c_e4fb43f047912dd6c2

Commit A introduces substantial new functionality and parser behavior across multiple modules. It adds deterministic typed block masking, prose tokenization for item references, URL and external reference handling, stricter DSL parsing rules for braced bodies versus code fences, improved linkification, breadcrumb fixes, resolver updates, and extensive automated tests covering the new behavior. These changes affect core parsing, rendering, and external item support. Commit B is primarily a UI/layout cleanup with some CSS improvements for ranking lists and removal of an extra wrapper element in the compare page. While useful, it is much smaller in scope and impact than the parser and linkification work in Commit A.

Side A introduces substantial new functionality: deterministic and typed block masking, a full prose tokenizer for item references, improved URL parsing (including punctuation trimming and newline handling), stricter DSL validation rules, and integration into HTML linkification. It also adds extensive tests and refactors multiple components to support these features. Side B is comparatively minor, mainly removing a wrapper element in a template and adding CSS refinements. The scope, complexity, and impact of A are significantly greater.

Side A introduces substantial functional changes across the DSL parser, block masking system, prose tokenization, linkification logic, external resolver behavior, and adds extensive new tests. It improves determinism, enforces stricter parsing rules, enhances URL/item reference handling, and refactors HTML link generation. Side B mainly removes a wrapper shell in one template and adjusts CSS styling for ranking lists, which is comparatively minor and largely presentational. Therefore, Side A contributed significantly more.

Commit A makes broader changes across the application by simplifying the vote-compare page structure and updating ontology ranking list styling in multiple themes, producing visible UI and layout improvements. Commit B is a useful maintenance change that replaces manually enumerated Kaocha test suites with automatic test discovery, reducing future configuration work and preventing missed tests, but it is limited to the project's testing infrastructure. Both are valuable, though A has the wider functional and user-facing impact.

Commit B meaningfully improves the project’s testing infrastructure by eliminating manual test enumeration and enabling automatic discovery of all test namespaces. This reduces maintenance overhead, prevents future omissions, and ensures CI and local runs stay consistent as new tests are added. Commit A is mostly a UI refactor and styling enhancement with limited functional impact, whereas B improves long-term reliability and developer workflow.

Side A makes broader, user-visible changes across multiple files, restructuring HTML layout and significantly updating ranking list styling in several themes. It impacts UI structure and presentation site-wide. Side B improves test infrastructure and maintainability, but is confined to a single configuration file and is smaller in scope. Overall, Side A represents a larger functional and visual contribution.

Side A makes substantive parser and rendering improvements: it introduces deterministic typed block masking, a prose item-reference tokenizer that correctly handles raw URLs, punctuation, newlines, and code fences, enforces braced DSL item bodies instead of ambiguous fenced bodies, fixes external breadcrumb root detection, and updates linkification plus extensive tests. Side B is primarily a UI/layout cleanup and CSS styling change (removing a wrapper element and improving ranking-number presentation), with little impact on core functionality or long-term correctness.

A introduces substantive parsing and correctness improvements: deterministic block tokens replacing RNG, typed block tracking, a new prose tokenizer that avoids code fences and trims punctuation, and stricter validation requiring braced item bodies—plus integrating this into linkification. B mainly removes a wrapper div and tweaks CSS/layout, with no comparable impact on core logic or correctness.

Side A introduces a typed, deterministic BlockMasker with BlockKind tracking, adds a prose tokenizer that avoids linkifying inside code fences, enforces braced DSL item bodies with explicit parse errors, and rewrites HTML linkification to use the tokenizer—backed by extensive new tests. Side B mainly removes a wrapping section in vote_compare and tweaks CSS for ranking lists, which is largely presentational and far less architecturally significant.

Side A implements a substantive, tested feature: a typed block-masking system, deterministic tokenization, and a proper prose tokenizer that correctly linkifies item refs while respecting code fences and line boundaries, backed by many new unit tests and bugfixes (e.g., requiring braced item bodies). Side B is a mix of minor CSS tweaks (ranking-list marker styling) and a cosmetic removal of a wrapping HTML section/indentation change with no functional test coverage or clear lasting design improvement.

A adds lasting core behavior: deterministic typed block masking, prose item-ref tokenization (tilde/dash/raw URLs, newline/punctuation boundaries, code-fence exclusion), braced body enforcement, linkify/resolver updates, and tests. B only unwraps a vote-compare shell div and tweaks ranking-list number CSS—cosmetic UI polish with little structural impact.

Side B replaces two manually maintained Kaocha test suites with a single auto-discovered `:all` suite using the `^test\..+` namespace pattern, so new test namespaces run automatically without editing `tests.edn`, reducing ongoing maintenance and preventing missed tests. Side A mostly removes a wrapping `section` from the vote-compare HTML and adjusts ranking-list CSS styling across themes, which is primarily structural and visual rather than a lasting functional improvement.

Side B replaces two manually enumerated Kaocha suites with a single auto-discovered suite using `:ns-patterns ["^test\\..+"]`, ensuring new tests run without config changes and preventing silent omissions. Side A mainly removes a wrapping `section.vote-compare-shell` and adds CSS styling for ranking lists, which is largely presentational and less impactful on long-term correctness or maintenance.

Side B replaces two manually enumerated Kaocha suites with a single auto-discovering `:all` suite using `^test\..+`, ensuring new `test.*` namespaces run automatically without editing `tests.edn`, which improves long-term test reliability and maintenance. Side A mainly removes a wrapping `section` in `garden.rs` and tweaks CSS for ranking list styling (font sizes, counters, layout), which are largely presentational changes without comparable impact on project correctness or workflow.

Side B makes a small, clear infrastructure fix that removes a maintenance trap (manually enumerating test namespaces), ensuring new test files are auto-discovered — a lasting, low-risk improvement to CI reliability. Side A is mostly cosmetic: removing an unused wrapper section (dead HTML markup) and CSS tweaks to ranking-list bullet styling, which is minor polish rather than substantive functionality or bug fixing.

B replaces brittle hand-enumerated kaocha suites with a single auto-discovered ^test\..+ suite so new tests run without editing tests.edn—a lasting maintainability fix used by CI and TEST.sh. A mainly drops a vote-compare-shell wrapper and retouches ranking-list number CSS across themes, which is minor presentational cleanup with less durable project impact.

Side A is entirely a planning markdown file plus a thin unused RouteContext wrapper that duplicates ThreadNav without any call sites migrated to it, contributing no functional change yet. Side B makes a concrete UI fix (removing an extraneous wrapper section/shell) and real, visible CSS improvements to rank markers across three themes, delivering tangible, lasting user-facing value.

A adds a real architectural seam (RouteContext wrapping ThreadNav with item_href/thread_url APIs) plus a concrete ItemId migration plan that steers lasting identity/routing work. B only unwraps a vote-compare-shell div and restyles ranking-list counters in CSS—cosmetic HTML/theme churn with little structural impact.

Side A introduces a new `RouteContext` abstraction in `server/src/html/routing.rs`, re-exports it, and documents a concrete migration path toward centralized, scope-aware URL generation, creating an architectural seam that future code can build on. Side B is primarily a UI refactor that removes a wrapper element in the vote-compare template and adjusts CSS styling for ranked lists, improving presentation but not changing core behavior or architecture.

Side A removes an unnecessary wrapper element/class, cleans up CSS duplication, and adds consistent, reusable ranking-list numbering styles across three theme files, representing real (if modest) UI/CSS improvements. Side B is a trivial one-line change swapping a span for an anchor tag with href='#', which is a minor/dubious fix (a disabled link pointing to '#' is arguably worse semantically than a span) with negligible lasting value.

A removes an unnecessary vote-compare-shell wrapper and lands real, cross-theme ranking-list typography (counters, flex, tabular nums) in default/retro/retro_craft CSS. B only swaps one disabled paginator control from span to a href="#", a tiny consistency tweak with little lasting impact.

Side A makes broader UI improvements: it removes an unnecessary wrapper around the vote-compare content and updates ontology ranking list styling across multiple themes by adding consistent numbered-list counters, typography, and alignment (`::before`, `counter-increment`, `font-variant-numeric`, etc.). Side B only changes a disabled paginator control from a `<span>` to an `<a href="#">`, a small markup adjustment with much narrower long-term impact.

Side B fixes an actual dead test (missing #[test] attribute so it silently never ran), removes clippy lints across multiple files (idiomatic Option handling, redundant deref, unused var, format-string improvements), and adds a documented dead_code allowance for shared test helpers — all real correctness/maintainability improvements. Side A is purely CSS/markup tweaks (removing a wrapper section, styling rank numbers) that are cosmetic and localized, with no functional or structural benefit beyond visual polish.

A’s lasting impact is the ranking-list presentation work (shared counter styling and filling in missing ont-ranking-list rules in retro/retro_craft) plus a real DOM simplification by dropping the unused vote-compare-shell wrapper. B is mostly mechanical clippy hygiene (needless refs, match→let, format tweaks, a type alias) with one genuine fix—annotating an unwired #[test]—so it ranks slightly behind.

Side B includes a functional improvement by adding a missing #[test] so the `set_new_thread_compose_expanded_true` test is actually executed, alongside targeted code-quality fixes such as replacing a complex return type with a type alias and modernizing APIs (`is_some_and`, pattern destructuring). Side A mostly removes a wrapper `section` around the vote-compare markup and adjusts CSS for ranking list presentation, which are primarily UI/layout changes with less lasting impact on correctness or maintainability.

Side B fixes a real off-by-one/indexing bug (thread_post_index semantics), removing a fallback that masked an invariant violation with .expect() and documenting the 0-based vs 1-based confusion, plus adds regression tests asserting correct index values. Side A is purely cosmetic: dedenting HTML markup and tweaking CSS counter styling across theme files, which is low-risk but non-functional churn.

B fixes a real indexing bug (1-based/unwrap_or(0) vs 0-based /t/tag/N routes), enforces the invariant with expect, always emits correct post links, and adds integration assertions. A only unwraps a vote-compare shell div and tweaks ranking-number CSS—cosmetic churn with little lasting design impact.

Side B fixes the semantics of `thread_post_index` by making rank history use the actual 0-based thread index consistently, replacing a silent `unwrap_or(0)` fallback with an `expect` that enforces the invariant, updating rendering to always generate the post link, correcting documentation, and adding integration tests for the behavior. Side A is largely a presentation change: it removes a wrapper element from the compare page and adjusts CSS styling for ranked lists across themes, with little impact on core functionality.

Side A makes a real, targeted fix: removing a dead/unused CSS wrapper section in the vote-compare template and consistently restyling the ranking-number markers across three theme files, producing a concrete visible/behavioral improvement. Side B is almost entirely mechanical rustfmt/reformatting noise across many files plus minor tooling config (toolchain components, VS Code settings) and one small CSS dedup, which has some value (tooling reliability) but contributes little functional substance and is largely reformatting churn.

Side A removes an unnecessary vote-compare-shell wrapper from the HTML template and ships real cross-theme CSS for ontology ranking lists (larger tabular rank numbers, missing ol.ont-ranking-list rules in retro/craft). Side B is almost entirely mechanical rustfmt whitespace/import churn plus minor toolchain pinning and a duplicate CSS merge, which adds little lasting product value.

Side A makes a user-visible structural change by removing the unnecessary `vote-compare-shell` wrapper from the vote comparison page and updates theme CSS to implement consistent numbered ranking lists across the default, retro, and retro_craft themes with improved counter styling. Side B is overwhelmingly a workspace-wide rustfmt reformat plus tooling configuration (pinning rustfmt/clippy, VS Code settings, minor CSS rule merge), with little functional impact beyond development ergonomics.

Side A is not a real commit at all: it dumps stray shell prompt output and duplicate copies of source files (forms.rs, ranking.rs, reducer.rs, vote.rs) into the repo root, seemingly by accident, adding noise rather than a coherent, working change. Side B is a small, legitimate, targeted commit that fixes real template/markup indentation and CSS styling (rank number sizing, retro theme list styles) with clear before/after diffs that actually improve the codebase.

A seeds core product systems—rank centrality and components, the event reducer/content model, vote-compare handlers, form-template UI actions, and a large tested URL parser—whereas B only unwraps a vote-compare HTML shell and tweaks ranking-list number CSS across themes. Notes and a few paste-noisy files weaken A slightly, but its lasting design and behavior still dwarf B’s small presentation cleanup.

Side A introduces substantial new project functionality, including a reducer and ranking engine with tests, a graph-based parser with extensive test coverage, vote handling, UI action parsing, form-template substitution, browser plumbing, and test scripts. Side B is primarily a presentation cleanup: it removes a wrapper element around the vote-compare page and adjusts CSS styling for ranked lists, improving layout but adding little enduring core functionality.

Side B ships substantial, functional infrastructure and features: a deployment pipeline (Dockerfile, fly.toml, GitHub Actions), a real /watch dashboard with live SSE audit events, security-relevant fixes (git auth headers, decimal context, single-contributor bypass, OpenRouter key guard), and new tests covering these behaviors. Side A is a small CSS/markup cleanup (removing a wrapper div and restyling rank numbers) that is tidy but of minor, cosmetic scope compared to B's operational and functional impact.

B ships production deployment (Docker, fly.toml, CI test-and-deploy), expands multi-repo/contributor configuration, and adds a real audited live /watch dashboard with SSE progress, status API, and tests—core lasting product and ops value. A only unwraps a vote-compare shell div and tweaks ontology ranking-list CSS counter styling, which is small presentational churn.

Side B adds substantial, lasting infrastructure and application functionality: production deployment (Dockerfile, Fly.io, GitHub Actions), a live audit dashboard with SSE-backed status updates, a new status API, improved Git authentication/configuration, deduplicated Git object verification, runtime error handling, and accompanying integration/unit tests. Side A mainly removes a wrapper element from the HTML template and adjusts CSS for ranking list styling, with only cosmetic/layout impact and no comparable functional improvement.

Side B includes an actual code fix in garden.rs (removing a wrapping <section> shell that likely fixed a layout/DOM bug), plus consistent CSS improvements (rank counter styling) applied across three themes with matching logic. Side A is purely cosmetic CSS tweaks (padding, colors, borders) duplicated across two theme files with no functional/behavioral change, making B's combination of a real markup fix plus coordinated styling slightly more substantive.

A delivers lasting vote UI polish: real button chrome, focus-visible rings, form/control styling, and a full retro-craft ontology/vote-compare theme that was missing. B only unwraps the vote-compare-shell div and enlarges ranking counters—useful but narrower cosmetic/structural tweaks.

Side A substantially improves the voting UI across both default and retro-craft themes by adding consistent styling, focus-visible accessibility, clearer interactive states, form controls, and layout refinements for the vote-compare page and pin controls. Side B mainly removes the unnecessary `vote-compare-shell` wrapper from the HTML and adjusts ranking-number CSS (larger counter styling), which is a comparatively modest structural cleanup and visual tweak with less lasting impact.

Side B fixes a real, user-visible layout bug (view counter overlapping login controls) with a small, targeted change to markup and CSS. Side A is mostly CSS styling tweaks for rank numbering plus a dedent/removal of a CSS class wrapper (minor cleanup with no functional bugfix), making both modest but B's fix is a clearer, more concrete improvement.

B fixes a real layout bug by moving the view counter into the top-nav flow and dropping fixed positioning so it no longer overlaps auth links—a small, necessary structural fix. A mostly unwraps a vote-compare shell (markup churn) and restyles ontology rank counters across themes, which is polish rather than a functional fix.

Side B fixes a concrete UI/layout bug by moving the view-count element into the existing top navigation and changing its CSS from `position: fixed` to normal flow with `margin-right: auto`, preventing overlap with the login/account controls in a maintainable way. Side A mostly removes a wrapper `section` around the vote-compare markup and updates ranking-list styling across themes; while the CSS improvements are useful, the HTML restructuring is largely cosmetic and the overall change is less directly tied to a lasting functional fix.

Side A makes an actual functional/UI change (removing an unused wrapper section and fixing rank-number styling across three theme CSS files), representing real, lasting improvements to markup and visuals. Side B is purely rustfmt/import-reorder churn with no behavioral change, offering minimal lasting value beyond style consistency.

Commit A removes an unnecessary vote-compare-shell wrapper from the HTML markup and adds concrete, lasting CSS improvements for ontology ranking-list counters (larger tabular numbers, flex layout, and full support in retro/craft themes). Commit B is pure mechanical reformatting of imports, signatures, and whitespace with zero behavioral or design effect.

Side A makes user-visible changes: it removes the outer `section.vote-compare-shell` wrapper from the vote comparison page and updates ontology ranking list CSS across the default, retro, and retro_craft themes by adding explicit counter styling, typography, and layout (`::before`, `counter-reset`, `counter-increment`, `font-variant-numeric`, etc.). Side B is purely formatting and import reordering with no behavioral or design change, so it contributes little lasting project value.