feat(config): Add per-analysis ignore globs for dupes, health, and dead-code

README.md

@@ -442,11 +442,15 @@ walking up from the analysis root to the nearest directory containing one
   "roots": ["MyApp.Program.Main"],
   "libraryProjects": ["MyLib"],
   "ignore": ["Migrations/**", "Generated/", "**/*.designer.cs"],
+  "deadCode": {
+    "ignore": ["Plugins/**"]
+  },
   "dupes": {
     "mode": "semantic",
     "minTokens": 100,
     "minLines": 10,
-    "minOccurrences": 2
+    "minOccurrences": 2,
+    "ignore": ["**/InsurerUnionQuery.cs"]
   },
   "health": {
     "maxComplexity": 15,
@@ -469,7 +473,11 @@ explicit `--aggressive`, `--root`, or `--library` always wins, otherwise the
 config's value applies, otherwise the built-in default (`false` / no extra
 roots / no extra library projects). The same `ignore` list also applies to
 `roe dupes`, since a duplicate spans multiple files and doesn't map cleanly
-onto a single-line inline suppression comment.
+onto a single-line inline suppression comment. Each command's section takes
+an `ignore` list of its own with the same rules, unioned with the top-level
+one and applied only to that command — so accepting an intentional duplicate
+through `dupes.ignore` doesn't cost the file its dead-code and health
+coverage.
 
 The `dupes` block sets defaults for `roe dupes`' matching mode and
 thresholds, and the `health` block does the same for `roe health`'s

_typos.toml

@@ -8,3 +8,5 @@ extend-exclude = ["Cargo.lock", "tests/fixtures/", "tests/snapshots/"]
 # `rejects_unknown_fields` in src/config.rs to prove serde rejects near-miss
 # keys instead of silently ignoring them.
 agressive = "agressive"
+# Same idea for the `ignore` field, used by `rejects_unknown_dead_code_fields`.
+ignor = "ignor"

docs/commands/dead-code.mdx

@@ -50,7 +50,10 @@ never flagged even though no executable in the workspace uses it:
 roe dead-code --library MyLib
 ```
 
-Both can also be set persistently in a [config file](/configuration).
+Both can also be set persistently in a [config file](/configuration). A
+config's `deadCode.ignore` glob list drops every dead-code finding in
+matching files without affecting the other commands — see
+[Suppressing findings](/suppressing-findings).
 
 ## What counts as an entry point
 

docs/commands/dupes.mdx

@@ -56,6 +56,16 @@ roe dupes --mode semantic path/to/solution
 ## Ignoring files
 
 Duplicates span multiple files, so there's no inline suppression comment for
-them. Instead, use the `ignore` glob list in a
-[config file](/configuration) — it applies to every command, dropping every
-finding in matching files.
+them. Instead, use ignore globs in a [config file](/configuration): the
+top-level `ignore` list applies to every command, and the `dupes.ignore`
+list adds patterns only the duplicate analysis honours — so accepting an
+intentional duplicate doesn't cost the file its dead-code and health
+coverage.
+
+```json roe.json
+{
+  "dupes": {
+    "ignore": ["**/InsurerUnionQuery.cs"]
+  }
+}
+```

docs/commands/health.mdx

@@ -115,8 +115,8 @@ or a `??` whose fallback contains a ternary is counted on its own merits.
 The footer's `N project(s), N file(s), N symbol(s) scanned` counts what was
 *eligible to be reported*, not what was parsed. Anything ruled out up front —
 test projects under `--exclude-tests`, files matched by the config's
-[`ignore` globs](/configuration) — is subtracted, and named on a line of its
-own:
+[`ignore` globs](/configuration) (top-level or `health.ignore`) — is
+subtracted, and named on a line of its own:
 
 ```text
 found 8 issues across 7 locations in 6 files — 2 project(s), 118 file(s), 1204 symbol(s) scanned in 153 ms
@@ -398,5 +398,6 @@ public void ParseEverything(string input)
 ```
 
 Circular dependencies span multiple files, so — like duplicates — they have
-no inline comment. Use the `ignore` globs in a
-[config file](/configuration) instead.
+no inline comment. Use ignore globs in a [config file](/configuration)
+instead — `health.ignore` scopes the suppression to health alone, so the
+file keeps its dead-code and dupes coverage.

docs/configuration.mdx

@@ -27,11 +27,15 @@ error.
   "roots": ["MyApp.Program.Main"],
   "libraryProjects": ["MyLib"],
   "ignore": ["Migrations/**", "Generated/", "**/*.designer.cs"],
+  "deadCode": {
+    "ignore": ["Plugins/**"]
+  },
   "dupes": {
     "mode": "semantic",
     "minTokens": 100,
     "minLines": 10,
-    "minOccurrences": 2
+    "minOccurrences": 2,
+    "ignore": ["**/InsurerUnionQuery.cs"]
   },
   "health": {
     "maxComplexity": 15,
@@ -41,7 +45,8 @@ error.
     "maxFileLines": 750,
     "maxTypeMembers": 25,
     "excludeTests": true,
-    "baseline": "roe-baseline.json"
+    "baseline": "roe-baseline.json",
+    "ignore": ["**/GeneratedModels.cs"]
   }
 }
 ```
@@ -51,9 +56,10 @@ error.
 | `aggressive` | boolean | Also flag enum members and public settable auto-properties. Default for the `--aggressive` flag. |
 | `roots` | string[] | Fully-qualified symbol names to treat as extra entry-point roots. Default for `--root`. |
 | `libraryProjects` | string[] | Project names to always treat in library mode (public API counts as used). Default for `--library`. |
-| `ignore` | string[] | Glob patterns; every finding in a matching file is dropped. Applies to all three commands. |
-| `dupes` | object | Matching mode and thresholds for [`roe dupes`](/commands/dupes). Every key is optional. |
-| `health` | object | Thresholds and the baseline path for [`roe health`](/commands/health). Every key is optional. |
+| `ignore` | string[] | Glob patterns; every finding in a matching file is dropped. Applies to all three commands; each command's section can add its own. |
+| `deadCode` | object | Extra `ignore` globs applied only to [`roe dead-code`](/commands/dead-code). |
+| `dupes` | object | Matching mode, thresholds, and extra `ignore` globs for [`roe dupes`](/commands/dupes). Every key is optional. |
+| `health` | object | Thresholds, the baseline path, and extra `ignore` globs for [`roe health`](/commands/health). Every key is optional. |
 
 Unknown fields are rejected, so a typo fails loudly instead of being
 silently ignored.
@@ -69,6 +75,7 @@ Every field under `dupes` is optional and corresponds to a flag on
 | `minTokens` | number | `--min-tokens` | `50` |
 | `minLines` | number | `--min-lines` | `5` |
 | `minOccurrences` | number | `--min-occurrences` | `2` |
+| `ignore` | string[] | — | none |
 
 This block is what calibrates a combined run: [`roe check`](/commands/check)
 takes no dupes flags of its own, so raising `minTokens` here is how a one-line
@@ -89,6 +96,7 @@ Every field under `health` is optional and corresponds to a flag on
 | `maxTypeMembers` | number | `--max-type-members` | `20` |
 | `excludeTests` | boolean | `--exclude-tests` | `false` |
 | `baseline` | string | `--baseline` | none |
+| `ignore` | string[] | — | none |
 
 Committing these is usually better than passing six flags on every CI
 invocation, and it keeps local runs and CI in agreement.
@@ -108,9 +116,33 @@ full-backlog run.
 A trailing `/` matches the whole directory, so `"Generated/"` needs no `**`.
 Patterns containing `..` are unsupported and produce a warning.
 
-The same `ignore` list also applies to `roe dupes` and to `roe health`'s
-circular dependencies, since both span multiple files and don't map cleanly
-onto a single-line [inline suppression comment](/suppressing-findings).
+The top-level `ignore` list applies to all three commands — including
+`roe dupes` and `roe health`'s circular dependencies, which span multiple
+files and don't map cleanly onto a single-line
+[inline suppression comment](/suppressing-findings).
+
+Each command's section takes an `ignore` list of its own with the same
+rules, unioned with the top-level list and applied only to that command.
+That keeps a suppression as narrow as the exception it accepts: ignoring an
+intentional duplicate through `dupes.ignore` doesn't cost the file its
+dead-code and health coverage.
+
+```yaml roe.yaml
+ignore:
+  - "Generated/**"
+deadCode:
+  ignore:
+    - "Plugins/**"
+dupes:
+  ignore:
+    - "**/InsurerUnionQuery.cs"
+health:
+  ignore:
+    - "**/GeneratedModels.cs"
+```
+
+A scoped list only ever adds patterns — it can't re-include a file the
+top-level list ignores — and an empty one is a no-op.
 
 ## Precedence
 
@@ -136,6 +168,10 @@ Dupes settings follow the same per-field order: `--min-tokens 100` overrides
 carries a value, so unlike `--aggressive` and `--exclude-tests` it has no
 on-only nuance — `--mode exact` overrides a config's `"semantic"` cleanly.
 
+Ignore lists sit outside this order entirely: they have no CLI flag, and a
+command's own `ignore` list is unioned with the top-level one rather than
+replacing it.
+
 To confirm a config-only `excludeTests` or `ignore` actually applied, read the
 `roe health` footer: the scanned counts narrow by whatever was excluded, and
 an `excluded:` line names it. See

docs/reference/json-output.mdx

@@ -216,7 +216,7 @@ including its own `version` and `root`, so anything already written against
 | --- | --- |
 | `version` | Schema version, `1`. |
 | `root` | The analysis root path. |
-| `summary` | Scan totals: `projects`, `filesScanned`, `symbols`, one count per check (`highComplexity`, `highCognitiveComplexity`, `longMethods`, `tooManyParameters`, `largeFiles`, `largeTypes`, `circularDependencies`), and `elapsedMs`. `commitsWalked` is present only when `--hotspots` was passed. The three scan totals count what was eligible to be reported, so they narrow under `--exclude-tests` and the config's `ignore` globs. |
+| `summary` | Scan totals: `projects`, `filesScanned`, `symbols`, one count per check (`highComplexity`, `highCognitiveComplexity`, `longMethods`, `tooManyParameters`, `largeFiles`, `largeTypes`, `circularDependencies`), and `elapsedMs`. `commitsWalked` is present only when `--hotspots` was passed. The three scan totals count what was eligible to be reported, so they narrow under `--exclude-tests` and the config's `ignore` globs (top-level and `health.ignore`). |
 | `summary.baselined` | How many findings and cycles a [baseline](/commands/health#baselines) hid — they are counted nowhere else in this document. Present only when a baseline was in force, so `0` means "the baseline is fully ratcheted" and absent means "no baseline was used". |
 | `summary.excluded` | What was ruled out before any check ran: `testProjects` (an array of project names, present under `--exclude-tests`) and `ignoredFiles` (a count). Omitted entirely when nothing was excluded. |
 | `findings[].kind` | `high-complexity`, `high-cognitive-complexity`, `long-method`, `too-many-parameters`, `large-file`, or `large-type` — the same names used by [inline suppressions](/suppressing-findings). |

docs/suppressing-findings.mdx

@@ -78,10 +78,25 @@ To drop every finding in matching files — for `dead-code`, `dupes`, and
 }
 ```
 
+To keep a suppression as narrow as the exception it accepts, scope it to one
+command instead: each command's section takes an `ignore` list of its own,
+unioned with the top-level one and applied only to that command. Accepting
+an intentional duplicate this way doesn't cost the file its dead-code and
+health coverage:
+
+```json roe.json
+{
+  "dupes": {
+    "ignore": ["**/InsurerUnionQuery.cs"]
+  }
+}
+```
+
 See [Configuration](/configuration) for glob semantics and config discovery.
 
 <Note>
   Duplicate groups and circular dependencies span multiple files, so neither
   maps onto a single-line comment and neither has an inline suppression.
-  Ignore globs are the only way to silence them.
+  Ignore globs — top-level, `dupes.ignore`, or `health.ignore` — are the only
+  way to silence them.
 </Note>