Optimize InsertionContext.InsertText with SearchValues and 2-phase algo for tab expansion and newline normalization

[udlose]

Note

I've attached the perf benchmark I used with the results. I found the optimal stackalloc size for the char[] passed to ValueStringBuilder to be 512 bytes. Benchmark results listed at end.

This pull request introduces several improvements and code quality updates across the codebase, focusing on performance, memory efficiency, code clarity, and modern C# idioms. The most significant change is a complete rewrite of the InsertionContext.InsertText method to use a high-performance, allocation-minimizing algorithm for snippet insertion. Additionally, the code now consistently uses GetValueOrDefault for dictionary lookups, improves resource management, and adds code analysis suppressions for clarity.

Performance and Memory Efficiency:

• InsertionContext.InsertText has been rewritten to use a two-phase, SIMD-accelerated algorithm for tab expansion and newline normalization, leveraging ValueStringBuilder and SearchValues<char> to minimize allocations and improve speed. This also ensures correct handling of custom tab strings and CRLF boundaries.

Modern C# Idioms and Code Quality:

• Dictionary lookups throughout the codebase now use GetValueOrDefault instead of TryGetValue(..., out ...) ? ... : null for improved clarity and conciseness. This applies to HighlightingManager, XmlHighlightingDefinition, and InsertionContext. [1] [2] [3] [4]
• Default argument assignments have been updated to use the null-coalescing assignment operator (??=) for brevity and clarity in InsertionContext. [1] [2]

Resource Management and Reliability:

• DataTransfer objects are now disposed using using statements in EditingCommandHandler to ensure proper resource cleanup. [1] [2]
• Code analysis suppressions have been added to clarify ownership and disposal responsibilities for disposable objects in several classes, including TextMate, V2Loader, and TextEditor. [1] [2] [3]

Compatibility and Documentation:

• Explicit use of Enumerable.Reverse is now enforced to avoid breaking changes with newer C# versions, with comments referencing the relevant compiler change.
• Extensive documentation has been added to InsertionContext.InsertText explaining the new algorithm, its rationale, and its benefits.

Miscellaneous:

• Minor formatting and copyright header fixes.
• .runsettings has been added to the solution items for improved test configuration management.

Benchmark Results

Area	Result
Memory Allocations	Varied between parity and 3.62x less memory with 20 of 30 cases decreasing memory from 1.58x to 3.62x
Fast paths (Plain, TabsHeavy)	Roughly 15x-21x faster than Original, depending on text length and tab mode.
Newline-heavy paths (LfHeavy, CrLfHeavy)	Usually allocate about 2.0x-2.8x less memory than Original, but are often near parity or slower in runtime.
Mixed workloads	Generally close to baseline, with modest wins in some Spaces4 cases and slight regressions in some KeepTabs cases.
Overall average across 30 cases	Average allocated bytes drop from about 75.7 KB in Original to about 41.4 KB in both optimized variants, while average mean time stays in the same broad range and hides the split between huge fast-path wins and slower newline-heavy cases.

Benchmark File

inserttext_benchmark_comparison_uncapped_reference_style.html

InsertTextBenchmark.linq.txt

Test Coverage