X-Git-Url: https://git.xandkar.net/?p=dups.git;a=blobdiff_plain;f=dups.ml;h=ebabc8af52d2a2bb54f21d30400c21f0abfa01ef;hp=8ff8874b7cf84ec76b965bda441e35cae1a4a22a;hb=7a0486beb770b58ec2d2d84ff6d7e5bfa8bcdbf9;hpb=a9a56d74122233b99d8d6401b1f50009a993cd0a

diff --git a/dups.ml b/dups.ml
index 8ff8874..ebabc8a 100644
--- a/dups.ml
+++ b/dups.ml
@@ -142,11 +142,19 @@ type output =
   | Stdout
   | Directory of string
 
+type opt =
+  { input  : input
+  ; output : output
+  ; ignore : Str.regexp option
+  ; sample : int
+  }
+
 type count =
   { considered  : int ref
   ; empty       : int ref
   ; ignored     : int ref
   ; unique_size : int ref
+  ; unique_sample : int ref
   ; hashed      : int ref
   }
 
@@ -190,7 +198,26 @@ let make_output_fun = function
         );
         close_out oc
 
-let main input output ignore =
+let sample path ~len =
+  let buf = Bytes.make len ' ' in
+  let ic = open_in_bin path in
+  let rec read pos len =
+    assert (len >= 0);
+    if len = 0 then
+      ()
+    else begin
+      let chunk_size = input ic buf pos len in
+      if chunk_size = 0 then (* EOF *)
+        ()
+      else
+        read (pos + chunk_size) (len - chunk_size)
+    end
+  in
+  read 0 len;
+  close_in ic;
+  Bytes.to_string buf
+
+let main {input; output; ignore; sample = sample_len} =
   let t0 = Sys.time () in
   let count =
     { considered  = ref 0
@@ -198,11 +225,13 @@ let main input output ignore =
     ; ignored     = ref 0
     ; unique_size = ref 0
     ; hashed      = ref 0
+    ; unique_sample     = ref 0
     }
   in
   let output = make_output_fun  output in
   let input  = make_input_stream input ignore count in
   let paths_by_size = Hashtbl.create 1_000_000 in
+  let paths_by_sample = Hashtbl.create 1_000_000 in
   let paths_by_digest = Hashtbl.create 1_000_000 in
   let process tbl path ~f =
     let key = f path in
@@ -224,14 +253,27 @@ let main input output ignore =
       if n > 1 then
         StrSet.iter
           (fun path ->
-            incr count.hashed;
-            process paths_by_digest path ~f:Digest.file
+            process paths_by_sample path ~f:(sample ~len:sample_len)
           )
           paths
       else
         incr count.unique_size;
     )
     paths_by_size;
+  Hashtbl.iter
+    (fun _ (n, paths) ->
+      (* Skip files with unique samples *)
+      if n > 1 then
+        StrSet.iter
+          (fun path ->
+            incr count.hashed;
+            process paths_by_digest path ~f:Digest.file
+          )
+          paths
+      else
+        incr count.unique_sample;
+    )
+    paths_by_sample;
   Hashtbl.iter (fun d (n, ps) -> if n > 1 then output d n ps) paths_by_digest;
   let t1 = Sys.time () in
   eprintf "Time                       : %f seconds\n%!" (t1 -. t0);
@@ -239,28 +281,26 @@ let main input output ignore =
   eprintf "Hashed                     : %d\n%!" !(count.hashed);
   eprintf "Skipped due to 0      size : %d\n%!" !(count.empty);
   eprintf "Skipped due to unique size : %d\n%!" !(count.unique_size);
+  eprintf "Skipped due to unique sample : %d\n%!" !(count.unique_sample);
   eprintf "Ignored due to regex match : %d\n%!" !(count.ignored)
 
-let () =
-  let input  = ref Stdin in
-  let output = ref Stdout in
-  let ignore = ref None in
-  let assert_file_exists path =
-    if Sys.file_exists path then
-      ()
-    else begin
-      eprintf "File does not exist: %S\n%!" path;
+let get_opt () : opt =
+  let assert_ test x msg =
+    if not (test x) then begin
+      eprintf "%s\n%!" msg;
       exit 1
     end
   in
+  let assert_file_exists path =
+    assert_ Sys.file_exists path (sprintf "File does not exist: %S" path)
+  in
   let assert_file_is_dir path =
-    if Sys.is_directory path then
-      ()
-    else begin
-      eprintf "File is not a directory: %S\n%!" path;
-      exit 1
-    end
+    assert_ Sys.is_directory path (sprintf "File is not a directory: %S" path)
   in
+  let input  = ref Stdin in
+  let output = ref Stdout in
+  let ignore = ref None in
+  let sample = ref 256 in
   let spec =
     [ ( "-out"
       , Arg.String (fun path ->
@@ -274,6 +314,10 @@ let () =
       , Arg.String (fun regexp -> ignore := Some (Str.regexp regexp))
       , " Ignore file paths which match this regexp pattern (see Str module)."
       )
+    ; ( "-sample"
+      , Arg.Set_int sample
+      , (sprintf " Byte size of file samples to use. Default: %d" !sample)
+      )
     ]
   in
   Arg.parse
@@ -288,4 +332,15 @@ let () =
           input := Directories (path :: paths)
     )
     "";
-  main !input !output !ignore
+  assert_
+    (fun x -> x > 0)
+    !sample
+    (sprintf "Sample size cannot be negative: %d" !sample);
+  { input  = !input
+  ; output = !output
+  ; ignore = !ignore
+  ; sample = !sample
+  }
+
+let () =
+  main (get_opt ())