Improve Library

app/API/PWChunkContext.Codeunit.al

@@ -14,6 +14,7 @@ codeunit 99001 "PW Chunk Context"
         KeyNotFoundErr: Label 'Key "%1" not found in chunk input.', Comment = '%1 = JSON key name';
         NotRecordChunkErr: Label 'This chunk was not created via RunForRecords.';
         NotInitializedErr: Label 'Chunk context has not been initialized.';
+        RecordSetChangedErr: Label 'The chunk''s record range no longer matches the data: the filtered record set is empty or has shrunk since the batch was created.';
 
     procedure Init(var Chunk: Record "PW Batch Chunk")
     var
@@ -202,9 +203,14 @@ codeunit 99001 "PW Chunk Context"
         if FilterView <> '' then
             RecRef.SetView(FilterView);
 
-        RecRef.FindSet();
+        // Chunk ranges are positional — if records were inserted or deleted
+        // since the batch was created, the ranges no longer match the data.
+        // Fail loudly instead of silently processing another chunk's records.
+        if not RecRef.FindSet() then
+            Error(RecordSetChangedErr);
         if StartIndex > 1 then
-            RecRef.Next(StartIndex - 1);
+            if RecRef.Next(StartIndex - 1) <> StartIndex - 1 then
+                Error(RecordSetChangedErr);
     end;
 
     /// <summary>
@@ -222,6 +228,7 @@ codeunit 99001 "PW Chunk Context"
     /// <param name="Result">The JSON object to store as the sole result.</param>
     procedure SetResult(Result: JsonObject)
     begin
+        CheckInitialized();
         Clear(Results);
         Results.Add(Result);
     end;
@@ -232,6 +239,7 @@ codeunit 99001 "PW Chunk Context"
     /// <param name="Result">The JSON object to append.</param>
     procedure AppendResult(Result: JsonObject)
     begin
+        CheckInitialized();
         Results.Add(Result);
     end;
 

app/Engine/PWTaskDispatcher.Codeunit.al

@@ -8,8 +8,16 @@ codeunit 99002 "PW Task Dispatcher"
         ErrorText: Text;
         ErrorCallStack: Text;
     begin
+        // UpdLock serializes the claim with the coordinator storing this chunk's
+        // session id in StartBatch and with dead-session recovery — a plain read
+        // could hit an optimistic-concurrency error on the Modify below.
+        Rec.ReadIsolation := IsolationLevel::UpdLock;
         if not Rec.Get(Rec."Batch Id", Rec."Chunk Index") then
             exit;
+        // Dead-session recovery may have already failed a chunk whose session
+        // was queued for too long — never resurrect a chunk that left Pending.
+        if Rec.Status <> "PW Chunk Status"::Pending then
+            exit;
 
         Rec.Status := "PW Chunk Status"::Running;
         Rec."Session Id" := SessionId();
@@ -29,21 +37,26 @@ codeunit 99002 "PW Task Dispatcher"
         if Codeunit.Run(Codeunit::"PW Worker Runner", Rec) then begin
             // Worker succeeded — results already saved by PW Worker Runner.
             // Re-read to pick up changes made inside Codeunit.Run.
-            Rec.Get(Rec."Batch Id", Rec."Chunk Index");
+            // The chunk disappears if the caller ran Cleanup mid-batch —
+            // nothing left to persist.
+            if not Rec.Get(Rec."Batch Id", Rec."Chunk Index") then
+                exit;
             Rec.Status := "PW Chunk Status"::Completed;
             Rec."Completed At" := CurrentDateTime();
             Rec.Modify();
             // Persist chunk's Completed status before updating batch counters.
             // If UpdateBatchCounters fails (e.g. lock timeout on PW Batch),
             // the chunk is still correctly marked as Completed.
-            // The next finishing chunk will recompute counters correctly.
+            // WaitForCompletion reconciles the counters if no chunk finishes after this one.
             Commit();
         end else begin
             // Worker failed — Codeunit.Run rolled back all DB changes.
-            // Re-read to ensure Rec is clean, consistent with the success branch.
-            Rec.Get(Rec."Batch Id", Rec."Chunk Index");
+            // Capture the error before any further record operations,
+            // then re-read to ensure Rec is clean, consistent with the success branch.
             ErrorText := GetLastErrorText();
             ErrorCallStack := GetLastErrorCallStack();
+            if not Rec.Get(Rec."Batch Id", Rec."Chunk Index") then
+                exit;
             Rec."Error Message" := CopyStr(ErrorText, 1, MaxStrLen(Rec."Error Message"));
             WriteFullErrorMessage(Rec, ErrorText);
             WriteErrorCallStack(Rec, ErrorCallStack);
@@ -83,7 +96,9 @@ codeunit 99002 "PW Task Dispatcher"
         // UpdLock on the Batch row serializes concurrent counter updates
         // without disabling Tri-State Locking for the entire table.
         Batch.ReadIsolation := IsolationLevel::UpdLock;
-        Batch.Get(BatchId);
+        // The batch disappears if the caller ran Cleanup mid-batch.
+        if not Batch.Get(BatchId) then
+            exit;
 
         // ReadCommitted ensures we see committed chunk statuses only.
         // Each dispatcher commits its chunk status before calling this procedure.

app/Monitoring/PWBatchCleanup.Codeunit.al

@@ -10,21 +10,18 @@ codeunit 99003 "PW Batch Cleanup"
     procedure CleanupOlderThan(Hours: Integer)
     var
         Batch: Record "PW Batch";
-        Chunk: Record "PW Batch Chunk";
+        CutoffOffset: Duration;
     begin
+        // Duration arithmetic — Hours * 3600000 as Integer overflows at ~596 hours.
+        CutoffOffset := Hours;
+        CutoffOffset := CutoffOffset * 3600000;
+
         Batch.SetFilter(Status, '%1|%2|%3',
             "PW Batch Status"::Completed,
             "PW Batch Status"::PartialFailure,
             "PW Batch Status"::Failed);
-        Batch.SetFilter("Completed At", '<>%1', 0DT);
-
-        if Batch.FindSet() then
-            repeat
-                if (CurrentDateTime() - Batch."Completed At") > Hours * 3600000 then begin
-                    Chunk.SetRange("Batch Id", Batch.Id);
-                    Chunk.DeleteAll();
-                    Batch.Delete();
-                end;
-            until Batch.Next() = 0;
+        Batch.SetFilter("Completed At", '<>%1&<%2', 0DT, CurrentDateTime() - CutoffOffset);
+        // OnDelete on PW Batch removes the chunks.
+        Batch.DeleteAll(true);
     end;
 }

app/Translations/Parallel Worker.g.xlf

@@ -118,11 +118,26 @@
           <note from="Developer" annotates="general" priority="2"></note>
           <note from="Xliff Generator" annotates="general" priority="3">Table PW Batch Chunk - Field Status - Property Caption</note>
         </trans-unit>
+        <trans-unit id="Codeunit 797980843 - NamedType 2314357299" size-unit="char" translate="yes" xml:space="preserve">
+          <source>Thread count must be at least 1.</source>
+          <note from="Developer" annotates="general" priority="2"></note>
+          <note from="Xliff Generator" annotates="general" priority="3">Codeunit PW Batch Coordinator - NamedType InvalidThreadCountErr</note>
+        </trans-unit>
+        <trans-unit id="Codeunit 797980843 - NamedType 3558797986" size-unit="char" translate="yes" xml:space="preserve">
+          <source>The payload key %1 is not allowed. Keys prefixed with $ are reserved for the framework.</source>
+          <note from="Developer" annotates="general" priority="2">%1 = JSON key name</note>
+          <note from="Xliff Generator" annotates="general" priority="3">Codeunit PW Batch Coordinator - NamedType ReservedKeyErr</note>
+        </trans-unit>
         <trans-unit id="Codeunit 797980843 - NamedType 1294502618" size-unit="char" translate="yes" xml:space="preserve">
           <source>Session terminated (timeout or server stop).</source>
           <note from="Developer" annotates="general" priority="2"></note>
           <note from="Xliff Generator" annotates="general" priority="3">Codeunit PW Batch Coordinator - NamedType SessionTerminatedErr</note>
         </trans-unit>
+        <trans-unit id="Codeunit 797980843 - NamedType 2503317769" size-unit="char" translate="yes" xml:space="preserve">
+          <source>Cannot wait for a parallel batch while a write transaction is active. WaitForCompletion may commit internally during recovery, which would silently commit your changes.</source>
+          <note from="Developer" annotates="general" priority="2"></note>
+          <note from="Xliff Generator" annotates="general" priority="3">Codeunit PW Batch Coordinator - NamedType WaitWriteTransactionErr</note>
+        </trans-unit>
         <trans-unit id="Codeunit 797980843 - NamedType 3772275472" size-unit="char" translate="yes" xml:space="preserve">
           <source>Cannot start a parallel batch while a write transaction is active. Commit or rollback your changes before calling RunFor* methods, otherwise they would be silently committed.</source>
           <note from="Developer" annotates="general" priority="2"></note>
@@ -143,6 +158,11 @@
           <note from="Developer" annotates="general" priority="2"></note>
           <note from="Xliff Generator" annotates="general" priority="3">Codeunit PW Chunk Context - NamedType NotRecordChunkErr</note>
         </trans-unit>
+        <trans-unit id="Codeunit 914310053 - NamedType 2954270052" size-unit="char" translate="yes" xml:space="preserve">
+          <source>The chunk's record range no longer matches the data: the filtered record set is empty or has shrunk since the batch was created.</source>
+          <note from="Developer" annotates="general" priority="2"></note>
+          <note from="Xliff Generator" annotates="general" priority="3">Codeunit PW Chunk Context - NamedType RecordSetChangedErr</note>
+        </trans-unit>
         <trans-unit id="Codeunit 3345984611 - NamedType 3198830347" size-unit="char" translate="yes" xml:space="preserve">
           <source>Chunk %1: </source>
           <note from="Developer" annotates="general" priority="2">%1 = chunk index</note>

test/disabledtests.json

@@ -202,5 +202,17 @@
         "codeunitId":  99209,
         "codeunitName":  "PW Integration Test",
         "method":  "BatchTimeoutReturnsFalseButSessionContinues"
+    },
+    {
+        "bug":  "Need to set isolation disabled test runner. Currently not supported by AL-Go.",
+        "codeunitId":  99209,
+        "codeunitName":  "PW Integration Test",
+        "method":  "SessionIdStoredOnChunksAtStart"
+    },
+    {
+        "bug":  "Need to set isolation disabled test runner. Currently not supported by AL-Go.",
+        "codeunitId":  99209,
+        "codeunitName":  "PW Integration Test",
+        "method":  "CleanupWhileSessionsRunningIsSafe"
     }
 ]

test/src/Helpers/PWTestHelper.Codeunit.al

@@ -97,4 +97,31 @@ codeunit 99204 "PW Test Helper"
         Chunk.Status := Status;
         Chunk.Insert();
     end;
+
+    procedure CreateChunkWithSession(BatchId: Guid; ChunkIndex: Integer; Status: Enum "PW Chunk Status"; SessionId: Integer)
+    var
+        Chunk: Record "PW Batch Chunk";
+    begin
+        Chunk."Batch Id" := BatchId;
+        Chunk."Chunk Index" := ChunkIndex;
+        Chunk.Status := Status;
+        Chunk."Session Id" := SessionId;
+        Chunk.Insert();
+    end;
+
+    procedure CreateChunkWithFullError(BatchId: Guid; ChunkIndex: Integer; ErrorText: Text)
+    var
+        Chunk: Record "PW Batch Chunk";
+        OutStream: OutStream;
+    begin
+        Chunk."Batch Id" := BatchId;
+        Chunk."Chunk Index" := ChunkIndex;
+        Chunk.Status := "PW Chunk Status"::Failed;
+        Chunk."Error Message" := CopyStr(ErrorText, 1, MaxStrLen(Chunk."Error Message"));
+        Chunk.Insert();
+
+        Chunk."Full Error Message".CreateOutStream(OutStream, TextEncoding::UTF8);
+        OutStream.WriteText(ErrorText);
+        Chunk.Modify();
+    end;
 }

test/src/Tests/PWBatchCleanupTest.Codeunit.al

@@ -126,6 +126,24 @@ codeunit 99203 "PW Batch Cleanup Test"
         LibraryAssert.IsFalse(Batch.Get(BatchId), 'OnRun should delete old completed batch');
     end;
 
+    [Test]
+    procedure CleanupOlderThanLargeHoursDoesNotOverflow()
+    // [SCENARIO 5.9] Hours large enough to overflow 32-bit millisecond math
+    // (Hours * 3600000 > MaxInt at ~596 hours) is handled without a runtime error
+    var
+        Cleanup: Codeunit "PW Batch Cleanup";
+        Batch: Record "PW Batch";
+        BatchId: Guid;
+    begin
+        BatchId := TestHelper.CreateBatchWithTimestamp(
+            "PW Batch Status"::Completed,
+            CurrentDateTime() - (48 * 3600000));
+
+        Cleanup.CleanupOlderThan(100000);
+
+        LibraryAssert.IsTrue(Batch.Get(BatchId), '48h-old batch should survive a 100000h threshold');
+    end;
+
     [Test]
     procedure MixedScenarioDeletesOnlyOldFinished()
     // [SCENARIO 5.7] Mixed — only old finished batches are deleted

test/src/Tests/PWChunkContextTest.Codeunit.al

@@ -507,4 +507,106 @@ codeunit 99201 "PW Chunk Context Test"
         Chunk.CalcFields("Result Payload");
         LibraryAssert.IsFalse(Chunk."Result Payload".HasValue(), 'Result blob should be empty');
     end;
+
+    [Test]
+    procedure SetResultErrorsBeforeInit()
+    // [SCENARIO 1.23] SetResult called before Init raises error
+    var
+        Ctx: Codeunit "PW Chunk Context";
+        Result: JsonObject;
+    begin
+        asserterror Ctx.SetResult(Result);
+
+        LibraryAssert.ExpectedError('Chunk context has not been initialized.');
+    end;
+
+    [Test]
+    procedure AppendResultErrorsBeforeInit()
+    // [SCENARIO 1.24] AppendResult called before Init raises error
+    var
+        Ctx: Codeunit "PW Chunk Context";
+        Result: JsonObject;
+    begin
+        asserterror Ctx.AppendResult(Result);
+
+        LibraryAssert.ExpectedError('Chunk context has not been initialized.');
+    end;
+
+    [Test]
+    procedure GetRecordRefErrorsWhenRecordSetIsEmpty()
+    // [SCENARIO 1.25] GetRecordRef fails loudly when the chunk's filtered set is now empty
+    var
+        Ctx: Codeunit "PW Chunk Context";
+        Chunk: Record "PW Batch Chunk";
+        SourceBatch: Record "PW Batch";
+        SourceRef: RecordRef;
+        TargetRef: RecordRef;
+        BatchId: Guid;
+        Payload: JsonObject;
+    begin
+        SourceBatch.SetRange("Company Name", 'PW_CTX_EMPTY');
+        SourceRef.GetTable(SourceBatch);
+
+        Payload.Add('$IsRecordChunk', true);
+        Payload.Add('$TableNo', Database::"PW Batch");
+        Payload.Add('$FilterView', SourceRef.GetView());
+        Payload.Add('$StartIndex', 1);
+        Payload.Add('$EndIndex', 5);
+        BatchId := TestHelper.CreateBatch("PW Batch Status"::Running);
+        TestHelper.CreateChunkWithPayload(BatchId, 1, Payload);
+        Chunk.Get(BatchId, 1);
+        Ctx.Init(Chunk);
+
+        asserterror Ctx.GetRecordRef(TargetRef);
+
+        LibraryAssert.ExpectedError('no longer matches the data');
+    end;
+
+    [Test]
+    procedure GetRecordRefErrorsWhenRecordSetShrank()
+    // [SCENARIO 1.26] GetRecordRef fails loudly when the set has fewer records
+    // than the chunk's start index — positioning would silently overlap another chunk
+    var
+        Ctx: Codeunit "PW Chunk Context";
+        Chunk: Record "PW Batch Chunk";
+        SourceBatch: Record "PW Batch";
+        SourceRef: RecordRef;
+        TargetRef: RecordRef;
+        BatchId: Guid;
+        Payload: JsonObject;
+        i: Integer;
+    begin
+        SourceBatch.SetRange("Company Name", 'PW_CTX_SHRUNK');
+        SourceBatch.DeleteAll();
+        for i := 1 to 2 do begin
+            Clear(SourceBatch);
+            SourceBatch.Id := CreateGuid();
+            SourceBatch.Status := "PW Batch Status"::Pending;
+            SourceBatch."Company Name" := 'PW_CTX_SHRUNK';
+            SourceBatch.Insert();
+        end;
+
+        SourceBatch.Reset();
+        SourceBatch.SetRange("Company Name", 'PW_CTX_SHRUNK');
+        SourceRef.GetTable(SourceBatch);
+
+        // Chunk expects to start at record 5 but only 2 records exist
+        Payload.Add('$IsRecordChunk', true);
+        Payload.Add('$TableNo', Database::"PW Batch");
+        Payload.Add('$FilterView', SourceRef.GetView());
+        Payload.Add('$StartIndex', 5);
+        Payload.Add('$EndIndex', 6);
+        BatchId := TestHelper.CreateBatch("PW Batch Status"::Running);
+        TestHelper.CreateChunkWithPayload(BatchId, 1, Payload);
+        Chunk.Get(BatchId, 1);
+        Ctx.Init(Chunk);
+
+        asserterror Ctx.GetRecordRef(TargetRef);
+
+        LibraryAssert.ExpectedError('no longer matches the data');
+
+        SourceBatch.Reset();
+        SourceBatch.SetRange("Company Name", 'PW_CTX_SHRUNK');
+        SourceBatch.DeleteAll();
+    end;
 }