Changes to support counting hash collisions.

ExceLint/ClusterModelBuilder.fs

@@ -448,6 +448,7 @@
                             weights = new Dictionary<AST.Address,double>();
                             clustering = m.ClusteringAtKnee;
                             fixes = [||];
+                            escapehatch = None;
                         }
                     )
                 else

ExceLint/CommonTypes.fs

@@ -152,6 +152,7 @@
             weights: Weights;
             clustering: Clustering;
             fixes: ProposedFix[];
+            escapehatch: obj option;
         }
 
         type Analysis =

ExceLint/EntropyModelBuilder.fs

@@ -325,6 +325,7 @@
                                 weights = EntropyModel.Weights fixes;    // this just returns entropy delta for now
                                 clustering = CommonFunctions.ToMutableClustering (EntropyModel.RankingToClusters fixes);
                                 fixes = [||];
+                                escapehatch = None;
                             }
                         )
                     else

ExceLint/EntropyModelBuilder2.fs

@@ -689,6 +689,7 @@
                                 cutoff_idx = m.Cutoff;
                                 weights = EntropyModel2.Weights fixeses;    // this just returns entropy delta for now
                                 clustering = CommonFunctions.ToMutableClustering (EntropyModel2.RankingToClusters fixes);
+                                escapehatch = Some (m :> obj);
                             }
                         )
                     else

ExceLint/ErrorModel.fs

@@ -151,6 +151,8 @@
             | Cluster a -> a.clustering
             | _ -> failwith "Not valid for non-cluster analysis."
 
+        member self.Analysis = analysis
+
         member self.inspectSelectorFor(addr: AST.Address, sel: Scope.Selector, dag: Depends.DAG) : KeyValuePair<AST.Address,(string*Countable)[]>[] =
             let sID = sel.id addr dag
 

ExceLint/ExceLint.fsproj

@@ -100,7 +100,6 @@
     </ProjectReference>
     <Reference Include="MathNet.Numerics">
       <HintPath>..\packages\MathNet.Numerics.3.20.0\lib\net40\MathNet.Numerics.dll</HintPath>
-      <Private>True</Private>
     </Reference>
     <Reference Include="Microsoft.Office.Interop.Excel, Version=15.0.0.0, Culture=neutral, PublicKeyToken=71e9bce111e9429c" />
     <Reference Include="mscorlib" />

ExceLint/OldClusterModel.fs

@@ -373,6 +373,7 @@
                             weights = new Dictionary<AST.Address,double>();
                             clustering = m.ClusteringAtKnee;
                             fixes = [||];
+                            escapehatch = None;
                         }
                     )
                 else

ExceLint/Vector.fs

@@ -846,6 +846,7 @@
                 (typeof<ShallowInputVectorMixedFullCVectorResultantNotOSI>.Name,
                     { enabled = false; kind = ConfigKind.Feature; runner = ShallowInputVectorMixedFullCVectorResultantNotOSI.run } )
 
+        // THIS IS THE IMPORTANT ONE USED IN THE PAPER
         type ShallowInputVectorMixedFullCVectorResultantOSI() =
             inherit BaseFeature()
             static member run(cell: AST.Address)(dag: DAG) : Countable =
@@ -862,6 +863,28 @@
             static member capability : string*Capability =
                 (typeof<ShallowInputVectorMixedFullCVectorResultantOSI>.Name,
                     { enabled = false; kind = ConfigKind.Feature; runner = ShallowInputVectorMixedFullCVectorResultantOSI.run } )
+            // THIS FUNCTION GETS THE VECTOR SET FOR THE ANALYSIS ABOVE
+            static member getPaperVectors(cell: AST.Address)(dag: DAG) : Countable[] =
+                let isMixed = true
+                let isTransitive = false
+                let isFormula = true
+                let isOffSheetInsensitive = true
+                let includeConstant = true
+                let includeLoc = true
+                let keepConstantValues = KeepConstantValue.No
+                let rebase_f = relativeToTail
+                let constant_f = makeConstantVectorsFromConstants keepConstantValues
+                let vs = getVectors cell dag (makeVector isMixed includeConstant) constant_f isTransitive isFormula
+                let rebased_vs = vs |> Array.map (fun v -> rebase_f v dag isOffSheetInsensitive includeLoc)
+                let rvarrs =
+                    rebased_vs |>
+                    Array.map (fun v ->
+                        match v with
+                        | ConstantWithLoc(x,y,z,x',y',z',c) -> Countable.FullCVectorResultant((double x, double y, double z, double x', double y', double z', double c))   
+                        | _ -> failwith "this should never happen"
+                    ) |>
+                    Array.map (fun v -> v.LocationFree)
+                rvarrs
 
         type ShallowOutputVectorMixedL2NormSum() =
             inherit BaseFeature()

ExceLintFileFormats/ExceLintStats.cs

@@ -115,5 +115,6 @@ public class ExceLintStatsRow
         public bool OptWeightConditionSetSz { get; set; }
         public double ExceLintJaccardDistance { get; set; }
         public int ExceLintDeltaK { get; set; }
+        public int Collisions { get; set; }
     }
 }

ExceLintRunner/ExceLintRunner.fsproj

@@ -26,7 +26,7 @@
     <PlatformTarget>AnyCPU</PlatformTarget>
     <DocumentationFile>bin\Debug\ExceLintRunner.XML</DocumentationFile>
     <Prefer32Bit>true</Prefer32Bit>
-    <StartArguments>"C:\Users\Daniel Barowy\Documents\Visual Studio 2017\Projects\ExceLintPaper\data\spreadsheets\custodes" "C:\Users\Daniel Barowy\Desktop\benchmarks-2017-11-16" "C:\Users\Daniel Barowy\Documents\Visual Studio 2017\Projects\ExceLintPaper\analysis\dissertation_annotations\true_ref_bugs.csv" "C:\Users\Daniel Barowy\Documents\Visual Studio 2017\Projects\ExceLintPaper\data\analyses\CUSTODES\smell_detection_result.csv" "C:\ProgramData\Oracle\Java\javapath\java.exe" "C:\Users\Daniel Barowy\Documents\Visual Studio 2017\Projects\ExceLintPaper\data\analyses\CUSTODES2\cc2.jar" -verbose -cluster -noexit -noshuffle</StartArguments>
+    <StartArguments>"C:\Users\Daniel Barowy\Documents\Visual Studio 2017\Projects\ExceLintPaper\data\spreadsheets\custodes" "C:\Users\Daniel Barowy\Desktop\benchmarks-2017-11-16" "C:\Users\Daniel Barowy\Documents\Visual Studio 2017\Projects\ExceLintPaper\analysis\dissertation_annotations\true_ref_bugs.csv" "C:\Users\Daniel Barowy\Documents\Visual Studio 2017\Projects\ExceLintPaper\data\analyses\CUSTODES\smell_detection_result.csv" "C:\ProgramData\Oracle\Java\javapath\java.exe" "C:\Users\Daniel Barowy\Documents\Visual Studio 2017\Projects\ExceLintPaper\data\analyses\CUSTODES2\cc2.jar" -verbose -cluster -noexit -noshuffle -nocustodes</StartArguments>
   </PropertyGroup>
   <PropertyGroup Condition=" '$(Configuration)|$(Platform)' == 'Release|AnyCPU' ">
     <DebugType>pdbonly</DebugType>

ExceLintRunner/Program.fs

@@ -36,6 +36,8 @@ open MathNet.Numerics.Distributions
         custodes_time: int64;
         excelint_jaccard: double;
         excelint_delta_k: int;
+        cells: int;
+        collisions: int;
     }
 
     let hs_difference<'a>(hs1: HashSet<'a>)(hs2: HashSet<'a>) : HashSet<'a> =
@@ -227,6 +229,8 @@ open MathNet.Numerics.Distributions
             else
                 Array.map (fun (kvp: KeyValuePair<AST.Address,double>) -> kvp.Value) (model.ranking()) |> Array.min
 
+        assert (model.AllCells.Count = stats.cells)
+
         // write stats
         let row = ExceLintStatsRow()
         row.BenchmarkName <- stats.shortname
@@ -284,6 +288,7 @@ open MathNet.Numerics.Distributions
         row.OptWeightConditionSetSz <- config.FeatureConf.IsEnabledOptWeightConditioningSetSize
         row.ExceLintJaccardDistance <- stats.excelint_jaccard
         row.ExceLintDeltaK <- stats.excelint_delta_k
+        row.Collisions <- stats.collisions
 
         csv.WriteRow row
 
@@ -397,6 +402,57 @@ open MathNet.Numerics.Distributions
                 i <- i + 1
         i
 
+    type SoundnessCount = { ncells: int; nnomatch: int; }
+
+    let soundness_count(model_opt: ErrorModel option)(dag: Depends.DAG) : SoundnessCount =
+        // get analysis base
+        let cells = match model_opt with | Some m -> m.AllCells | None -> failwith "does not apply"
+
+        // save set of cells that hashes to the same fingerprint
+        let fd = new Dict<Countable,HashSet<AST.Address>>()
+
+        // save all vectors for cells at given address
+        let addrv = new Dict<AST.Address,Countable[]>()
+
+        // for each cell, get vectors and fingerprint
+        cells |>
+        Seq.iter (fun cell ->
+            let vs = Vector.ShallowInputVectorMixedFullCVectorResultantOSI.getPaperVectors cell dag |> Array.map (fun v -> v)
+            let fingerprint = (Vector.ShallowInputVectorMixedFullCVectorResultantOSI.run cell dag).LocationFree
+
+            // save vectors
+            addrv.Add(cell, vs)
+
+            // init hashset
+            if not (fd.ContainsKey(fingerprint)) then
+                fd.Add(fingerprint, new HashSet<AST.Address>())
+
+            // get set
+            let hs = fd.[fingerprint]
+
+            // add to set
+            hs.Add cell |> ignore
+        )
+
+        // for each fingerprint, count
+        // how many of those cells' vector sets do not match
+        let mutable nomatch = 0
+        fd |>
+        Seq.iter (fun (kvp: KeyValuePair<Countable,HashSet<AST.Address>>) -> 
+            let addrs = kvp.Value |> Seq.toArray
+            if addrs.Length > 1 then
+                // get the first set of vectors
+                let vs0 = addrv.[addrs.[0]] |> Set.ofArray
+                for addr in addrs do
+                    // get the second set of vectors
+                    let vsi = addrv.[addr] |> Set.ofArray
+                    if vs0 <> vsi then
+                        nomatch <- nomatch + 1
+        )
+
+        { ncells = Seq.length cells; nnomatch = nomatch; }
+
+
     let analyze (file: String)(app: Application)(config: Args.Config)(etruth: ExceLintGroundTruth)(ctruth: CUSTODES.GroundTruth)(csv: ExceLintStats)(debug_csv: DebugInfo) =
         let shortf = (System.IO.Path.GetFileName file)
 
@@ -410,6 +466,8 @@ open MathNet.Numerics.Distributions
 
         let model_opt = ExceLint.ModelBuilder.analyze (app.XLApplication()) config.FeatureConf graph (config.alpha) (Depends.Progress.NOPProgress())
 
+        let scount = soundness_count model_opt graph
+
         let (jdist,delta_k) =
             match model_opt with
             | Some model ->
@@ -479,7 +537,11 @@ open MathNet.Numerics.Distributions
                 let esz = Math.Min(excelint_flags.Count, model.Cutoff + 1)
                 assert (esz = excelint_true_ref_TP + excelint_true_ref_FP)
                 let csz = custodes_flags.Count
-                assert (csz = custodes_true_ref_TP + custodes_true_ref_FP)
+                // TODO: something funny happening here: 2017-11-16
+                //let foo1 = custodes_true_ref_TP
+                //let foo2 = custodes_true_ref_FP
+                //let foo3 = foo1 + foo2
+                //assert (csz = custodes_true_ref_TP + custodes_true_ref_FP)
 
                 let stats = {
                     shortname = shortf;
@@ -508,6 +570,8 @@ open MathNet.Numerics.Distributions
                     custodes_time = custodes_time;
                     excelint_jaccard = jdist;
                     excelint_delta_k = delta_k;
+                    cells = scount.ncells;
+                    collisions = scount.nnomatch;
                 }
 
                 // write to per-workbook CSV