Fix monotonic discovery path filters

USER_GUIDE.md

@@ -2171,6 +2171,18 @@ A positional value containing an unescaped `*` or `?` is treated exactly like
 for example `cdidx files '**/*.cs'`. Positionals without those glob
 metacharacters remain filename-substring queries.
 
+For `files` and `map`, `--exclude-tests` also activates the implicit production
+source baseline. Its `src/**` include group is intersected with the user's
+repeatable `--path` group; patterns within either group remain OR-combined.
+The implicit source exclusions and every explicit `--exclude-path` are then
+applied together. Therefore adding a broad or narrow `--path` cannot re-enable
+files outside the baseline or files excluded by it. Generated-file policy,
+`--since`, path matching semantics, and count/row scope remain unchanged.
+Count, compact, summary, and map JSON expose this composition under
+`query_context.effective_path_scope`: `include_groups` and `exclude_groups`
+identify `implicit_source_baseline` versus `explicit_cli` provenance, while the
+operator fields describe how the groups and their patterns are combined.
+
 Output:
 
 ```
@@ -5853,6 +5865,17 @@ cdidx files --format compact --max-json-bytes 8000
 `cdidx files '**/*.cs'` のように引用してください。これらの glob metacharacter を
 含まない positional 値は、従来どおり filename substring query として扱われます。
 
+`files` と `map` では、`--exclude-tests` によって暗黙の production source baseline
+も有効になります。baseline の `src/**` include group は、繰り返し指定できるユーザーの
+`--path` group と AND で交差し、各 group 内の pattern は従来どおり OR で結合されます。
+その後、暗黙の source 除外とすべての明示 `--exclude-path` が合わせて適用されます。
+したがって、広いまたは狭い `--path` を追加しても baseline 外や baseline 除外済みの file
+が再び有効になることはありません。generated-file policy、`--since`、path match semantics、
+count / row の scope は維持されます。count、compact、summary、map の JSON は、この合成を
+`query_context.effective_path_scope` に公開します。`include_groups` と `exclude_groups` は
+`implicit_source_baseline` / `explicit_cli` の由来を示し、operator field は group と pattern
+の結合方法を示します。
+
 出力:
 
 ```

changelog.d/unreleased/5229.fixed.md

@@ -0,0 +1,23 @@
+---
+category: fixed
+issues:
+  - 5229
+affected:
+  - src/CodeIndex/Cli/SourceScopeDefaults.cs
+  - src/CodeIndex/Cli/QueryCommandRunner.Discovery.cs
+  - src/CodeIndex/Cli/QueryCommandRunner.Map.cs
+  - src/CodeIndex/Cli/QueryCommandRunner.ResultEnvelopes.cs
+  - src/CodeIndex/Database/DbReader.cs
+  - src/CodeIndex/Database/RepoMapBuilder.cs
+  - tests/CodeIndex.Tests/DbReaderTests.cs
+  - tests/CodeIndex.Tests/QueryCommandRunnerFilesTests.cs
+  - USER_GUIDE.md
+---
+
+## English
+
+- **`files --exclude-tests` path filters are now monotonic (#5229)** — explicit `--path` values are intersected with the implicit production-source baseline instead of replacing it, so broad or narrow includes cannot re-enable baseline-excluded files. Count, row, and map scopes now share the same composition, and JSON query metadata reports effective include/exclude provenance.
+
+## 日本語
+
+- **`files --exclude-tests` の path filter が単調になりました (#5229)** — 明示した `--path` は暗黙の production-source baseline を置き換えず、その baseline と交差するようになったため、広いまたは狭い include で baseline 除外済みの file が再び有効になることはありません。count、row、map は同じ scope 合成を共有し、JSON query metadata は有効な include / exclude の由来を報告します。

src/CodeIndex/Cli/QueryCommandRunner.ArgumentParser.cs

@@ -389,8 +389,8 @@ private void ApplySearchSourceOptionDefaults()
             }
 
             if (pathPatterns.Count == 0)
-                AddDistinct(pathPatterns, SearchAuditRecipes.DefaultSourcePathPatterns);
-            AddDistinct(excludePaths, SearchAuditRecipes.DefaultSourceExcludePaths);
+                AddDistinct(pathPatterns, SourceScopeDefaults.IncludePaths);
+            AddDistinct(excludePaths, SourceScopeDefaults.ExcludePaths);
             AddSourceOnlyDefaultExcludeOrigin(excludeOrigins, matchOrigins, SearchMatchClassifier.Comment);
             AddSourceOnlyDefaultExcludeOrigin(excludeOrigins, matchOrigins, SearchMatchClassifier.HelpText);
             AddSourceOnlyDefaultExcludeOrigin(excludeOrigins, matchOrigins, SearchMatchClassifier.SchemaDescription);

src/CodeIndex/Cli/QueryCommandRunner.Discovery.cs

@@ -532,7 +532,14 @@ public static int RunFiles(string[] cmdArgs, JsonSerializerOptions jsonOptions)
         {
             if (options.CountOnly)
             {
-                var counts = reader.CountListFiles(options.Query, options.Lang, filesScope.PathPatterns, filesScope.ExcludePaths, filesScope.ExcludeTests, options.Since);
+                var counts = reader.CountListFiles(
+                    options.Query,
+                    options.Lang,
+                    filesScope.PathPatterns,
+                    filesScope.ExcludePaths,
+                    filesScope.ExcludeTests,
+                    options.Since,
+                    requiredPathPatterns: filesScope.RequiredPathPatterns);
                 var generatedFileCountExcluded = CountGeneratedFilesExcluded(reader, options, filesScope);
                 if (options.Json)
                 {
@@ -572,7 +579,8 @@ public static int RunFiles(string[] cmdArgs, JsonSerializerOptions jsonOptions)
                 filesScope.ExcludeTests,
                 options.Since,
                 orderBySize: options.RawBytes,
-                offset: JsonEnvelopeWrapper.GetBoundedResponseOffset("files"));
+                offset: JsonEnvelopeWrapper.GetBoundedResponseOffset("files"),
+                requiredPathPatterns: filesScope.RequiredPathPatterns);
             Func<FileResult, JsonNode?> rowFactory =
                 result => ToFileDiscoveryJsonNode(result, jsonOptions, options.OutputFormat == OutputFormatCompact);
             if (results.Count == 0)
@@ -610,7 +618,14 @@ public static int RunFiles(string[] cmdArgs, JsonSerializerOptions jsonOptions)
 
             if (IsDiscoveryNdjson(options))
             {
-                var counts = reader.CountListFiles(options.Query, options.Lang, filesScope.PathPatterns, filesScope.ExcludePaths, filesScope.ExcludeTests, options.Since);
+                var counts = reader.CountListFiles(
+                    options.Query,
+                    options.Lang,
+                    filesScope.PathPatterns,
+                    filesScope.ExcludePaths,
+                    filesScope.ExcludeTests,
+                    options.Since,
+                    requiredPathPatterns: filesScope.RequiredPathPatterns);
                 var stream = WriteDiscoveryNdjson(
                     reader,
                     options,
@@ -625,7 +640,14 @@ public static int RunFiles(string[] cmdArgs, JsonSerializerOptions jsonOptions)
 
             if (ShouldWriteBoundedDiscoveryJsonPayload(options))
             {
-                var counts = reader.CountListFiles(options.Query, options.Lang, filesScope.PathPatterns, filesScope.ExcludePaths, filesScope.ExcludeTests, options.Since);
+                var counts = reader.CountListFiles(
+                    options.Query,
+                    options.Lang,
+                    filesScope.PathPatterns,
+                    filesScope.ExcludePaths,
+                    filesScope.ExcludeTests,
+                    options.Since,
+                    requiredPathPatterns: filesScope.RequiredPathPatterns);
                 return WriteBoundedDiscoveryJsonPayload(
                     reader,
                     options,
@@ -693,24 +715,26 @@ internal static (List<string>? Queries, bool HadExplicitInput) BuildSymbolQueryL
 
     private sealed record DiscoveryFileScopeFilters(
         IReadOnlyList<string> PathPatterns,
+        IReadOnlyList<string> RequiredPathPatterns,
         IReadOnlyList<string> ExcludePaths,
         bool ExcludeTests);
 
     private static DiscoveryFileScopeFilters BuildDiscoveryFileScopeFilters(QueryCommandOptions options)
     {
-        if (!options.ExcludeTests || options.PathPatterns.Count > 0)
+        if (!options.ExcludeTests)
         {
             return new(
                 options.PathPatterns,
+                [],
                 options.ExcludePaths,
                 options.ExcludeTests);
         }
 
-        var pathPatterns = new List<string>(options.PathPatterns);
-        AddDistinct(pathPatterns, SearchAuditRecipes.DefaultSourcePathPatterns);
         var excludePaths = new List<string>(options.ExcludePaths);
-        AddDistinct(excludePaths, SearchAuditRecipes.DefaultSourceExcludePaths);
-        return new(pathPatterns, excludePaths, ExcludeTests: true);
+        AddDistinct(excludePaths, SourceScopeDefaults.ExcludePaths);
+        options.DiscoveryBaselineIncludePaths = SourceScopeDefaults.IncludePaths;
+        options.DiscoveryBaselineExcludePaths = SourceScopeDefaults.ExcludePaths;
+        return new(options.PathPatterns, SourceScopeDefaults.IncludePaths, excludePaths, ExcludeTests: true);
     }
 
     private static int? CountGeneratedFilesExcluded(DbReader reader, QueryCommandOptions options, DiscoveryFileScopeFilters filesScope)
@@ -725,7 +749,8 @@ private static DiscoveryFileScopeFilters BuildDiscoveryFileScopeFilters(QueryCom
                 filesScope.ExcludePaths,
                 filesScope.ExcludeTests,
                 options.Since,
-                generatedOnly: true).Count;
+                generatedOnly: true,
+                requiredPathPatterns: filesScope.RequiredPathPatterns).Count;
 
     private static void AddGeneratedFileFilterJsonFields(
         JsonObject payload,

src/CodeIndex/Cli/QueryCommandRunner.Map.cs

@@ -80,7 +80,8 @@ public static int RunMap(string[] cmdArgs, JsonSerializerOptions jsonOptions)
                 oversizedByteThreshold: evaluateIssueDraftCandidates ? MapIssueDraftByteThreshold : null,
                 offset: JsonEnvelopeWrapper.GetBoundedResponseOffset("map"),
                 requestedCollection: JsonEnvelopeWrapper.GetBoundedMapCollection(),
-                summaryProjection: JsonEnvelopeWrapper.IsBoundedMapScalarProjection());
+                summaryProjection: JsonEnvelopeWrapper.IsBoundedMapScalarProjection(),
+                requiredPathPatterns: filesScope.RequiredPathPatterns);
             var generatedFileCountExcluded = options.IncludeGenerated
                 ? 0
                 : !reader.GeneratedFileFilterAvailable
@@ -90,7 +91,8 @@ public static int RunMap(string[] cmdArgs, JsonSerializerOptions jsonOptions)
                     pathPatterns: filesScope.PathPatterns,
                     excludePathPatterns: filesScope.ExcludePaths,
                     excludeTests: filesScope.ExcludeTests,
-                    generatedOnly: true).Count;
+                    generatedOnly: true,
+                    requiredPathPatterns: filesScope.RequiredPathPatterns).Count;
             WorkspaceMetadataEnricher.Enrich(map, options.DbPath, options.DbPathExplicit);
             var compactTruncation = options.Compact ? ApplyRepoMapCompactCaps(map, compactLimit, options) : null;
 
@@ -266,6 +268,8 @@ private static JsonObject BuildRepoMapJsonPayload(
             options,
             generatedFileCountExcluded,
             generatedFileFilterAvailable);
+        if (options.DiscoveryBaselineIncludePaths.Count > 0 || options.DiscoveryBaselineExcludePaths.Count > 0)
+            payload["query_context"] = BuildQueryContextJson(options, jsonOptions);
         if (options.MapSummaryOnly)
         {
             KeepRepoMapJsonProperties(payload, RepoMapSummaryJsonProperties);
@@ -582,6 +586,7 @@ private static string BuildRepoMapIssueDraftBody(RepoFileSummaryResult file, Jso
         "worktree_head_changed",
         "head_freshness",
         "graph_table_available",
+        "query_context",
     };
 
     private static readonly IReadOnlyDictionary<string, string[]> RepoMapSectionJsonProperties = new Dictionary<string, string[]>(StringComparer.Ordinal)

src/CodeIndex/Cli/QueryCommandRunner.ResultEnvelopes.cs

@@ -246,6 +246,8 @@ private static JsonObject BuildQueryContextJson(QueryCommandOptions options, Jso
             query["sort"] = options.SymbolSortMode.ToString().ToLowerInvariant();
         if (options.ExcludeTests)
             query["exclude_tests"] = true;
+        if (options.DiscoveryBaselineIncludePaths.Count > 0 || options.DiscoveryBaselineExcludePaths.Count > 0)
+            query["effective_path_scope"] = BuildEffectiveDiscoveryPathScopeJson(options, jsonOptions);
         if (options.ExcludeComments)
             query["exclude_comments"] = true;
         if (options.ExcludeStrings)
@@ -321,6 +323,55 @@ private static JsonObject BuildQueryContextJson(QueryCommandOptions options, Jso
         return query;
     }
 
+    private static JsonObject BuildEffectiveDiscoveryPathScopeJson(
+        QueryCommandOptions options,
+        JsonSerializerOptions jsonOptions)
+    {
+        var context = CliJsonSerializerContextFactory.Create(jsonOptions);
+        var includeGroups = new JsonArray
+        {
+            new JsonObject
+            {
+                ["origin"] = "implicit_source_baseline",
+                ["patterns"] = JsonSerializer.SerializeToNode(options.DiscoveryBaselineIncludePaths.ToList(), context.ListString),
+            },
+        };
+        if (options.PathPatterns.Count > 0)
+        {
+            includeGroups.Add(new JsonObject
+            {
+                ["origin"] = "explicit_cli",
+                ["patterns"] = JsonSerializer.SerializeToNode(options.PathPatterns, context.ListString),
+            });
+        }
+
+        var excludeGroups = new JsonArray
+        {
+            new JsonObject
+            {
+                ["origin"] = "implicit_source_baseline",
+                ["patterns"] = JsonSerializer.SerializeToNode(options.DiscoveryBaselineExcludePaths.ToList(), context.ListString),
+            },
+        };
+        if (options.ExcludePaths.Count > 0)
+        {
+            excludeGroups.Add(new JsonObject
+            {
+                ["origin"] = "explicit_cli",
+                ["patterns"] = JsonSerializer.SerializeToNode(options.ExcludePaths, context.ListString),
+            });
+        }
+
+        return new JsonObject
+        {
+            ["include_group_operator"] = "and",
+            ["patterns_within_include_group_operator"] = "or",
+            ["include_groups"] = includeGroups,
+            ["exclude_group_operator"] = "or",
+            ["exclude_groups"] = excludeGroups,
+        };
+    }
+
     private static void AddReferenceRankingQueryContextJson(
         JsonObject payload,
         QueryCommandOptions options,

src/CodeIndex/Cli/QueryCommandRunner.cs

@@ -139,6 +139,8 @@ public sealed class QueryCommandOptions
     public string? SolutionFilter { get; init; }
     public List<string> ExcludePaths { get; init; } = [];
     public bool ExcludeTests { get; init; }
+    internal IReadOnlyList<string> DiscoveryBaselineIncludePaths { get; set; } = [];
+    internal IReadOnlyList<string> DiscoveryBaselineExcludePaths { get; set; } = [];
     public bool IncludeGenerated { get; init; }
     public bool CountOnly { get; init; }
     public bool GroupPartials { get; init; }