Skip to content

armgit / Filee

Public Imported from GitHub
0

Free, open-source file converter. Convert 263 formats offline by drag & drop — images, PDF, Word, Excel, PowerPoint, HWP/HWPX, video, audio, e-books and archives. No upload, no Office needed.

OCR: other PDF options, non-ASCII paths, cancel (review of #429)

cre8ive-marketing327224b4e594
diff --git a/src/Filee.App/ViewModels/PresetEditorViewModel.cs b/src/Filee.App/ViewModels/PresetEditorViewModel.cs
index 4cdddb97..1fee4e9b 100644
--- a/src/Filee.App/ViewModels/PresetEditorViewModel.cs
+++ b/src/Filee.App/ViewModels/PresetEditorViewModel.cs
@@ -214,7 +214,7 @@ public sealed partial class PresetEditorViewModel : ObservableObject
[ObservableProperty]
[NotifyPropertyChangedFor(nameof(ShowImage), nameof(ShowTurnForPdf), nameof(ShowQuality), nameof(ShowTiff), nameof(ShowMultiPageTiff), nameof(ShowWebp), nameof(ShowIco),
nameof(ShowZipResult), nameof(ShowPdf), nameof(ShowPdfBuild), nameof(ShowDocument), nameof(ShowLeaveOutPictures), nameof(ShowPdfSplit), nameof(ShowPdfCombine),
- nameof(ShowPdfCompression), nameof(ShowPdfCompressionOptions), nameof(ShowOcr),
+ nameof(ShowPdfCompression), nameof(ShowPdfCompressionOptions), nameof(ShowOcr), nameof(ShowPageSize),
nameof(ShowMedia), nameof(ShowMediaQuality), nameof(ShowMaxHeight), nameof(ShowAudioBitrate), nameof(ShowRemoveAudio),
nameof(ShowAudioChannels), nameof(ShowWav), nameof(VideoCodecs), nameof(VideoCodecChoice), nameof(ShowVideoCodec), nameof(ShowHardwareEncoding),
nameof(AudioCodecs), nameof(AudioCodecChoice), nameof(ShowAudioCodec),
@@ -400,7 +400,15 @@ public sealed partial class PresetEditorViewModel : ObservableObject
[ObservableProperty] private bool _compressWithGhostscript;
/// <summary>Recognize the text of scans and pictures (PdfOptions.Ocr), in <see cref="OcrLanguage"/>.</summary>
- [ObservableProperty] private bool _ocr;
+ [ObservableProperty]
+ [NotifyPropertyChangedFor(nameof(ShowPdfSplit), nameof(ShowPdfCompression), nameof(ShowPdfCompressionOptions), nameof(ShowPageSize))]
+ private bool _ocr;
+
+ /// <summary>
+ /// Text recognition makes the PDF itself (Tesseract): one file with each page the size of its scan or picture, so one
+ /// file per page, compression (and with it removing metadata), page size and margin don't apply and are hidden.
+ /// </summary>
+ private bool Recognizes => ShowOcr && Ocr;
[ObservableProperty] private Choice<string> _ocrLanguage;
/// <summary>PDF → image: one tall image of all pages.</summary>
@@ -508,10 +516,13 @@ public sealed partial class PresetEditorViewModel : ObservableObject
/// <summary>"Save as ZIP" is offered for geospatial targets, whose results are several files (a Shapefile) or travel zipped.</summary>
public bool ShowZipResult => TargetCategory == FormatCategory.Geo;
public bool ShowPdfBuild => TargetId == "pdf";
- public bool ShowPdfSplit => TargetId == "pdf";
+
+ /// <summary>Page size and margin of a PDF built from pictures; not with text recognition (<see cref="Recognizes"/>).</summary>
+ public bool ShowPageSize => ShowPdfBuild && !Recognizes;
+ public bool ShowPdfSplit => TargetId == "pdf" && !Recognizes;
/// <summary>Compression is for PDF → PDF; Ghostscript and metadata only matter while it is on.</summary>
- public bool ShowPdfCompression => TargetId == "pdf";
+ public bool ShowPdfCompression => TargetId == "pdf" && !Recognizes;
public bool ShowPdfCompressionOptions => ShowPdfCompression && CompressionLevel.Value != PdfCompression.None;
/// <summary>Text from scans goes into the formats Filee writes from a PDF's text itself (PdfOptions.OcrTargets).</summary>
@@ -624,8 +635,17 @@ public sealed partial class PresetEditorViewModel : ObservableObject
pdf.Compression = CompressionLevel.Value;
pdf.CompressWithGhostscript = CompressWithGhostscript;
// Only where the option is shown, as with SaveEveryFrame below.
- pdf.Ocr = ShowOcr && Ocr;
+ pdf.Ocr = Recognizes;
pdf.OcrLanguage = OcrLanguage.Value;
+ if (pdf.Ocr)
+ {
+ // Hidden while text is recognized, so nothing is kept behind it that the PDF wouldn't do (Recognizes).
+ pdf.SplitPages = false;
+ pdf.Compression = PdfCompression.None;
+ pdf.CompressWithGhostscript = false;
+ pdf.PageSize = PdfPageSize.FitImage;
+ pdf.MarginMm = 0;
+ }
_original.Document.PdfA = PdfA;
// Only where the option is shown, as with SaveEveryFrame below.
_original.Document.LeaveOutPictures = ShowLeaveOutPictures && LeaveOutPictures;
diff --git a/src/Filee.App/Views/PresetEditorView.axaml b/src/Filee.App/Views/PresetEditorView.axaml
index 8e105efd..81a1db4d 100644
--- a/src/Filee.App/Views/PresetEditorView.axaml
+++ b/src/Filee.App/Views/PresetEditorView.axaml
@@ -355,7 +355,7 @@
<TextBlock Classes="caption" TextWrapping="Wrap" Margin="28,0,0,0" Text="{DynamicResource presets.split_tall_image_hint}" />
</StackPanel>
</StackPanel>
- <Grid ColumnDefinitions="*,12,*" IsVisible="{Binding ShowPdfBuild}">
+ <Grid ColumnDefinitions="*,12,*" IsVisible="{Binding ShowPageSize}">
<StackPanel>
<TextBlock Classes="label" Text="{DynamicResource presets.page_size}" />
<ComboBox ItemsSource="{Binding PageSizes}" SelectedItem="{Binding PageSize}" HorizontalAlignment="Stretch" />
diff --git a/src/Filee.Core/Presets/BuiltInData.cs b/src/Filee.Core/Presets/BuiltInData.cs
index af5cd208..e6fe44ac 100644
--- a/src/Filee.Core/Presets/BuiltInData.cs
+++ b/src/Filee.Core/Presets/BuiltInData.cs
@@ -197,7 +197,7 @@ public static class BuiltInData
];
/// <summary>
- /// "Searchable PDF": scans and pictures with their text recognized by Tesseract (added in settings schema 33). On no
+ /// "Searchable PDF": scans and pictures with their text recognized by Tesseract (added in settings schema 40). On no
/// donut, like "Compress PDF".
/// </summary>
public static List<Preset> OcrPresets() =>
diff --git a/src/Filee.Engines/Ocr/TesseractConverter.cs b/src/Filee.Engines/Ocr/TesseractConverter.cs
index a5c8f9f5..89c5a09f 100644
--- a/src/Filee.Engines/Ocr/TesseractConverter.cs
+++ b/src/Filee.Engines/Ocr/TesseractConverter.cs
@@ -92,7 +92,7 @@ public sealed class TesseractConverter : IConverter
var tessdata = Path.Combine(scratch, "tessdata");
Directory.CreateDirectory(tessdata);
File.Copy(data, Path.Combine(tessdata, Path.GetFileName(data)));
- var ocr = new Run(tesseract, tessdata, language.Code);
+ var ocr = new Run(tesseract, scratch, language.Code);
progress?.Report(0.05);
var output = step.Output.Allocate("pdf");
@@ -104,9 +104,11 @@ public sealed class TesseractConverter : IConverter
}
else
{
- var target = Path.Combine(scratch, "page");
- await ocr.ToPdfAsync(step.InputPath, target, HasResolution(step.InputPath) ? null : Dpi, cancellationToken);
- File.Move(target + ".pdf", output, overwrite: true);
+ // A copy with a plain name: the picture's own folder and name may not be readable for Tesseract (Run).
+ var picture = "picture" + Path.GetExtension(step.InputPath).ToLowerInvariant();
+ File.Copy(step.InputPath, Path.Combine(scratch, picture));
+ await ocr.ToPdfAsync(picture, "page", HasResolution(step.InputPath) ? null : Dpi, cancellationToken);
+ File.Move(Path.Combine(scratch, "page.pdf"), output, overwrite: true);
}
progress?.Report(1);
return [output];
@@ -136,6 +138,8 @@ public sealed class TesseractConverter : IConverter
using var free = new SemaphoreSlim(Runs);
var runs = new List<Task>();
var done = 0;
+ try
+ {
foreach (var page in scans)
{
await free.WaitAsync(cancellationToken);
@@ -154,14 +158,13 @@ public sealed class TesseractConverter : IConverter
catch
{
free.Release();
- await Task.WhenAll(runs).ContinueWith(_ => { }, TaskScheduler.Default); // let the running pages end first
throw;
}
runs.Add(Task.Run(async () =>
{
try
{
- await ocr.ToPdfAsync(picture, Path.Combine(scratch, $"p{page}"), Dpi, cancellationToken);
+ await ocr.ToPdfAsync($"p{page}.png", $"p{page}", Dpi, cancellationToken);
File.Delete(picture);
progress?.Report(0.05 + 0.9 * Interlocked.Increment(ref done) / scans.Count);
}
@@ -171,15 +174,25 @@ public sealed class TesseractConverter : IConverter
}
}, cancellationToken));
}
+ }
+ finally
+ {
+ // Rendering stopped (a failed page, a cancel): the running pages end before the work folder can go.
+ await Task.WhenAll(runs).ContinueWith(_ => { }, TaskScheduler.Default);
+ }
await Task.WhenAll(runs);
- // A new document: the pages and the document information come along, bookmarks and form fields don't.
+ // A new document: the pages and the document information come along (not when the preset removes metadata),
+ // bookmarks and form fields don't.
using var source = PdfSharpConverter.Open(step.InputPath, PdfDocumentOpenMode.Import);
using var result = new PdfDocument();
- result.Info.Title = source.Info.Title;
- result.Info.Author = source.Info.Author;
- result.Info.Subject = source.Info.Subject;
- result.Info.Keywords = source.Info.Keywords;
+ if (!step.Preset.Pdf.RemoveMetadata)
+ {
+ result.Info.Title = source.Info.Title;
+ result.Info.Author = source.Info.Author;
+ result.Info.Subject = source.Info.Subject;
+ result.Info.Keywords = source.Info.Keywords;
+ }
var recognized = new List<PdfDocument>();
try
{
@@ -254,18 +267,22 @@ public sealed class TesseractConverter : IConverter
}
}
- /// <summary>One language, one Tesseract.</summary>
- private sealed record Run(string Executable, string Tessdata, string Language)
+ /// <summary>
+ /// One language, one Tesseract, working in <paramref name="Folder"/> (the job's scratch folder, with the language in
+ /// its <c>tessdata</c>). Tesseract only ever sees plain relative names there: some Windows builds of Tesseract and
+ /// Leptonica open files through the ANSI code page and fail on a user name or folder in Korean or Cyrillic.
+ /// </summary>
+ private sealed record Run(string Executable, string Folder, string Language)
{
- /// <summary>Writes <c>{outputBase}.pdf</c>.</summary>
+ /// <summary>Reads <paramref name="input"/> and writes <c>{outputBase}.pdf</c>, both names in <see cref="Folder"/>.</summary>
public async Task ToPdfAsync(string input, string outputBase, int? dpi, CancellationToken cancellationToken)
{
// tessedit_create_pdf instead of the "pdf" config file, which lives in Tesseract's own tessdata folder.
- List<string> args = [input, outputBase, "--tessdata-dir", Tessdata, "-l", Language, "-c", "tessedit_create_pdf=1"];
+ List<string> args = [input, outputBase, "--tessdata-dir", "tessdata", "-l", Language, "-c", "tessedit_create_pdf=1"];
if (dpi is { } value)
args.AddRange(["--dpi", value.ToString(System.Globalization.CultureInfo.InvariantCulture)]);
- var result = await ProcessRunner.RunAsync(Executable, args, Timeout, cancellationToken, environment: Environment);
- if (result.ExitCode != 0 || !File.Exists(outputBase + ".pdf"))
+ var result = await ProcessRunner.RunAsync(Executable, args, Timeout, cancellationToken, workingDirectory: Folder, environment: Environment);
+ if (result.ExitCode != 0 || !File.Exists(Path.Combine(Folder, outputBase + ".pdf")))
throw new InvalidOperationException($"Tesseract failed (exit {result.ExitCode}). {result.StandardError.Trim()}");
}
}
diff --git a/tests/Filee.App.Tests/RenderTests.cs b/tests/Filee.App.Tests/RenderTests.cs
index cffc0059..eb6346d9 100644
--- a/tests/Filee.App.Tests/RenderTests.cs
+++ b/tests/Filee.App.Tests/RenderTests.cs
@@ -1194,6 +1194,36 @@ public class RenderTests
Assert.False(preset.Pdf.Ocr);
}
+ [AvaloniaFact]
+ public void Preset_editor_hides_and_clears_what_text_recognition_does_not_do()
+ {
+ TestServices.EnsureInitialized("en");
+ var loc = AppHost.Get<Filee.Core.Localization.ILocalizer>();
+ var preset = new Filee.Core.Presets.Preset
+ {
+ TargetFormat = "pdf", Name = "Scans",
+ Pdf = { SplitPages = true, Compression = Filee.Core.Presets.PdfCompression.Screen, PageSize = Filee.Core.Presets.PdfPageSize.A4, MarginMm = 10 },
+ };
+ var vm = new PresetEditorViewModel(preset, loc);
+ Assert.True(vm.ShowPdfSplit && vm.ShowPdfCompression && vm.ShowPageSize);
+
+ vm.Ocr = true;
+
+ Assert.False(vm.ShowPdfSplit);
+ Assert.False(vm.ShowPdfCompression);
+ Assert.False(vm.ShowPdfCompressionOptions);
+ Assert.False(vm.ShowPageSize);
+ Assert.True(vm.Apply());
+ var pdf = preset.Pdf;
+ Assert.True(pdf.Ocr);
+ Assert.Equal((false, Filee.Core.Presets.PdfCompression.None, Filee.Core.Presets.PdfPageSize.FitImage, 0.0),
+ (pdf.SplitPages, pdf.Compression, pdf.PageSize, pdf.MarginMm));
+
+ // Without it they are back.
+ vm.Ocr = false;
+ Assert.True(vm.ShowPdfSplit && vm.ShowPdfCompression && vm.ShowPageSize);
+ }
+
[AvaloniaFact]
public void Preset_editor_keeps_rotation_visible_for_pdf_and_drops_it_for_targets_that_cannot_turn()
{
diff --git a/tests/Filee.Engines.Tests/OcrTests.cs b/tests/Filee.Engines.Tests/OcrTests.cs
index b8582e33..ea0d514b 100644
--- a/tests/Filee.Engines.Tests/OcrTests.cs
+++ b/tests/Filee.Engines.Tests/OcrTests.cs
@@ -149,6 +149,51 @@ public class OcrTests(EngineFixture fx) : IClassFixture<EngineFixture>
Assert.Contains("searchable", text);
}
+ [Fact]
+ public async Task Pictures_and_scans_with_korean_and_cyrillic_names_are_recognized()
+ {
+ Assert.SkipUnless(Installed, Skip);
+ var folder = Path.Combine(fx.NewFolder(), "김민수 Документы");
+ Directory.CreateDirectory(folder);
+ var picture = Scan(folder, "스캔.png", "Filee reads scanned pages");
+ var scanPdf = (await fx.ConvertAsync([Scan(folder, "Скан.png", "The quarterly report is ready")], new Preset { TargetFormat = "pdf" })).Outputs.Single();
+
+ var job = await fx.ConvertAsync([picture, scanPdf], Ocr("pdf"));
+
+ AssertDone(job);
+ Assert.Contains("scanned", PageText(job.Outputs.Single(o => Path.GetFileName(o).StartsWith("스캔", StringComparison.Ordinal)), 1));
+ Assert.Contains("quarterly", PageText(job.Outputs.Single(o => Path.GetFileName(o).StartsWith("Скан", StringComparison.Ordinal)), 1));
+ }
+
+ [Fact]
+ public async Task Removing_metadata_leaves_the_document_information_behind()
+ {
+ Assert.SkipUnless(Installed, Skip);
+ var folder = fx.NewFolder();
+ var scanPdf = (await fx.ConvertAsync([Scan(folder, "scan.png", "Private letter")], new Preset { TargetFormat = "pdf" })).Outputs.Single();
+ using (var document = PdfReader.Open(scanPdf, PdfDocumentOpenMode.Modify))
+ {
+ document.Info.Title = "Secret title";
+ document.Info.Author = "Someone";
+ document.Save(scanPdf);
+ }
+ var keep = Ocr("pdf");
+ var remove = Ocr("pdf");
+ remove.Pdf.RemoveMetadata = true;
+
+ var kept = (await fx.ConvertAsync([scanPdf], keep)).Outputs.Single();
+ var removed = (await fx.ConvertAsync([scanPdf], remove)).Outputs.Single();
+
+ using (var document = PdfReader.Open(kept, PdfDocumentOpenMode.Import))
+ Assert.Equal("Secret title", document.Info.Title);
+ using (var document = PdfReader.Open(removed, PdfDocumentOpenMode.Import))
+ {
+ Assert.Equal("", document.Info.Title);
+ Assert.Equal("", document.Info.Author);
+ }
+ Assert.Contains("Private", PageText(removed, 1));
+ }
+
[Fact]
public async Task A_scanned_pdf_becomes_a_docx_with_its_text_and_without_the_scan()
{