Scan, DryRun, Deduplicate and Stats each copy the path prompt and the scan→hash→group pipeline, so every fix to that path must be made four times

[matt-edmondson]

What's wrong

The four verbs in FileDeduplicator/Verbs/ duplicate two blocks almost word for word.

• The path prompt. ValidateArgs() in Scan.cs:14-30, DryRun.cs:14-30, Deduplicate.cs:14-30 and Stats.cs:15-31 is the same method: PathString is "." or "" → prompt → Console.ReadLine()?.Trim() → abort when the input is empty.
• The pipeline. Steps 1-3 of each Run (Scan.cs:37-61, DryRun.cs:37-61, Deduplicate.cs:37-61, Stats.cs:38-56) are the same sequence: FileScanner.ScanForFiles → print the count → return on zero → FileHasher.HashFiles → Deduplicator.GroupByHash → Deduplicator.FindDuplicates. For Scan and DryRun, lines 14-61 differ only in the verb type and one banner line.

On top of that, Scan.cs:64-83 renders its own KEEP/DELETE listing and wasted-space total instead of using DuplicateReport.PlanDeletions. DuplicateReport's own remarks say all verbs should build this listing in one place so they cannot drift apart.

Why it matters

Several open issues change exactly this duplicated code. Each fix has to land four times, identically, or the verbs start to behave differently:

• Passing -p . explicitly is treated as "no path" and prompts, swallowing the next stdin line as the path #145: an explicit -p . is treated as "no path" (the PathString is "." or "" check, in four places).
• A quoted path pasted at the "Enter the path" prompt is scanned as <cwd>/"<path>" and reported as "Directory not found" #152: a quoted path pasted at the prompt (the ReadLine()?.Trim(), in four places).
• A file deleted or moved during hashing crashes DryRun/Deduplicate/Stats with FileNotFoundException after all the hashing is done #141: a file vanishes during hashing (the pipeline; PR Tolerate a file that vanishes during the hash pass #156 is open).
• Every verb reads and SHA-256-hashes every file in full, even files whose size no other file shares #147: pre-group files by size before hashing (the pipeline, in four places).
• Every verb exits with code 0 even when the directory is missing, deletions fail, or the arguments are invalid #144: exit codes (the early returns in each copy).

If a fix reaches three verbs and misses the fourth, one verb reports a different set of duplicates, or accepts different input, than the others. The codebase already treats that kind of divergence as its worst defect: DryRunEquivalenceTests exists because DryRun has to predict Deduplicate exactly.

Suggested fix

• Move the prompt into BaseVerb<T>, and have each verb supply only its prompt text (for example, through an abstract PromptText property).
• Add one method, for example DuplicateFinder.Find(AbsoluteDirectoryPath) -> ScanResult, that returns the scanned files, the hashes and the duplicate groups. Each verb calls it once and keeps only its own reporting.
• Make Scan render from DuplicateReport.PlanDeletions, like DryRun and Deduplicate do. It can keep its [KEEP]/[DELETE] suffix format.

Acceptance criteria

• ValidateArgs and the scan/hash/group sequence each exist in exactly one place.
• Scan's KEEP/DELETE decisions and byte total come from DuplicateReport.
• The existing VerbOutputTests, VerbReuseTests and DryRunEquivalenceTests pass unchanged.
• Ideally, land this before or together with the fixes for Passing -p . explicitly is treated as "no path" and prompts, swallowing the next stdin line as the path #145, A quoted path pasted at the "Enter the path" prompt is scanned as <cwd>/"<path>" and reported as "Directory not found" #152, Every verb reads and SHA-256-hashes every file in full, even files whose size no other file shares #147 and Every verb exits with code 0 even when the directory is missing, deletions fail, or the arguments are invalid #144, so each of those is made once.

[matt-edmondson]

Triage

• Category: Improvement (a refactor that removes duplication)
• Priority: Medium. There is no user-visible defect of its own. It is prioritised because it gates Passing -p . explicitly is treated as "no path" and prompts, swallowing the next stdin line as the path #145, A quoted path pasted at the "Enter the path" prompt is scanned as <cwd>/"<path>" and reported as "Directory not found" #152, Every verb reads and SHA-256-hashes every file in full, even files whose size no other file shares #147 and Every verb exits with code 0 even when the directory is missing, deletions fail, or the arguments are invalid #144: without it, each of those fixes has to be made four times identically, and a miss makes the verbs disagree. That kind of drift is the codebase's worst class of defect, which is why DryRunEquivalenceTests exists.
• Area / suggested assignment: The CLI verbs (FileDeduplicator/Verbs/), BaseVerb<T> and DuplicateReport. Whoever owns the verbs should take it.
• Duplicates / in progress: No duplicate found, and no open PR covers the refactor itself. PR Tolerate a file that vanishes during the hash pass #156 ("Tolerate a file that vanishes during the hash pass", for A file deleted or moved during hashing crashes DryRun/Deduplicate/Stats with FileNotFoundException after all the hashing is done #141) is open and changes the duplicated pipeline. Either merge Tolerate a file that vanishes during the hash pass #156 first and extract from the result, or fold its change into the new single DuplicateFinder.Find, so it isn't lost.
• Next step: Land this refactor before Passing -p . explicitly is treated as "no path" and prompts, swallowing the next stdin line as the path #145, A quoted path pasted at the "Enter the path" prompt is scanned as <cwd>/"<path>" and reported as "Directory not found" #152, Every verb reads and SHA-256-hashes every file in full, even files whose size no other file shares #147 and Every verb exits with code 0 even when the directory is missing, deletions fail, or the arguments are invalid #144 so each of them is fixed in one place. Keep the existing VerbOutputTests, VerbReuseTests and DryRunEquivalenceTests unchanged as the regression gate.

---

Generated by Claude Code