What this supports
Review usability claims
The examples support claims about alignment, visible statuses, one-sided insertions, configurable noise handling, and manual correction.
Observed product behaviours
Line diff versus structure-aware diff: an evidence matrix
This page records what the current AlignDiff examples are designed to demonstrate. It separates observed comparison behaviour from claims the tool does not make.
Transformation matrix
What changes, what AlignDiff surfaces, and what still needs review
| Transformation | Structure-aware observation | Reviewer still checks |
|---|---|---|
| Function moved | The likely counterpart can remain paired and receive a moved or moved-and-modified status. | Whether calls, parameters, and behaviour also changed. |
| Function renamed | A structurally similar function can appear as a rename candidate with matching reasons. | Whether the candidate is the intended counterpart. |
| Statements reordered | Related calls, assignments, and returns can receive spacer rows instead of cascading into misleading row matches. | Whether execution order changed meaningfully. |
| Comments changed | The reviewer can include comment changes or choose the current ignore-comments option. | Whether comments encode requirements or operational warnings. |
| Whitespace changed | The current ignore-whitespace option can suppress formatting-only noise. | Whether whitespace is syntactically significant in the chosen language. |
| Return expression changed | The paired function remains together while the detailed return tokens are highlighted. | Whether the new result is correct for all inputs. |
| Similar functions compete | Possible candidates can remain reviewable and the automatic pair can be overridden locally. | Which candidate reflects developer intent. |
Visible evidence
One public fixture combines movement, renaming, insertion, and detailed edits
What this does not support
Behavioural equivalence claims
A structure-aware result does not prove that two programs behave identically, that a pull request is safe, or that an automatic pair is always correct.
Repeat the observation
Load the examples and inspect the actual result
Use the recommended Python product demo first, then compare the JavaScript and TypeScript fixtures or paste your own non-sensitive files.
Continue exploring