up both to % 16 anyway, e.g. see sharplab

[Chasmical]

I see. 😄 Then, I guess it could still improve readability of the decompiled code. It was confusing to see 47 characters allocated when the actual maximum length is 43.

On another note, looks like Number.Int32NumberBufferLength isn't used anywhere else, besides the Number class and here, and the extra 1 to length is actually for the terminating nulls. Would make more sense not to use it here, in case Number.NumberBuffer changes.

runtime/src/libraries/Common/src/System/Number.NumberBuffer.cs

Lines 13 to 16 in 153a94b

	// We need 1 additional byte, per length, for the terminating null
	internalconstintDecimalNumberBufferLength=29+1+1;// 29 for the longest input + 1 for rounding
	internalconstintDoubleNumberBufferLength=767+1+1;// 767 for the longest input + 1 for rounding: 4.9406564584124654E-324
	internalconstintInt32NumberBufferLength=10+1;// 10 for the longest input: 2,147,483,647

[stephentoub]

Thank you, but I'd rather keep it how it is today. I think it's more maintainable as it is currently and this doesn't actually improve perf.

[Chasmical]

Change the length of the char buffer used in Version.ToString() from 47 to 43 chars (4 non-negative Int32s + 3 periods). 2147483647.2147483647.2147483647.2147483647. Even if a version has invalid internal state (any component less than -1), the TryFormat method just casts them to UInt32 anyway.

[Chasmical]

Oh, I didn't realize that Number.UInt32NumberBufferLength is the same as Int32NumberBufferLength - 11, for some reason.

[EgorBo]

To be fair, this doesn't reduce the actual stack allocation size - JIT is going to round up both to % 16 anyway, e.g. see sharplab

[Chasmical]

I see. 😄 Then, I guess it could still improve readability of the decompiled code. It was confusing to see 47 characters allocated when the actual maximum length is 43.

On another note, looks like Number.Int32NumberBufferLength isn't used anywhere else, besides the Number class and here, and the extra 1 to length is actually for the terminating nulls. Would make more sense not to use it here, in case Number.NumberBuffer changes.

runtime/src/libraries/Common/src/System/Number.NumberBuffer.cs

Lines 13 to 16 in 153a94b

	// We need 1 additional byte, per length, for the terminating null
	internalconstintDecimalNumberBufferLength=29+1+1;// 29 for the longest input + 1 for rounding
	internalconstintDoubleNumberBufferLength=767+1+1;// 767 for the longest input + 1 for rounding: 4.9406564584124654E-324
	internalconstintInt32NumberBufferLength=10+1;// 10 for the longest input: 2,147,483,647

[stephentoub]

Thank you, but I'd rather keep it how it is today. I think it's more maintainable as it is currently and this doesn't actually improve perf.

[Chasmical]

Change the length of the char buffer used in Version.ToString() from 47 to 43 chars (4 non-negative Int32s + 3 periods). 2147483647.2147483647.2147483647.2147483647. Even if a version has invalid internal state (any component less than -1), the TryFormat method just casts them to UInt32 anyway.

[Chasmical]

Oh, I didn't realize that Number.UInt32NumberBufferLength is the same as Int32NumberBufferLength - 11, for some reason.

[EgorBo]

To be fair, this doesn't reduce the actual stack allocation size - JIT is going to round up both to % 16 anyway, e.g. see sharplab

[Chasmical]

I see. 😄 Then, I guess it could still improve readability of the decompiled code. It was confusing to see 47 characters allocated when the actual maximum length is 43.

On another note, looks like Number.Int32NumberBufferLength isn't used anywhere else, besides the Number class and here, and the extra 1 to length is actually for the terminating nulls. Would make more sense not to use it here, in case Number.NumberBuffer changes.

runtime/src/libraries/Common/src/System/Number.NumberBuffer.cs

Lines 13 to 16 in 153a94b

	// We need 1 additional byte, per length, for the terminating null
	internalconstintDecimalNumberBufferLength=29+1+1;// 29 for the longest input + 1 for rounding
	internalconstintDoubleNumberBufferLength=767+1+1;// 767 for the longest input + 1 for rounding: 4.9406564584124654E-324
	internalconstintInt32NumberBufferLength=10+1;// 10 for the longest input: 2,147,483,647

[stephentoub]

Thank you, but I'd rather keep it how it is today. I think it's more maintainable as it is currently and this doesn't actually improve perf.

[Chasmical]

Change the length of the char buffer used in Version.ToString() from 47 to 43 chars (4 non-negative Int32s + 3 periods). 2147483647.2147483647.2147483647.2147483647. Even if a version has invalid internal state (any component less than -1), the TryFormat method just casts them to UInt32 anyway.

[Chasmical]

Oh, I didn't realize that Number.UInt32NumberBufferLength is the same as Int32NumberBufferLength - 11, for some reason.

[EgorBo]

To be fair, this doesn't reduce the actual stack allocation size - JIT is going to round up both to % 16 anyway, e.g. see sharplab

[Chasmical]

I see. 😄 Then, I guess it could still improve readability of the decompiled code. It was confusing to see 47 characters allocated when the actual maximum length is 43.

On another note, looks like Number.Int32NumberBufferLength isn't used anywhere else, besides the Number class and here, and the extra 1 to length is actually for the terminating nulls. Would make more sense not to use it here, in case Number.NumberBuffer changes.

runtime/src/libraries/Common/src/System/Number.NumberBuffer.cs

Lines 13 to 16 in 153a94b

	// We need 1 additional byte, per length, for the terminating null
	internalconstintDecimalNumberBufferLength=29+1+1;// 29 for the longest input + 1 for rounding
	internalconstintDoubleNumberBufferLength=767+1+1;// 767 for the longest input + 1 for rounding: 4.9406564584124654E-324
	internalconstintInt32NumberBufferLength=10+1;// 10 for the longest input: 2,147,483,647

[stephentoub]

Thank you, but I'd rather keep it how it is today. I think it's more maintainable as it is currently and this doesn't actually improve perf.