Merge pull request #173 from Tiktack/threading-performance

Improve performance of API Threading strategy.

Wino.Core.Domain/Interfaces/IThreadingStrategy.cs

@@ -7,6 +7,11 @@ namespace Wino.Core.Domain.Interfaces
 {
     public interface IThreadingStrategy
     {
+        /// <summary>
+        /// Attach thread mails to the list.
+        /// </summary>
+        /// <param name="items">Original mails.</param>
+        /// <returns>Original mails with thread mails.</returns>
         Task<List<IMailItem>> ThreadItemsAsync(List<MailCopy> items);
         bool ShouldThreadWithItem(IMailItem originalItem, IMailItem targetItem);
     }

Wino.Core/Integration/Threading/APIThreadingStrategy.cs

@@ -26,87 +26,55 @@ namespace Wino.Core.Integration.Threading
             return originalItem.ThreadId != null && originalItem.ThreadId == targetItem.ThreadId;
         }
 
+        ///<inheritdoc/>
         public async Task<List<IMailItem>> ThreadItemsAsync(List<MailCopy> items)
         {
-            var accountId = items.First().AssignedAccount.Id;
+            var assignedAccount = items[0].AssignedAccount;
 
-            var threads = new List<ThreadMailItem>();
+            var sentFolder = await _folderService.GetSpecialFolderByAccountIdAsync(assignedAccount.Id, SpecialFolderType.Sent);
-            var assignedAccount = items.First().AssignedAccount;
+            var draftFolder = await _folderService.GetSpecialFolderByAccountIdAsync(assignedAccount.Id, SpecialFolderType.Draft);
-
-            // TODO: Can be optimized by moving to the caller.
-            var sentFolder = await _folderService.GetSpecialFolderByAccountIdAsync(accountId, Domain.Enums.SpecialFolderType.Sent);
-            var draftFolder = await _folderService.GetSpecialFolderByAccountIdAsync(accountId, Domain.Enums.SpecialFolderType.Draft);
 
             if (sentFolder == null || draftFolder == null) return default;
 
-            // Child -> Parent approach.
+            // True: Non threaded items.
+            // False: Potentially threaded items.
+            var nonThreadedOrThreadedMails = items
+                .Distinct()
+                .GroupBy(x => string.IsNullOrEmpty(x.ThreadId))
+                .ToDictionary(x => x.Key, x => x);
 
-            var potentialThreadItems = items.Distinct().Where(a => !string.IsNullOrEmpty(a.ThreadId));
+            _ = nonThreadedOrThreadedMails.TryGetValue(true, out var nonThreadedMails);
+            var isThreadedItems = nonThreadedOrThreadedMails.TryGetValue(false, out var potentiallyThreadedMails);
 
-            var mailLookupTable = new Dictionary<string, bool>();
+            List<IMailItem> resultList = nonThreadedMails is null ? [] : [.. nonThreadedMails];
 
-            // Fill up the mail lookup table to prevent double thread creation.
+            if (isThreadedItems)
-            foreach (var mail in items)
-                if (!mailLookupTable.ContainsKey(mail.Id))
-                    mailLookupTable.Add(mail.Id, false);
-
-            foreach (var potentialItem in potentialThreadItems)
             {
-                if (mailLookupTable[potentialItem.Id])
+                var threadItems = (await GetThreadItemsAsync(potentiallyThreadedMails.Select(x => (x.ThreadId, x.AssignedFolder)).ToList(), assignedAccount.Id, sentFolder.Id, draftFolder.Id))
+                .GroupBy(x => x.ThreadId);
+
+                foreach (var threadItem in threadItems)
+                {
+                    if (threadItem.Count() == 1)
+                    {
+                        resultList.Add(threadItem.First());
                         continue;
+                    }
 
-                mailLookupTable[potentialItem.Id] = true;
+                    var thread = new ThreadMailItem();
-
+                    foreach (var childThreadItem in threadItem)
-                var allThreadItems = await GetThreadItemsAsync(potentialItem.ThreadId, accountId, potentialItem.AssignedFolder, sentFolder.Id, draftFolder.Id);
-
-                if (allThreadItems.Count == 1)
                     {
-                    // It's a single item.
+                        thread.AddThreadItem(childThreadItem);
-                    // Mark as not-processed as thread.
-
-                    mailLookupTable[potentialItem.Id] = false;
                     }
-                else
+                    resultList.Add(thread);
-                {
-                    // Thread item. Mark all items as true in dict.
-                    var threadItem = new ThreadMailItem();
-
-                    foreach (var childThreadItem in allThreadItems)
-                    {
-                        if (mailLookupTable.ContainsKey(childThreadItem.Id))
-                            mailLookupTable[childThreadItem.Id] = true;
-
-                        childThreadItem.AssignedAccount = assignedAccount;
-                        childThreadItem.AssignedFolder = await _folderService.GetFolderAsync(childThreadItem.FolderId);
-
-                        threadItem.AddThreadItem(childThreadItem);
-                    }
-
-                    // Multiple mail copy ids from different folders are thing for Gmail.
-                    if (threadItem.ThreadItems.Count == 1)
-                        mailLookupTable[potentialItem.Id] = false;
-                    else
-                        threads.Add(threadItem);
                 }
             }
 
-            // At this points all mails in the list belong to single items.
+            return resultList;
-            // Merge with threads.
-            // Last sorting will be done later on in MailService.
-
-            // Remove single mails that are included in thread.
-            items.RemoveAll(a => mailLookupTable.ContainsKey(a.Id) && mailLookupTable[a.Id]);
-
-            var finalList = new List<IMailItem>(items);
-
-            finalList.AddRange(threads);
-
-            return finalList;
         }
 
-        private async Task<List<MailCopy>> GetThreadItemsAsync(string threadId,
+        private async Task<List<MailCopy>> GetThreadItemsAsync(List<(string threadId, MailItemFolder threadingFolder)> potentialThread,
                                                                Guid accountId,
-                                                               MailItemFolder threadingFolder,
                                                                Guid sentFolderId,
                                                                Guid draftFolderId)
         {
@@ -118,24 +86,20 @@ namespace Wino.Core.Integration.Threading
 
             // TODO: Convert to SQLKata query.
 
-            string query = string.Empty;
+            var query = @$"SELECT DISTINCT MC.* FROM MailCopy MC
-
-            if (threadingFolder.SpecialFolderType == SpecialFolderType.Draft || threadingFolder.SpecialFolderType == SpecialFolderType.Sent)
-            {
-                query = @$"SELECT DISTINCT MC.* FROM MailCopy MC
                            INNER JOIN MailItemFolder MF on MF.Id = MC.FolderId
-                           WHERE MF.MailAccountId == '{accountId}' AND MC.ThreadId = '{threadId}'";
+                           WHERE MF.MailAccountId == '{accountId}' AND
-            }
+                           ({string.Join(" OR ", potentialThread.Select(x => ConditionForItem(x, sentFolderId, draftFolderId)))})";
-            else
-            {
-                query = @$"SELECT DISTINCT MC.* FROM MailCopy MC
-                           INNER JOIN MailItemFolder MF on MF.Id = MC.FolderId
-                           WHERE MF.MailAccountId == '{accountId}' AND MC.FolderId IN ('{threadingFolder.Id}','{sentFolderId}','{draftFolderId}')
-                           AND MC.ThreadId = '{threadId}'";
-            }
-
 
             return await _databaseService.Connection.QueryAsync<MailCopy>(query);
+
+            static string ConditionForItem((string threadId, MailItemFolder threadingFolder) potentialThread, Guid sentFolderId, Guid draftFolderId)
+            {
+                if (potentialThread.threadingFolder.SpecialFolderType == SpecialFolderType.Draft || potentialThread.threadingFolder.SpecialFolderType == SpecialFolderType.Sent)
+                    return $"(MC.ThreadId = '{potentialThread.threadId}')";
+
+                return $"(MC.ThreadId = '{potentialThread.threadId}' AND MC.FolderId IN ('{potentialThread.threadingFolder.Id}','{sentFolderId}','{draftFolderId}'))";
+            }
         }
     }
 }

Wino.Core/Services/MailService.cs

@@ -2,6 +2,7 @@
 using System.Collections.Generic;
 using System.Linq;
 using System.Threading.Tasks;
+using Microsoft.Kiota.Abstractions.Extensions;
 using MimeKit;
 using MimeKit.Text;
 using MoreLinq;
@@ -186,46 +187,35 @@ namespace Wino.Core.Services
 
             var mails = await Connection.QueryAsync<MailCopy>(query);
 
-            // Fill in assigned account and folder for each mail.
+            Dictionary<Guid, MailItemFolder> folderCache = [];
-            // To speed things up a bit, we'll load account and assigned folder in groups
+            Dictionary<Guid, MailAccount> accountCache = [];
-            // to reduce the query time.
 
-            var groupedByFolders = mails.GroupBy(a => a.FolderId);
+            // Populate Folder Assignment for each single mail, to be able later group by "MailAccountId".
-
+            // This is needed to execute threading strategy by account type.
-            foreach (var group in groupedByFolders)
+            // Avoid DBs calls as possible, storing info in a dictionary.
+            foreach (var mail in mails)
             {
-                MailItemFolder folderAssignment = null;
+                await LoadAssignedPropertiesWithCacheAsync(mail, folderCache, accountCache).ConfigureAwait(false);
-                MailAccount accountAssignment = null;
-
-                folderAssignment = await _folderService.GetFolderAsync(group.Key).ConfigureAwait(false);
-
-                if (folderAssignment != null)
-                {
-                    accountAssignment = await _accountService.GetAccountAsync(folderAssignment.MailAccountId).ConfigureAwait(false);
-                }
-
-                group.ForEach(a =>
-                {
-                    a.AssignedFolder = folderAssignment;
-                    a.AssignedAccount = accountAssignment;
-                });
             }
 
             // Remove items that has no assigned account or folder.
             mails.RemoveAll(a => a.AssignedAccount == null || a.AssignedFolder == null);
 
-            // Each account items must be threaded separately.
+            if (!options.CreateThreads)
-
-            if (options.CreateThreads)
             {
+                // Threading is disabled. Just return everything as it is.
+                mails.Sort(options.SortingOptionType == SortingOptionType.ReceiveDate ? new DateComparer() : new NameComparer());
+
+                return new List<IMailItem>(mails);
+            }
+
+            // Populate threaded items.
+
             var threadedItems = new List<IMailItem>();
 
-                var groupedByAccounts = mails.GroupBy(a => a.AssignedAccount.Id);
+            // Each account items must be threaded separately.
-
+            foreach (var group in mails.GroupBy(a => a.AssignedAccount.Id))
-                foreach (var group in groupedByAccounts)
             {
-                    if (!group.Any()) continue;
-
                 var accountId = group.Key;
                 var groupAccount = mails.First(a => a.AssignedAccount.Id == accountId).AssignedAccount;
 
@@ -233,7 +223,14 @@ namespace Wino.Core.Services
 
                 // Only thread items from Draft and Sent folders must present here.
                 // Otherwise this strategy will fetch the items that are in Deleted folder as well.
-                    var accountThreadedItems = await threadingStrategy.ThreadItemsAsync(group.ToList());
+                var accountThreadedItems = await threadingStrategy.ThreadItemsAsync([.. group]);
+
+                // Populate threaded items with folder and account assignments.
+                // Almost everything already should be in cache from initial population.
+                foreach (var mail in accountThreadedItems)
+                {
+                    await LoadAssignedPropertiesWithCacheAsync(mail, folderCache, accountCache).ConfigureAwait(false);
+                }
 
                 if (accountThreadedItems != null)
                 {
@@ -244,14 +241,42 @@ namespace Wino.Core.Services
             threadedItems.Sort(options.SortingOptionType == SortingOptionType.ReceiveDate ? new DateComparer() : new NameComparer());
 
             return threadedItems;
-            }
+
-            else
+            // Recursive function to populate folder and account assignments for each mail item.
+            async Task LoadAssignedPropertiesWithCacheAsync(IMailItem mail, Dictionary<Guid, MailItemFolder> folderCache, Dictionary<Guid, MailAccount> accountCache)
             {
-                // Threading is disabled. Just return everything as it is.
+                if (mail is ThreadMailItem threadMailItem)
+                {
+                    foreach (var childMail in threadMailItem.ThreadItems)
+                    {
+                        await LoadAssignedPropertiesWithCacheAsync(childMail, folderCache, accountCache).ConfigureAwait(false);
+                    }
+                }
 
-                mails.Sort(options.SortingOptionType == SortingOptionType.ReceiveDate ? new DateComparer() : new NameComparer());
+                if (mail is MailCopy mailCopy)
+                {
+                    MailAccount accountAssignment = null;
 
-                return new List<IMailItem>(mails);
+                    var isFolderCached = folderCache.TryGetValue(mailCopy.FolderId, out MailItemFolder folderAssignment);
+                    accountAssignment = null;
+                    if (!isFolderCached)
+                    {
+                        folderAssignment = await _folderService.GetFolderAsync(mailCopy.FolderId).ConfigureAwait(false);
+                        _ = folderCache.TryAdd(mailCopy.FolderId, folderAssignment);
+                    }
+                    if (folderAssignment != null)
+                    {
+                        var isAccountCached = accountCache.TryGetValue(folderAssignment.MailAccountId, out accountAssignment);
+                        if (!isAccountCached)
+                        {
+                            accountAssignment = await _accountService.GetAccountAsync(folderAssignment.MailAccountId).ConfigureAwait(false);
+                            _ = accountCache.TryAdd(folderAssignment.MailAccountId, accountAssignment);
+                        }
+                    }
+
+                    mailCopy.AssignedFolder = folderAssignment;
+                    mailCopy.AssignedAccount = accountAssignment;
+                }
             }
         }