Fixed the Batch Processing assistant with an output folder above the input folder (#1098)

app/MindWork AI Studio/Assistants/BatchProcessing/AssistantBatchProcessing.razor.Validation.cs

@@ -83,39 +83,18 @@ public partial class AssistantBatchProcessing
     private IReadOnlyList<string> FindInputFiles(string resolvedOutputDirectory)
     {
         var selection = new FolderFileSelection(FileNamePatterns.Parse(this.filePatterns), FileExclusions.Parse(this.excludePatterns), this.includeSubdirectories, this.includeHiddenFilesAndFolders, this.maxFileSizeMegabytes);
+        var outputFolder = new BatchProcessingOutputFolder(this.inputDirectory, resolvedOutputDirectory, this.ResolveResultsFileName());
         var files = new SortedSet<string>(StringComparer.OrdinalIgnoreCase);
         this.numTooLargeFiles = 0;
 
-        var normalizedInputDirectory = TrimDirectorySeparator(Path.GetFullPath(this.inputDirectory));
-        var normalizedOutputDirectory = TrimDirectorySeparator(Path.GetFullPath(resolvedOutputDirectory));
-
-        // When the output folder is a folder of its own, we skip everything
-        // inside it. When it is the input folder itself, we must not skip the
-        // whole folder: we would not find any document at all. We then skip
-        // our own output artifacts instead.
-        var isOutputSeparateFolder = !string.Equals(normalizedInputDirectory, normalizedOutputDirectory, StringComparison.OrdinalIgnoreCase);
-
-        // The separator is essential: without it, an output folder named 'out'
-        // would also exclude a document named 'output-notes.md':
-        var outputDirectoryPrefix = normalizedOutputDirectory + Path.DirectorySeparatorChar;
-
         foreach (var file in selection.EnumerateCandidates(
                      this.inputDirectory,
-                     includeDirectory: directory => !isOutputSeparateFolder
-                                                    || !string.Equals(TrimDirectorySeparator(Path.GetFullPath(directory)), normalizedOutputDirectory, StringComparison.OrdinalIgnoreCase),
+                     includeDirectory: outputFolder.Enters,
                      reportInaccessibleDirectory: (directory, exception) =>
                          throw new IOException($"Could not access directory '{directory}'.", exception)))
         {
             var normalizedFile = Path.GetFullPath(file);
-            if (!selection.Selects(normalizedFile) || IsTranscriptArtifact(normalizedFile))
-                continue;
-
-            if (isOutputSeparateFolder)
-            {
-                if (normalizedFile.StartsWith(outputDirectoryPrefix, StringComparison.OrdinalIgnoreCase))
-                    continue;
-            }
-            else if (this.IsOwnOutputArtifact(normalizedFile))
+            if (!selection.Selects(normalizedFile) || IsTranscriptArtifact(normalizedFile) || outputFolder.IsOutputOfRun(normalizedFile))
                 continue;
 
             //
@@ -137,25 +116,6 @@ public partial class AssistantBatchProcessing
         return [.. files];
     }
 
-    private static string TrimDirectorySeparator(string path) => path.TrimEnd(Path.DirectorySeparatorChar, Path.AltDirectorySeparatorChar);
-
-    /// <summary>
-    /// Checks whether a file is an output artifact of this assistant. We need
-    /// this when the output folder is the input folder: without it, the results
-    /// of a previous run would be processed as documents.
-    /// </summary>
-    private bool IsOwnOutputArtifact(string filePath)
-    {
-        var fileName = Path.GetFileName(filePath);
-        if (string.Equals(fileName, LOG_FILENAME, StringComparison.OrdinalIgnoreCase))
-            return true;
-
-        if (string.Equals(fileName, this.ResolveResultsFileName(), StringComparison.OrdinalIgnoreCase))
-            return true;
-
-        return fileName.EndsWith(RESULT_FILE_SUFFIX, StringComparison.OrdinalIgnoreCase);
-    }
-
     /// <summary>
     /// Checks for persistent or interrupted media transcript artifacts. They
     /// always live beside their source file, independently of the output folder.

app/MindWork AI Studio/Assistants/BatchProcessing/AssistantBatchProcessing.razor.cs

@@ -18,7 +18,7 @@ public partial class AssistantBatchProcessing : AssistantBaseCore<SettingsDialog
     private const string DEFAULT_OUTPUT_DIRECTORY_NAME = "ai-results";
     private const string DEFAULT_RESULTS_FILENAME = "batch-results.csv";
     private const string CSV_EXTENSION = ".csv";
-    private const string RESULT_FILE_SUFFIX = "_result";
+    internal const string RESULT_FILE_SUFFIX = "_result";
     private const string TIME_FORMAT = "yyyy-MM-dd HH:mm:ss";
     private const char LOG_SEPARATOR = ';';
 
@@ -26,7 +26,7 @@ public partial class AssistantBatchProcessing : AssistantBaseCore<SettingsDialog
     /// The name of the log file. It is fixed, so that a later batch run finds
     /// the log of a previous run and can continue it.
     /// </summary>
-    private const string LOG_FILENAME = "log.csv";
+    internal const string LOG_FILENAME = "log.csv";
 
     protected override Tools.Components Component => Tools.Components.BATCH_PROCESSING_ASSISTANT;
 

app/MindWork AI Studio/Assistants/BatchProcessing/BatchProcessingOutputFolder.cs

@@ -0,0 +1,88 @@
+namespace AIStudio.Assistants.BatchProcessing;
+
+/// <summary>
+/// Tells the search for documents which of the files it finds a batch run wrote itself.
+/// </summary>
+/// <remarks>
+/// A run writes its result files, its log, and its results table directly into the output folder.
+/// What the search must leave out therefore depends on where this folder lies. Below the input
+/// folder, the search leaves out all of it. When it is the input folder itself, leaving out the
+/// whole folder would leave no document at all, so the search leaves out what a run writes, by
+/// name. Anywhere else, above the input folder included, the search never comes across the output
+/// folder and leaves out nothing.
+/// </remarks>
+/// <param name="inputDirectory">The folder the run reads its documents from.</param>
+/// <param name="outputDirectory">The folder the run writes into.</param>
+/// <param name="resultsFileName">The file name of the results table.</param>
+internal sealed class BatchProcessingOutputFolder(string inputDirectory, string outputDirectory, string resultsFileName)
+{
+    private readonly string fullOutputDirectory = Path.GetFullPath(outputDirectory);
+
+    private readonly bool isInputFolder = IsSameFolder(inputDirectory, outputDirectory);
+
+    //
+    // The same folder with a separator at its end would count as lying below itself:
+    //
+    private readonly bool liesBelowInputFolder = !IsSameFolder(inputDirectory, outputDirectory) && PathTools.IsInsideDirectory(inputDirectory, outputDirectory);
+
+    /// <summary>
+    /// Whether the search enters a subfolder of the input folder: every one but the output folder.
+    /// </summary>
+    /// <param name="directory">The subfolder.</param>
+    public bool Enters(string directory) => !this.liesBelowInputFolder || !IsSameFolder(directory, this.fullOutputDirectory);
+
+    /// <summary>
+    /// Whether a file the search found was written by a run instead of being a document.
+    /// </summary>
+    /// <param name="filePath">The full path of the file.</param>
+    public bool IsOutputOfRun(string filePath)
+    {
+        //
+        // Checked with the separator after the name of the folder: otherwise, an output folder
+        // named 'out' would also leave out a document named 'output-notes.md':
+        //
+        if (this.liesBelowInputFolder)
+            return PathTools.IsInsideDirectory(this.fullOutputDirectory, filePath);
+
+        //
+        // A run writes into the output folder itself, never into one of its subfolders:
+        //
+        return this.isInputFolder
+               && Path.GetDirectoryName(Path.GetFullPath(filePath)) is { } fileDirectory
+               && IsSameFolder(fileDirectory, this.fullOutputDirectory)
+               && this.IsOwnOutputArtifact(filePath);
+    }
+
+    /// <summary>
+    /// Checks whether a file is an output artifact of the assistant. We need
+    /// this when the output folder is the input folder: without it, the results
+    /// of a previous run would be processed as documents.
+    /// </summary>
+    /// <remarks>
+    /// A result file is recognized by its name without the extension, e.g., report_result.md or
+    /// report_result_2.docx, cf. AssistantBatchProcessing.CreateResultFileName. Its extension does
+    /// not count, because the user may have chosen another file format since a previous run.
+    /// </remarks>
+    private bool IsOwnOutputArtifact(string filePath)
+    {
+        var fileName = Path.GetFileName(filePath);
+        if (string.Equals(fileName, AssistantBatchProcessing.LOG_FILENAME, StringComparison.OrdinalIgnoreCase)
+            || string.Equals(fileName, resultsFileName, StringComparison.OrdinalIgnoreCase))
+            return true;
+
+        var stem = Path.GetFileNameWithoutExtension(fileName);
+        var suffixIndex = stem.LastIndexOf(AssistantBatchProcessing.RESULT_FILE_SUFFIX, StringComparison.OrdinalIgnoreCase);
+        if (suffixIndex < 0)
+            return false;
+
+        //
+        // A second result of the same name gets a counter, e.g., report_result_2.md:
+        //
+        var counter = stem[(suffixIndex + AssistantBatchProcessing.RESULT_FILE_SUFFIX.Length)..];
+        return counter.Length is 0 || (counter.Length > 1 && counter[0] is '_' && counter[1..].All(char.IsAsciiDigit));
+    }
+
+    private static bool IsSameFolder(string firstDirectory, string secondDirectory) => string.Equals(TrimDirectorySeparator(Path.GetFullPath(firstDirectory)), TrimDirectorySeparator(Path.GetFullPath(secondDirectory)), PathTools.COMPARISON);
+
+    private static string TrimDirectorySeparator(string path) => path.TrimEnd(Path.DirectorySeparatorChar, Path.AltDirectorySeparatorChar);
+}