[specs] put specs files operations in their own module

Summary:
Summary.ml defines both a bunch of types and how to use them and a
mechanism to save and store summaries on disk while maintaining a
complex in-memory cache of what's on disk. Make the distinction clear.

Reviewed By: ngorogiannis

Differential Revision: D16358869

fbshipit-source-id: 9d4c6cb77

infer/src/absint/NodePrinter.ml

@@ -13,7 +13,7 @@ open! IStd
 let new_session node =
   let pname = Procdesc.Node.get_proc_name node in
   let node_id = (Procdesc.Node.get_id node :> int) in
-  match Summary.get pname with
+  match Summary.OnDisk.get pname with
   | None ->
       0
   | Some summary ->

infer/src/backend/InferAnalyze.ml

@@ -13,7 +13,7 @@ module L = Logging
 
 let clear_caches () =
   Ondemand.clear_cache () ;
-  Summary.clear_cache () ;
+  Summary.OnDisk.clear_cache () ;
   Typ.Procname.SQLite.clear_cache ()
 
 
@@ -91,7 +91,7 @@ let main ~changed_files =
   register_active_checkers () ;
   if Config.reanalyze then (
     L.progress "Invalidating procedures to be reanalyzed@." ;
-    Summary.reset_all ~filter:(Lazy.force Filtering.procedures_filter) () ;
+    Summary.OnDisk.reset_all ~filter:(Lazy.force Filtering.procedures_filter) () ;
     L.progress "Done@." )
   else DB.Results_dir.clean_specs_dir () ;
   let n_all_source_files = ref 0 in

infer/src/backend/SpecsFiles.ml

@@ -50,7 +50,7 @@ let spec_files_from_cmdline () =
 let summary_iterator spec_files =
   let sorted_spec_files = List.sort ~compare:String.compare (spec_files ()) in
   let do_spec f fname =
-    match Summary.load_from_file (DB.filename_from_string fname) with
+    match Summary.OnDisk.load_from_file (DB.filename_from_string fname) with
     | None ->
         L.(die UserError) "Error: cannot open file %s@." fname
     | Some summary ->
@@ -65,5 +65,7 @@ let iter_from_config ~f = summary_iterator spec_files_from_cmdline f
 let iter ~f = summary_iterator load_specfiles f
 
 let delete pname =
-  let filename = Summary.specs_filename_of_procname pname |> DB.filename_to_string in
-  Unix.unlink filename ; Ondemand.remove_from_cache pname ; Summary.remove_from_cache pname
+  let filename = Summary.OnDisk.specs_filename_of_procname pname |> DB.filename_to_string in
+  Unix.unlink filename ;
+  Ondemand.remove_from_cache pname ;
+  Summary.OnDisk.remove_from_cache pname

infer/src/backend/Summary.ml

@@ -8,7 +8,6 @@
 
 open! IStd
 module F = Format
-open PolyVariantEqual
 
 module Stats = struct
   type t =
@@ -96,14 +95,6 @@ let get_err_log summary = summary.err_log
 
 let get_loc summary = (get_attributes summary).ProcAttributes.loc
 
-type cache = t Typ.Procname.Hash.t
-
-let cache : cache = Typ.Procname.Hash.create 128
-
-let clear_cache () = Typ.Procname.Hash.clear cache
-
-let remove_from_cache pname = Typ.Procname.Hash.remove cache pname
-
 let pp_errlog fmt err_log =
   F.fprintf fmt "ERRORS: @[<h>%a@]@\n%!" Errlog.pp_errors err_log ;
   F.fprintf fmt "WARNINGS: @[<h>%a@]" Errlog.pp_warnings err_log
@@ -139,6 +130,17 @@ let pp_html source fmt summary =
   F.fprintf fmt "</LISTING>@\n"
 
 
+module OnDisk = struct
+  open PolyVariantEqual
+
+  type cache = t Typ.Procname.Hash.t
+
+  let cache : cache = Typ.Procname.Hash.create 128
+
+  let clear_cache () = Typ.Procname.Hash.clear cache
+
+  let remove_from_cache pname = Typ.Procname.Hash.remove cache pname
+
   (** Add the summary to the table for the given function *)
   let add (proc_name : Typ.Procname.t) (summary : t) : unit =
     Typ.Procname.Hash.replace cache proc_name summary
@@ -165,7 +167,9 @@ let specs_models_filename pname =
     DB.filename_from_string (Filename.concat Config.models_dir (specs_filename pname))
 
 
-let has_model pname = Sys.file_exists (DB.filename_to_string (specs_models_filename pname)) = `Yes
+  let has_model pname =
+    Sys.file_exists (DB.filename_to_string (specs_models_filename pname)) = `Yes
+
 
   let summary_serializer : t Serialization.serializer =
     Serialization.create_serializer Serialization.Key.summary
@@ -203,8 +207,11 @@ let load_summary_to_spec_table =
 
 
   let get proc_name =
-  try Some (Typ.Procname.Hash.find cache proc_name)
-  with Caml.Not_found -> load_summary_to_spec_table proc_name
+    match Typ.Procname.Hash.find cache proc_name with
+    | summary ->
+        Some summary
+    | exception Caml.Not_found ->
+        load_summary_to_spec_table proc_name
 
 
   (** Check if the procedure is from a library:
@@ -239,7 +246,7 @@ let store (summ : t) =
       ~data:final_summary
 
 
-let init_summary proc_desc =
+  let reset proc_desc =
     let summary =
       { sessions= ref 0
       ; payloads= Payloads.empty
@@ -258,11 +265,8 @@ let dummy =
       ProcAttributes.default (SourceFile.invalid __FILE__) Typ.Procname.empty_block
     in
     let dummy_proc_desc = Procdesc.from_proc_attributes dummy_attributes in
-  init_summary dummy_proc_desc
-
+    reset dummy_proc_desc
 
-(** Reset a summary rebuilding the dependents and preserving the proc attributes if present. *)
-let reset proc_desc = init_summary proc_desc
 
   let reset_all ~filter () =
     let reset proc_name =
@@ -273,7 +277,7 @@ let reset_all ~filter () =
              Serialization.write_to_file summary_serializer filename ~data:blank_summary )
     in
     Procedures.get_all ~filter () |> List.iter ~f:reset
-
+end
 
 module SummaryValue = struct
   type nonrec t = t option

infer/src/backend/Summary.mli

@@ -51,22 +51,6 @@ type t =
 
 val poly_fields : t PolyFields.t
 
-val dummy : t
-(** dummy summary for testing *)
-
-val has_model : Typ.Procname.t -> bool
-(** Check if a summary for a given procedure exists in the models directory *)
-
-val clear_cache : unit -> unit
-(** Remove all the elements from the cache of summaries *)
-
-val remove_from_cache : Typ.Procname.t -> unit
-(** Remove an element from the cache of summaries. Contrast to reset which re-initializes a summary
-    keeping the same Procdesc and updates the cache accordingly. *)
-
-val get : Typ.Procname.t -> t option
-(** Return the summary option for the procedure name *)
-
 val get_proc_name : t -> Typ.Procname.t
 (** Get the procedure name *)
 
@@ -88,6 +72,26 @@ val get_signature : t -> string
 val get_status : t -> Status.t
 (** Return the status (active v.s. inactive) of a procedure summary *)
 
+val pp_html : SourceFile.t -> Format.formatter -> t -> unit
+(** Print the summary in html format *)
+
+val pp_text : Format.formatter -> t -> unit
+(** Print the summary in text format *)
+
+module OnDisk : sig
+  val has_model : Typ.Procname.t -> bool
+  (** Check if a summary for a given procedure exists in the models directory *)
+
+  val clear_cache : unit -> unit
+  (** Remove all the elements from the cache of summaries *)
+
+  val remove_from_cache : Typ.Procname.t -> unit
+  (** Remove an element from the cache of summaries. Contrast to reset which re-initializes a summary
+    keeping the same Procdesc and updates the cache accordingly. *)
+
+  val get : Typ.Procname.t -> t option
+  (** Return the summary option for the procedure name *)
+
   val reset : Procdesc.t -> t
   (** Reset a summary rebuilding the dependents and preserving the proc attributes if present. *)
 
@@ -97,12 +101,6 @@ val specs_filename_of_procname : Typ.Procname.t -> DB.filename
   val load_from_file : DB.filename -> t option
   (** Load procedure summary from the given file *)
 
-val pp_html : SourceFile.t -> Format.formatter -> t -> unit
-(** Print the summary in html format *)
-
-val pp_text : Format.formatter -> t -> unit
-(** Print the summary in text format *)
-
   val proc_resolve_attributes : Typ.Procname.t -> ProcAttributes.t option
   (** Try to find the attributes for a defined proc.
     First look at specs (to get attributes computed by analysis)
@@ -119,6 +117,10 @@ val store : t -> unit
 
   val reset_all : filter:Filtering.procedures_filter -> unit -> unit
 
+  val dummy : t
+  (** dummy summary for testing *)
+end
+
 module SummaryValue : Memcached.Value with type t = t option
 
 module SummaryServer : Memcached.Server with module Value = SummaryValue

infer/src/backend/callbacks.ml

@@ -42,7 +42,7 @@ let get_procedure_definition exe_env proc_name =
   Procdesc.load proc_name
   |> Option.map ~f:(fun proc_desc ->
          let tenv = Exe_env.get_tenv exe_env proc_name in
-         (tenv, Summary.reset proc_desc) )
+         (tenv, Summary.OnDisk.reset proc_desc) )
 
 
 (** Invoke all registered procedure callbacks on the given procedure. *)

infer/src/backend/ondemand.ml

@@ -41,7 +41,7 @@ let is_active, add_active, remove_active =
 
 
 let already_analyzed proc_name =
-  match Summary.get proc_name with
+  match Summary.OnDisk.get proc_name with
   | Some summary ->
       Summary.(Status.is_analyzed (get_status summary))
   | None ->
@@ -57,7 +57,7 @@ let should_be_analyzed proc_attributes =
 
 let get_proc_attr proc_name =
   IList.force_until_first_some
-    [lazy (Summary.proc_resolve_attributes proc_name); lazy (Topl.get_proc_attr proc_name)]
+    [lazy (Summary.OnDisk.proc_resolve_attributes proc_name); lazy (Topl.get_proc_attr proc_name)]
 
 
 let procedure_should_be_analyzed proc_name =
@@ -158,12 +158,12 @@ let run_proc_analysis ~caller_pdesc callee_pdesc =
       Typ.Procname.pp callee_pname ;
   let preprocess () =
     incr nesting ;
-    let initial_callee_summary = Summary.reset callee_pdesc in
+    let initial_callee_summary = Summary.OnDisk.reset callee_pdesc in
     add_active callee_pname ; initial_callee_summary
   in
   let postprocess summary =
     decr nesting ;
-    Summary.store summary ;
+    Summary.OnDisk.store summary ;
     remove_active callee_pname ;
     Printer.write_proc_html callee_pdesc ;
     log_elapsed_time () ;
@@ -180,7 +180,10 @@ let run_proc_analysis ~caller_pdesc callee_pdesc =
       {summary.payloads with biabduction}
     in
     let new_summary = {summary with stats; payloads} in
-    Summary.store new_summary ; remove_active callee_pname ; log_elapsed_time () ; new_summary
+    Summary.OnDisk.store new_summary ;
+    remove_active callee_pname ;
+    log_elapsed_time () ;
+    new_summary
   in
   let old_state = save_global_state () in
   let initial_callee_summary = preprocess () in
@@ -293,7 +296,7 @@ let register_callee ?caller_summary callee_pname =
 let get_proc_desc callee_pname =
   IList.force_until_first_some
     [ lazy (Procdesc.load callee_pname)
-    ; lazy (Option.map ~f:Summary.get_proc_desc (Summary.get callee_pname))
+    ; lazy (Option.map ~f:Summary.get_proc_desc (Summary.OnDisk.get callee_pname))
     ; lazy (Topl.get_proc_desc callee_pname) ]
 
 
@@ -318,10 +321,10 @@ let analyze_proc ?caller_summary callee_pname callee_pdesc should_be_analyzed =
                          callee_pdesc)
                   , true )
               | None ->
-                  (Summary.get callee_pname, true)
+                  (Summary.OnDisk.get callee_pname, true)
             else (
               EventLogger.log_skipped_pname (F.asprintf "%a" Typ.Procname.pp callee_pname) ;
-              (Summary.get callee_pname, true) )
+              (Summary.OnDisk.get callee_pname, true) )
       in
       if update_memcached then memcache_set callee_pname callee_summary_option ;
       Typ.Procname.Hash.add cache callee_pname callee_summary_option ;

infer/src/backend/preanal.ml

@@ -209,7 +209,7 @@ let do_funptr_sub summary tenv =
 
 
 let do_preanalysis pdesc tenv =
-  let summary = Summary.reset pdesc in
+  let summary = Summary.OnDisk.reset pdesc in
   if
     Config.function_pointer_specialization
     && not (Typ.Procname.is_java (Procdesc.get_proc_name pdesc))

infer/src/backend/printer.ml

@@ -69,7 +69,7 @@ let curr_html_formatter = ref F.std_formatter
 
 (** Return true if the node was visited during analysis *)
 let is_visited node =
-  match Summary.get (Procdesc.Node.get_proc_name node) with
+  match Summary.OnDisk.get (Procdesc.Node.get_proc_name node) with
   | None ->
       false
   | Some summary ->
@@ -182,7 +182,7 @@ end = struct
          [])
       linenum ;
     pp_node_link_seq [] ~description:true fmt nodes ;
-    ( match Summary.get pname with
+    ( match Summary.OnDisk.get pname with
     | None ->
         ()
     | Some summary ->
@@ -230,7 +230,7 @@ end = struct
       Hashtbl.replace table_nodes_at_linenum lnum (n :: curr_nodes)
     in
     List.iter ~f:process_node (Procdesc.get_nodes proc_desc) ;
-    match Summary.get proc_name with
+    match Summary.OnDisk.get proc_name with
     | None ->
         ()
     | Some summary ->
@@ -259,7 +259,7 @@ end = struct
               | Procdesc.Node.Start_node ->
                   let proc_name = Procdesc.Node.get_proc_name n in
                   let proc_name_escaped = Escape.escape_xml (Typ.Procname.to_string proc_name) in
-                  if Summary.get proc_name |> Option.is_some then (
+                  if Summary.OnDisk.get proc_name |> Option.is_some then (
                     F.pp_print_char fmt ' ' ;
                     let label = F.asprintf "summary for %s" proc_name_escaped in
                     Io_infer.Html.pp_proc_link [fname_encoding] proc_name fmt label )

infer/src/backend/reporting.ml

@@ -59,7 +59,7 @@ let log_issue_from_summary severity summary ~node ~session ~loc ~ltr ?extras exn
 
 let log_issue_deprecated_using_state severity proc_name ?node ?loc ?ltr exn =
   if !BiabductionConfig.footprint then
-    match Summary.get proc_name with
+    match Summary.OnDisk.get proc_name with
     | Some summary ->
         let node =
           let node = match node with None -> State.get_node_exn () | Some node -> node in

infer/src/biabduction/SymExec.ml

@@ -519,7 +519,7 @@ let method_exists right_proc_name methods =
     | Some attrs ->
         attrs.ProcAttributes.is_defined
     | None ->
-        Summary.has_model right_proc_name
+        Summary.OnDisk.has_model right_proc_name
 
 
 let resolve_method tenv class_name proc_name =
@@ -1125,7 +1125,7 @@ let resolve_and_analyze_clang current_summary tenv prop_r n_actual_params callee
     (* to be extended to other methods *)
   then
     try
-      let has_clang_model = Summary.has_model callee_pname in
+      let has_clang_model = Summary.OnDisk.has_model callee_pname in
       let resolve_and_analyze_result =
         resolve_and_analyze tenv ~caller_summary:current_summary ~has_clang_model prop_r
           n_actual_params callee_pname call_flags
@@ -1721,7 +1721,7 @@ and unknown_or_scan_call ~is_scan ~reason ret_typ ret_annots
     let callee_loc_opt =
       Option.map
         ~f:(fun attributes -> attributes.ProcAttributes.loc)
-        (Summary.proc_resolve_attributes callee_pname)
+        (Summary.OnDisk.proc_resolve_attributes callee_pname)
     in
     let skip_path = Paths.Path.add_skipped_call path callee_pname reason callee_loc_opt in
     [(prop_with_undef_attr, skip_path)]
@@ -1763,7 +1763,7 @@ and check_variadic_sentinel ?(fails_on_nil = false) n_formals (sentinel, null_po
 
 
 and check_variadic_sentinel_if_present ({Builtin.prop_; path; proc_name} as builtin_args) =
-  match Summary.proc_resolve_attributes proc_name with
+  match Summary.OnDisk.proc_resolve_attributes proc_name with
   | None ->
       [(prop_, path)]
   | Some callee_attributes -> (

infer/src/biabduction/Tabulation.ml

@@ -834,7 +834,7 @@ let prop_set_exn tenv pname prop se_exn =
 
 (** Include a subtrace for a procedure call if the callee is not a model. *)
 let include_subtrace callee_pname =
-  match Summary.proc_resolve_attributes callee_pname with
+  match Summary.OnDisk.proc_resolve_attributes callee_pname with
   | Some attrs ->
       (not attrs.ProcAttributes.is_model)
       && SourceFile.is_under_project_root attrs.ProcAttributes.loc.Location.file

infer/src/checkers/NullabilityPreanalysis.ml

@@ -96,9 +96,8 @@ let analysis cfg tenv =
   let f proc_name pdesc domain =
     if Procdesc.is_defined pdesc && Typ.Procname.is_constructor proc_name then
       match
-        FieldsAssignedInConstructorsChecker.compute_post
-          (ProcData.make (Summary.reset pdesc) tenv (Ident.Hash.create 10))
-          ~initial
+        FieldsAssignedInConstructorsChecker.compute_post ~initial
+          (ProcData.make (Summary.OnDisk.reset pdesc) tenv (Ident.Hash.create 10))
       with
       | Some new_domain ->
           FieldsAssignedInConstructors.union new_domain domain

infer/src/checkers/annotationReachability.ml

@@ -51,7 +51,8 @@ let is_allocator tenv pname =
 
 let check_attributes check tenv pname =
   PatternMatch.check_class_attributes check tenv pname
-  || Annotations.pname_has_return_annot pname ~attrs_of_pname:Summary.proc_resolve_attributes check
+  || Annotations.pname_has_return_annot pname
+       ~attrs_of_pname:Summary.OnDisk.proc_resolve_attributes check
 
 
 let method_overrides is_annotated tenv pname =

infer/src/clang/cFrontend_decl.ml

@@ -41,7 +41,7 @@ let protect ~f ~recover ~pp_context (trans_unit_ctx : CFrontend_config.translati
 
 
 module CFrontend_decl_funct (T : CModule_type.CTranslation) : CModule_type.CFrontend = struct
-  let model_exists procname = (not Config.models_mode) && Summary.has_model procname
+  let model_exists procname = (not Config.models_mode) && Summary.OnDisk.has_model procname
 
   (** Translates the method/function's body into nodes of the cfg. *)
   let add_method ?(is_destructor_wrapper = false) trans_unit_ctx tenv cfg class_decl_opt procname

infer/src/concurrency/RacerD.ml

@@ -524,7 +524,7 @@ let analyze_procedure {Callbacks.exe_env; summary} =
     let initial =
       let threads =
         if
-          runs_on_ui_thread ~attrs_of_pname:Summary.proc_resolve_attributes tenv proc_desc
+          runs_on_ui_thread ~attrs_of_pname:Summary.OnDisk.proc_resolve_attributes tenv proc_desc
           |> Option.is_some
           || is_thread_confined_method tenv proc_desc
         then ThreadsDomain.AnyThreadButSelf

infer/src/concurrency/RacerDModels.ml

@@ -155,7 +155,8 @@ let should_skip =
 
 
 let has_return_annot predicate pn =
-  Annotations.pname_has_return_annot pn ~attrs_of_pname:Summary.proc_resolve_attributes predicate
+  Annotations.pname_has_return_annot pn ~attrs_of_pname:Summary.OnDisk.proc_resolve_attributes
+    predicate
 
 
 let is_functional pname =
@@ -342,8 +343,8 @@ let is_thread_safe_class pname tenv =
 
 
 let is_thread_safe_method pname tenv =
-  find_method_or_override_annotated ~attrs_of_pname:Summary.proc_resolve_attributes is_thread_safe
-    pname tenv
+  find_method_or_override_annotated ~attrs_of_pname:Summary.OnDisk.proc_resolve_attributes
+    is_thread_safe pname tenv
   |> Option.is_some
 
 

infer/src/concurrency/starvation.ml

@@ -159,8 +159,8 @@ let analyze_procedure {Callbacks.exe_env; summary} =
         StarvationDomain.acquire tenv StarvationDomain.bottom loc (Option.to_list lock)
     in
     let initial =
-      ConcurrencyModels.runs_on_ui_thread ~attrs_of_pname:Summary.proc_resolve_attributes tenv
-        proc_desc
+      ConcurrencyModels.runs_on_ui_thread ~attrs_of_pname:Summary.OnDisk.proc_resolve_attributes
+        tenv proc_desc
       |> Option.value_map ~default:initial ~f:(StarvationDomain.set_on_ui_thread initial loc)
     in
     let filter_blocks =

infer/src/nullsafe/NullabilityCheck.ml

@@ -41,7 +41,7 @@ module TransferFunctions (CFG : ProcCfg.S) = struct
           ~f:(fun attributes ->
             ClangMethodKind.equal attributes.ProcAttributes.clang_method_kind
               ClangMethodKind.CPP_INSTANCE )
-          (Summary.proc_resolve_attributes callee_pname)
+          (Summary.OnDisk.proc_resolve_attributes callee_pname)
 
 
   let is_objc_instance_method callee_pname =
@@ -49,7 +49,7 @@ module TransferFunctions (CFG : ProcCfg.S) = struct
       ~f:(fun attributes ->
         ClangMethodKind.equal attributes.ProcAttributes.clang_method_kind
           ClangMethodKind.OBJC_INSTANCE )
-      (Summary.proc_resolve_attributes callee_pname)
+      (Summary.OnDisk.proc_resolve_attributes callee_pname)
 
 
   let is_blacklisted_method : Typ.Procname.t -> bool =
@@ -91,7 +91,7 @@ module TransferFunctions (CFG : ProcCfg.S) = struct
   let lookup_local_attributes = function
     | Typ.Procname.Java _ as pname ->
         (* Looking up the attribute according to the classpath *)
-        Summary.proc_resolve_attributes pname
+        Summary.OnDisk.proc_resolve_attributes pname
     | pname ->
         (* Looking up the attributes locally, i.e. either from the file of from the includes *)
         Option.map ~f:Procdesc.get_attributes (Ondemand.get_proc_desc pname)

infer/src/nullsafe/typeCheck.ml

@@ -754,7 +754,7 @@ let typecheck_instr tenv calls_this checks (node : Procdesc.Node.t) idenv curr_p
         in
         let resolved_ret_ =
           let ret_ia, ret_typ = callee_annotated_signature.AnnotatedSignature.ret in
-          let is_library = Summary.proc_is_library callee_attributes in
+          let is_library = Summary.OnDisk.proc_is_library callee_attributes in
           let origin =
             TypeOrigin.Proc
               { TypeOrigin.pname= callee_pname

infer/src/unit/TaintTests.ml

@@ -154,7 +154,7 @@ let tests =
     ; ( "sink without source not tracked"
       , [assign_to_non_source "ret_id"; call_sink "ret_id"; assert_empty] ) ]
     |> TestInterpreter.create_tests ~pp_opt:pp_sparse
-         {formal_map= FormalMap.empty; summary= Summary.dummy}
+         {formal_map= FormalMap.empty; summary= Summary.OnDisk.dummy}
          ~initial:(MockTaintAnalysis.Domain.bottom, Bindings.empty)
   in
   "taint_test_suite" >::: test_list

infer/src/unit/analyzerTester.ml

@@ -243,7 +243,7 @@ struct
     let exit_node = create_node Procdesc.Node.Exit_node [] in
     set_succs last_node [exit_node] ~exn_handlers:no_exn_handlers ;
     Procdesc.set_exit_node pdesc exit_node ;
-    (Summary.reset pdesc, assert_map)
+    (Summary.OnDisk.reset pdesc, assert_map)
 
 
   let create_test test_program extras ~initial pp_opt test_pname _ =