Add churn mode for incremental correctness testing

- Add -churn n CLI option to simulate file add/remove cycles
- Alternates between removing n random files and adding them back
- Shows issue count changes per run with detailed diff
- Add remove_batch to ReactiveFileCollection for batched removals (28x faster)
- Add reset_stats to Reactive for per-operation stat tracking
- Include churn time in timing report with proper accounting
- Print aggregate stats at end: mean/std for churn time and issue changes
- Add skip_file parameter to filter removed files from processing

Author: cristianoc

analysis/reactive/src/Reactive.ml

@@ -144,6 +144,21 @@ module Registry = struct
     Hashtbl.clear combinators;
     dirty_nodes := []
 
+  let reset_stats () =
+    Hashtbl.iter
+      (fun _ info ->
+        info.stats.deltas_received <- 0;
+        info.stats.entries_received <- 0;
+        info.stats.adds_received <- 0;
+        info.stats.removes_received <- 0;
+        info.stats.process_count <- 0;
+        info.stats.process_time_ns <- 0L;
+        info.stats.deltas_emitted <- 0;
+        info.stats.entries_emitted <- 0;
+        info.stats.adds_emitted <- 0;
+        info.stats.removes_emitted <- 0)
+      nodes
+
   (** Generate Mermaid diagram of the pipeline *)
   let to_mermaid () =
     let buf = Buffer.create 256 in
@@ -460,8 +475,8 @@ let flatMap ~name (src : ('k1, 'v1) t) ~f ?merge () : ('k2, 'v2) t =
       Hashtbl.remove provenance k1;
       List.iter
         (fun k2 ->
-          match Hashtbl.find_opt contributions k2 with
-          | None -> ()
+             match Hashtbl.find_opt contributions k2 with
+             | None -> ()
           | Some contribs -> Hashtbl.remove contribs k1)
         target_keys;
       target_keys
@@ -472,37 +487,37 @@ let flatMap ~name (src : ('k1, 'v1) t) ~f ?merge () : ('k2, 'v2) t =
     Hashtbl.replace provenance k1 target_keys;
     List.iter
       (fun (k2, v2) ->
-        let contribs =
-          match Hashtbl.find_opt contributions k2 with
-          | Some c -> c
-          | None ->
-            let c = Hashtbl.create 4 in
-            Hashtbl.replace contributions k2 c;
-            c
-        in
+           let contribs =
+             match Hashtbl.find_opt contributions k2 with
+             | Some c -> c
+             | None ->
+               let c = Hashtbl.create 4 in
+               Hashtbl.replace contributions k2 c;
+               c
+           in
         Hashtbl.replace contribs k1 v2)
       entries;
     target_keys
   in
 
   let process_entry (k1, v1_opt) =
-    let old_affected = remove_source k1 in
+        let old_affected = remove_source k1 in
     let new_affected =
       match v1_opt with
       | None -> []
       | Some v1 ->
         let entries = f k1 v1 in
         add_source k1 entries
     in
-    let all_affected = old_affected @ new_affected in
+        let all_affected = old_affected @ new_affected in
     (* Deduplicate *)
-    let seen = Hashtbl.create (List.length all_affected) in
+        let seen = Hashtbl.create (List.length all_affected) in
     List.filter_map
       (fun k2 ->
-        if Hashtbl.mem seen k2 then None
-        else (
-          Hashtbl.replace seen k2 ();
-          recompute_target k2))
+               if Hashtbl.mem seen k2 then None
+               else (
+                 Hashtbl.replace seen k2 ();
+                 recompute_target k2))
       all_affected
   in
 
@@ -626,8 +641,8 @@ let join ~name (left : ('k1, 'v1) t) (right : ('k2, 'v2) t) ~key_of ~f ?merge ()
       Hashtbl.remove provenance k1;
       List.iter
         (fun k3 ->
-          match Hashtbl.find_opt contributions k3 with
-          | None -> ()
+             match Hashtbl.find_opt contributions k3 with
+             | None -> ()
           | Some contribs -> Hashtbl.remove contribs k1)
         target_keys;
       target_keys
@@ -638,14 +653,14 @@ let join ~name (left : ('k1, 'v1) t) (right : ('k2, 'v2) t) ~key_of ~f ?merge ()
     Hashtbl.replace provenance k1 target_keys;
     List.iter
       (fun (k3, v3) ->
-        let contribs =
-          match Hashtbl.find_opt contributions k3 with
-          | Some c -> c
-          | None ->
-            let c = Hashtbl.create 4 in
-            Hashtbl.replace contributions k3 c;
-            c
-        in
+           let contribs =
+             match Hashtbl.find_opt contributions k3 with
+             | Some c -> c
+             | None ->
+               let c = Hashtbl.create 4 in
+               Hashtbl.replace contributions k3 c;
+               c
+           in
         Hashtbl.replace contribs k1 v3)
       entries;
     target_keys
@@ -742,7 +757,7 @@ let join ~name (left : ('k1, 'v1) t) (right : ('k2, 'v2) t) ~key_of ~f ?merge ()
         | Some left_keys ->
           List.iter
             (fun k1 ->
-              match Hashtbl.find_opt left_entries k1 with
+                 match Hashtbl.find_opt left_entries k1 with
               | Some v1 ->
                 let affected = process_left_entry k1 v1 in
                 all_affected := affected @ !all_affected
@@ -975,7 +990,7 @@ let fixpoint ~name ~(init : ('k, unit) t) ~(edges : ('k, 'k list) t) () :
     let frontier = Queue.create () in
 
     (* Start from all roots *)
-    Hashtbl.iter
+      Hashtbl.iter
       (fun k () ->
         Hashtbl.replace new_current k ();
         Queue.add k frontier)
@@ -985,9 +1000,9 @@ let fixpoint ~name ~(init : ('k, unit) t) ~(edges : ('k, 'k list) t) () :
     while not (Queue.is_empty frontier) do
       let k = Queue.pop frontier in
       match Hashtbl.find_opt edge_map k with
-      | None -> ()
+            | None -> ()
       | Some successors ->
-        List.iter
+          List.iter
           (fun succ ->
             if not (Hashtbl.mem new_current succ) then (
               Hashtbl.replace new_current succ ();
@@ -1025,7 +1040,7 @@ let fixpoint ~name ~(init : ('k, unit) t) ~(edges : ('k, 'k list) t) () :
     let needs_full_recompute = ref false in
 
     (* Apply edge updates *)
-    List.iter
+      List.iter
       (fun (k, v_opt) ->
         match v_opt with
         | Some successors ->
@@ -1041,7 +1056,7 @@ let fixpoint ~name ~(init : ('k, unit) t) ~(edges : ('k, 'k list) t) () :
       edges_entries;
 
     (* Apply init updates *)
-    List.iter
+      List.iter
       (fun (k, v_opt) ->
         match v_opt with
         | Some () -> Hashtbl.replace roots k ()
@@ -1057,7 +1072,7 @@ let fixpoint ~name ~(init : ('k, unit) t) ~(edges : ('k, 'k list) t) () :
        let new_current = recompute_all () in
 
        (* Find removed entries *)
-       Hashtbl.iter
+        Hashtbl.iter
          (fun k () ->
            if not (Hashtbl.mem new_current k) then
              output_entries := (k, None) :: !output_entries)
@@ -1091,7 +1106,7 @@ let fixpoint ~name ~(init : ('k, unit) t) ~(edges : ('k, 'k list) t) () :
          match Hashtbl.find_opt edge_map k with
          | None -> ()
          | Some successors ->
-           List.iter
+      List.iter
              (fun succ ->
                if not (Hashtbl.mem current succ) then (
                  Hashtbl.replace current succ ();
@@ -1167,3 +1182,4 @@ let fixpoint ~name ~(init : ('k, unit) t) ~(edges : ('k, 'k list) t) () :
 let to_mermaid () = Registry.to_mermaid ()
 let print_stats () = Registry.print_stats ()
 let reset () = Registry.clear ()
+let reset_stats () = Registry.reset_stats ()

analysis/reactive/src/Reactive.mli

@@ -164,3 +164,6 @@ val print_stats : unit -> unit
 
 val reset : unit -> unit
 (** Clear all registered nodes (for tests) *)
+
+val reset_stats : unit -> unit
+(** Reset all node statistics to zero (keeps nodes intact) *)

analysis/reactive/src/ReactiveFileCollection.ml

@@ -82,6 +82,19 @@ let remove t path =
   Hashtbl.remove t.internal.cache path;
   emit t (Reactive.Remove path)
 
+(** Remove multiple files as a batch *)
+let remove_batch t paths =
+  let entries =
+    paths
+    |> List.filter_map (fun path ->
+           if Hashtbl.mem t.internal.cache path then (
+             Hashtbl.remove t.internal.cache path;
+             Some (path, None))
+           else None)
+  in
+  if entries <> [] then emit t (Reactive.Batch entries);
+  List.length entries
+
 (** Clear all cached data *)
 let clear t = Hashtbl.clear t.internal.cache
 

analysis/reactive/src/ReactiveFileCollection.mli

@@ -54,6 +54,10 @@ val process_if_changed : ('raw, 'v) t -> string -> bool
 val remove : ('raw, 'v) t -> string -> unit
 (** Remove a file from the collection. *)
 
+val remove_batch : ('raw, 'v) t -> string list -> int
+(** Remove multiple files as a batch. Returns the number of files removed.
+    More efficient than calling [remove] multiple times. *)
+
 (** {1 Cache Management} *)
 
 val invalidate : ('raw, 'v) t -> string -> unit

analysis/reanalyze/src/Cli.ml

@@ -34,5 +34,8 @@ let reactive = ref false
 (* number of analysis runs (for benchmarking reactive mode) *)
 let runs = ref 1
 
+(* number of files to churn (remove/re-add) between runs for incremental testing *)
+let churn = ref 0
+
 (* output mermaid diagram of reactive pipeline *)
 let mermaid = ref false

analysis/reanalyze/src/Log_.ml

@@ -197,6 +197,7 @@ module Stats = struct
   let issues = ref []
   let addIssue (issue : Issue.t) = issues := issue :: !issues
   let clear () = issues := []
+  let get_issue_count () = List.length !issues
 
   let getSortedIssues () =
     let counters2 = Hashtbl.create 1 in