refactor(Noder): optimize line splitting using ReadOnlySpan

[Lemon73-Computing]

Summary

Improved the processes of Noder.

Details

In Noder processing, replacing operations on string and string[] with ReadOnlySpan and List<string> has resulted in faster and more lightweight processing.

See also: https://github.com/AliceNovel/CommonNovel/pull/34/changes#diff-51e1cf45389002daed215ce0327155a875852fddba52326385b7afa56daabe09R10-R47

Breaking Changes

Caution

This PR contains breaking changes.

• (Change in interpretation) Changed the handling of line break codes from being environment-dependent to being character-dependent (Not specified in the specifications)
  • (This need to be specified in the specifications in the future) → [Spec] Specify how line break codes are handled #36
  • Before:
    • On Linux \n (LF) or \r\n (CRLF) → \n (LF)
    • On Windows \n (LF) or \r\n (CRLF) → \r\n (CRLF)
  • After:
    • Every where \n → \n, \r\n → \r\n
  • See also:
    • [Spec] Specify how line break codes are handled #36
  • Diff:
    • ba9e898#diff-8d94b27edaf5e79eac69939e49f85abd96cdb22d75754272178a1f32a391231eR26-R41
• Adjustments have been made so that nodes are created for all consecutive two-line sections where multiple line breaks occur.
  • (The previous implementation was incorrect and has been corrected to match the specifications.)
  • (In the previous implementation, even if lines were consecutive, no nodes were generated at all if the content was empty.)
  • Before:
    • \n\n\n → [] (no elements)
  • After:
    • \n\n\n → [ "", "" ] (two elements)
  • See also:

    • CommonNovel/docs/v0.1.x/v0.1.0.md

      Line 51 in 627580e

      A node indicates multiple lines from a blank line (a line containing only spaces, tabs, or line feed codes) to the next blank line.

    • CommonNovel/docs/v0.1.x/v0.1.0.md

      Lines 68 to 79 in 627580e

      	### Example 2
      	```sh
      	- Alice
      	[Hi!]
      	- Rabbit
      	[Hi, there.]
      	```
      	This example is similar to [Example 1](#example-1), however there are two blank lines between `[Hi!]` and `- Rabbit`. So, the first node includes the two lines; `- Alice` and `[Hi!]`. But the next node does not include any line. (Treat as blank ` ` internally) Then, the last node includes the two lines; `- Rabbit` and `[Hi, there.]`.

  • Diff:
    • ba9e898#diff-8d94b27edaf5e79eac69939e49f85abd96cdb22d75754272178a1f32a391231eR42-R46

Benchmark

BenchmarkDotNet v0.15.8, Windows 11 (10.0.26200.8328/25H2/2025Update/HudsonValley2)
13th Gen Intel Core i7-13700H 2.40GHz, 1 CPU, 20 logical and 14 physical cores
.NET SDK 10.0.103
  [Host]     : .NET 10.0.3 (10.0.3, 10.0.326.7603), X64 RyuJIT x86-64-v3
  DefaultJob : .NET 10.0.3 (10.0.3, 10.0.326.7603), X64 RyuJIT x86-64-v3

Before

Method	Mean	Error	StdDev	Gen0	Allocated
RunNoder	104.307 ns	1.9910 ns	3.8361 ns	0.0497	624 B

After

Method	Mean	Error	StdDev	Gen0	Allocated
RunNoder	51.757 ns	1.0359 ns	1.1929 ns	0.0197	248 B

[github-actions]

Code Coverage

Result: success

Show coverage summary

Summary

Generated on:	05/09/2026 - 07:46:31
Coverage date:	05/09/2026 - 07:46:28
Parser:	Cobertura
Assemblies:	1
Classes:	1
Files:	3
Line coverage:	100% (73 of 73)
Covered lines:	73
Uncovered lines:	0
Coverable lines:	73
Total lines:	145
Branch coverage:	100% (28 of 28)
Covered branches:	28
Total branches:	28
Method coverage:	Feature is only available for sponsors

Risk Hotspots

Assembly	Class	Method	Crap Score	Cyclomatic complexity
CommonNovel	CommonNovel.Compiler	Noder(...)	16	16

Coverage

Name	Covered	Uncovered	Coverable	Total	Line coverage	Covered	Total	Branch coverage
CommonNovel	73	0	73	145	100%	28	28	100%
CommonNovel.Compiler	73	0	73	145	100%	28	28	100%

[Lemon73-Computing]

/gemini summary

[gemini-code-assist]

Summary of Changes

This pull request refactors the Noder component to optimize performance by leveraging ReadOnlySpan for string manipulation. The changes include a shift in how line breaks are interpreted and processed, resulting in more predictable and efficient node generation. While this introduces breaking changes regarding how empty lines and line breaks are handled, it significantly reduces memory overhead and improves processing speed as demonstrated by the provided benchmarks.

Highlights

• Performance Optimization: Replaced string-based processing with ReadOnlySpan and List to reduce memory allocations and improve execution speed.
• Line Splitting Logic: Updated the Noder logic to handle line breaks based on character sequences rather than environment-dependent settings, ensuring consistent behavior across platforms.
• Breaking Changes: Adjusted node generation to correctly handle consecutive line breaks, which now produce empty string elements instead of being ignored.

Activity

• The pull request was created by Lemon73-Computing.
• A code coverage report was generated by the CI bot, showing 100% coverage.
• Automated review comments were posted highlighting potential functional regressions and concerns regarding empty node handling.
• The author requested a summary of the pull request.