Reduce memory usage by caching git objects on their sha hash

Author: Pascal van Buijtene

new-cli/GitVersion.Core.Libgit2Sharp/GitVersion.Core.Libgit2Sharp.csproj

@@ -21,6 +21,9 @@
       <Compile Include="..\..\src\GitVersion.LibGit2Sharp\Git\CommitCollection.cs">
         <Link>Git\CommitCollection.cs</Link>
       </Compile>
+      <Compile Include="..\..\src\GitVersion.LibGit2Sharp\Git\GitCache.cs">
+        <Link>Git\GitCache.cs</Link>
+      </Compile>
       <Compile Include="..\..\src\GitVersion.LibGit2Sharp\Git\GitObject.cs">
         <Link>Git\GitObject.cs</Link>
       </Compile>

src/GitVersion.LibGit2Sharp/Git/Branch.cs

@@ -10,16 +10,16 @@ internal sealed class Branch : IBranch
 
     private readonly LibGit2Sharp.Branch innerBranch;
 
-    internal Branch(LibGit2Sharp.Branch branch, LibGit2Sharp.Diff diff)
+    internal Branch(LibGit2Sharp.Branch branch, LibGit2Sharp.Diff diff, GitCache gitCache)
     {
         this.innerBranch = branch.NotNull();
         Name = new(branch.CanonicalName);
 
         var commit = this.innerBranch.Tip;
-        Tip = commit is null ? null : new Commit(commit, diff);
+        Tip = commit is null ? null : gitCache.GetOrCreate(commit, diff);
 
         var commits = this.innerBranch.Commits;
-        Commits = new CommitCollection(commits, diff);
+        Commits = new CommitCollection(commits, diff, gitCache);
     }
 
     public ReferenceName Name { get; }

src/GitVersion.LibGit2Sharp/Git/BranchCollection.cs

@@ -8,12 +8,14 @@ internal sealed class BranchCollection : IBranchCollection
     private readonly LibGit2Sharp.BranchCollection innerCollection;
     private readonly Lazy<IReadOnlyCollection<IBranch>> branches;
     private readonly Diff diff;
+    private readonly GitCache gitCache;
 
-    internal BranchCollection(LibGit2Sharp.BranchCollection collection, Diff diff)
+    internal BranchCollection(LibGit2Sharp.BranchCollection collection, Diff diff, GitCache gitCache)
     {
         this.innerCollection = collection.NotNull();
-        this.branches = new Lazy<IReadOnlyCollection<IBranch>>(() => [.. this.innerCollection.Select(branch => new Branch(branch, diff))]);
+        this.branches = new Lazy<IReadOnlyCollection<IBranch>>(() => [.. this.innerCollection.Select(branch => gitCache.GetOrCreate(branch, diff))]);
         this.diff = diff.NotNull();
+        this.gitCache = gitCache;
     }
 
     public IEnumerator<IBranch> GetEnumerator()
@@ -27,7 +29,7 @@ public IBranch? this[string name]
         {
             name = name.NotNull();
             var branch = this.innerCollection[name];
-            return branch is null ? null : new Branch(branch, this.diff);
+            return branch is null ? null : this.gitCache.GetOrCreate(branch, this.diff);
         }
     }
 

src/GitVersion.LibGit2Sharp/Git/Commit.cs

@@ -14,10 +14,10 @@ internal sealed class Commit : GitObject, ICommit
     private readonly LibGit2Sharp.Commit innerCommit;
     private readonly LibGit2Sharp.Diff repoDiff;
 
-    internal Commit(LibGit2Sharp.Commit innerCommit, LibGit2Sharp.Diff repoDiff) : base(innerCommit)
+    internal Commit(LibGit2Sharp.Commit innerCommit, LibGit2Sharp.Diff repoDiff, GitCache gitCache) : base(innerCommit)
     {
         this.innerCommit = innerCommit.NotNull();
-        this.parentsLazy = new(() => innerCommit.Parents.Select(parent => new Commit(parent, repoDiff)).ToList());
+        this.parentsLazy = new(() => innerCommit.Parents.Select(parent => gitCache.GetOrCreate(parent, repoDiff)).ToList());
         When = innerCommit.Committer.When;
         this.repoDiff = repoDiff;
     }

src/GitVersion.LibGit2Sharp/Git/CommitCollection.cs

@@ -8,12 +8,14 @@ internal sealed class CommitCollection : ICommitCollection
     private readonly ICommitLog innerCollection;
     private readonly Lazy<IReadOnlyCollection<ICommit>> commits;
     private readonly Diff diff;
+    private readonly GitCache gitCache;
 
-    internal CommitCollection(ICommitLog collection, Diff diff)
+    internal CommitCollection(ICommitLog collection, Diff diff, GitCache gitCache)
     {
         this.innerCollection = collection.NotNull();
-        this.commits = new Lazy<IReadOnlyCollection<ICommit>>(() => [.. this.innerCollection.Select(commit => new Commit(commit, diff))]);
+        this.commits = new Lazy<IReadOnlyCollection<ICommit>>(() => [.. this.innerCollection.Select(commit => gitCache.GetOrCreate(commit, diff))]);
         this.diff = diff.NotNull();
+        this.gitCache = gitCache;
     }
 
     public IEnumerator<ICommit> GetEnumerator()
@@ -36,7 +38,7 @@ public IEnumerable<ICommit> QueryBy(CommitFilter commitFilter)
             SortBy = (LibGit2Sharp.CommitSortStrategies)commitFilter.SortBy
         };
         var commitLog = ((IQueryableCommitLog)this.innerCollection).QueryBy(filter);
-        return new CommitCollection(commitLog, this.diff);
+        return new CommitCollection(commitLog, this.diff, this.gitCache);
 
         static object? GetReacheableFrom(object? item) =>
             item switch

src/GitVersion.LibGit2Sharp/Git/GitCache.cs

@@ -0,0 +1,23 @@
+using System.Collections.Concurrent;
+
+namespace GitVersion.Git;
+
+internal sealed class GitCache
+{
+    private static readonly ConcurrentDictionary<string, Branch> cachedBranches = new();
+    private static readonly ConcurrentDictionary<string, Commit> cachedCommits = new();
+    private static readonly ConcurrentDictionary<string, Reference> cachedReferences = new();
+    private static readonly ConcurrentDictionary<string, Remote> cachedRemotes = new();
+    private static readonly ConcurrentDictionary<string, Tag> cachedTags = new();
+
+    public Branch GetOrCreate(LibGit2Sharp.Branch innerBranch, LibGit2Sharp.Diff repoDiff) =>
+        cachedBranches.GetOrAdd(innerBranch.CanonicalName, new Branch(innerBranch, repoDiff, this));
+    public Commit GetOrCreate(LibGit2Sharp.Commit innerCommit, LibGit2Sharp.Diff repoDiff) =>
+        cachedCommits.GetOrAdd(innerCommit.Sha, new Commit(innerCommit, repoDiff, this));
+    public Reference GetOrCreate(LibGit2Sharp.Reference innerReference) =>
+        cachedReferences.GetOrAdd(innerReference.CanonicalName, new Reference(innerReference));
+    public Remote GetOrCreate(LibGit2Sharp.Remote innerRemote) =>
+        cachedRemotes.GetOrAdd(innerRemote.Name, new Remote(innerRemote));
+    public Tag GetOrCreate(LibGit2Sharp.Tag innerTag, LibGit2Sharp.Diff repoDiff) =>
+        cachedTags.GetOrAdd(innerTag.CanonicalName, new Tag(innerTag, repoDiff, this));
+}

src/GitVersion.LibGit2Sharp/Git/GitRepository.cs

@@ -7,6 +7,7 @@ namespace GitVersion.Git;
 internal sealed partial class GitRepository
 {
     private Lazy<IRepository>? repositoryLazy;
+    private readonly GitCache gitCache = new();
 
     private IRepository RepositoryInstance
     {
@@ -16,17 +17,18 @@ private IRepository RepositoryInstance
             return lazy.Value;
         }
     }
+
     public string Path => RepositoryInstance.Info.Path;
     public string WorkingDirectory => RepositoryInstance.Info.WorkingDirectory;
     public bool IsHeadDetached => RepositoryInstance.Info.IsHeadDetached;
     public bool IsShallow => RepositoryInstance.Info.IsShallow;
-    public IBranch Head => new Branch(RepositoryInstance.Head, RepositoryInstance.Diff);
+    public IBranch Head => this.gitCache.GetOrCreate(RepositoryInstance.Head, RepositoryInstance.Diff);
 
-    public ITagCollection Tags => new TagCollection(RepositoryInstance.Tags, RepositoryInstance.Diff);
-    public IReferenceCollection Refs => new ReferenceCollection(RepositoryInstance.Refs);
-    public IBranchCollection Branches => new BranchCollection(RepositoryInstance.Branches, RepositoryInstance.Diff);
-    public ICommitCollection Commits => new CommitCollection(RepositoryInstance.Commits, RepositoryInstance.Diff);
-    public IRemoteCollection Remotes => new RemoteCollection(RepositoryInstance.Network.Remotes);
+    public ITagCollection Tags => new TagCollection(RepositoryInstance.Tags, RepositoryInstance.Diff, this.gitCache);
+    public IReferenceCollection Refs => new ReferenceCollection(RepositoryInstance.Refs, this.gitCache);
+    public IBranchCollection Branches => new BranchCollection(RepositoryInstance.Branches, RepositoryInstance.Diff, this.gitCache);
+    public ICommitCollection Commits => new CommitCollection(RepositoryInstance.Commits, RepositoryInstance.Diff, this.gitCache);
+    public IRemoteCollection Remotes => new RemoteCollection(RepositoryInstance.Network.Remotes, this.gitCache);
 
     public void DiscoverRepository(string? gitDirectory)
     {
@@ -48,7 +50,7 @@ public void DiscoverRepository(string? gitDirectory)
             var first = (Commit)commit;
             var second = (Commit)otherCommit;
             var mergeBase = RepositoryInstance.ObjectDatabase.FindMergeBase(first, second);
-            return mergeBase == null ? null : new Commit(mergeBase, RepositoryInstance.Diff);
+            return mergeBase == null ? null : this.gitCache.GetOrCreate(mergeBase, RepositoryInstance.Diff);
         });
     }
 

src/GitVersion.LibGit2Sharp/Git/ReferenceCollection.cs

@@ -5,13 +5,18 @@ namespace GitVersion.Git;
 internal sealed class ReferenceCollection : IReferenceCollection
 {
     private readonly LibGit2Sharp.ReferenceCollection innerCollection;
+    private readonly GitCache gitCache;
     private IReadOnlyCollection<IReference>? references;
 
-    internal ReferenceCollection(LibGit2Sharp.ReferenceCollection collection) => this.innerCollection = collection.NotNull();
+    internal ReferenceCollection(LibGit2Sharp.ReferenceCollection collection, GitCache gitCache)
+    {
+        this.innerCollection = collection.NotNull();
+        this.gitCache = gitCache;
+    }
 
     public IEnumerator<IReference> GetEnumerator()
     {
-        this.references ??= [.. this.innerCollection.Select(reference => new Reference(reference))];
+        this.references ??= [.. this.innerCollection.Select(reference => this.gitCache.GetOrCreate(reference))];
         return this.references.GetEnumerator();
     }
 
@@ -30,13 +35,13 @@ public IReference? this[string name]
         get
         {
             var reference = this.innerCollection[name];
-            return reference is null ? null : new Reference(reference);
+            return reference is null ? null : this.gitCache.GetOrCreate(reference);
         }
     }
 
     public IReference? this[ReferenceName referenceName] => this[referenceName.Canonical];
 
     public IReference? Head => this["HEAD"];
 
-    public IEnumerable<IReference> FromGlob(string prefix) => this.innerCollection.FromGlob(prefix).Select(reference => new Reference(reference));
+    public IEnumerable<IReference> FromGlob(string prefix) => this.innerCollection.FromGlob(prefix).Select(reference => this.gitCache.GetOrCreate(reference));
 }

src/GitVersion.LibGit2Sharp/Git/RemoteCollection.cs

@@ -5,13 +5,18 @@ namespace GitVersion.Git;
 internal sealed class RemoteCollection : IRemoteCollection
 {
     private readonly LibGit2Sharp.RemoteCollection innerCollection;
+    private readonly GitCache gitCache;
     private IReadOnlyCollection<IRemote>? remotes;
 
-    internal RemoteCollection(LibGit2Sharp.RemoteCollection collection) => this.innerCollection = collection.NotNull();
+    internal RemoteCollection(LibGit2Sharp.RemoteCollection collection, GitCache gitCache)
+    {
+        this.innerCollection = collection.NotNull();
+        this.gitCache = gitCache;
+    }
 
     public IEnumerator<IRemote> GetEnumerator()
     {
-        this.remotes ??= [.. this.innerCollection.Select(reference => new Remote(reference))];
+        this.remotes ??= [.. this.innerCollection.Select(reference => this.gitCache.GetOrCreate(reference))];
         return this.remotes.GetEnumerator();
     }
 
@@ -22,7 +27,7 @@ public IRemote? this[string name]
         get
         {
             var remote = this.innerCollection[name];
-            return remote is null ? null : new Remote(remote);
+            return remote is null ? null : this.gitCache.GetOrCreate(remote);
         }
     }
 

src/GitVersion.LibGit2Sharp/Git/Tag.cs

@@ -11,12 +11,14 @@ internal sealed class Tag : ITag
     private readonly LibGit2Sharp.Tag innerTag;
     private readonly Diff diff;
     private readonly Lazy<ICommit?> commitLazy;
+    private readonly GitCache gitCache;
 
-    internal Tag(LibGit2Sharp.Tag tag, Diff diff)
+    internal Tag(LibGit2Sharp.Tag tag, Diff diff, GitCache gitCache)
     {
         this.innerTag = tag.NotNull();
         this.commitLazy = new(PeeledTargetCommit);
         this.diff = diff.NotNull();
+        this.gitCache = gitCache;
         Name = new(this.innerTag.CanonicalName);
     }
 
@@ -35,7 +37,7 @@ internal Tag(LibGit2Sharp.Tag tag, Diff diff)
             target = annotation.Target;
         }
 
-        return target is LibGit2Sharp.Commit commit ? new Commit(commit, this.diff) : null;
+        return target is LibGit2Sharp.Commit commit ? this.gitCache.GetOrCreate(commit, this.diff) : null;
     }
 
     public override bool Equals(object? obj) => Equals(obj as ITag);