DocumentMemoryCache Class

Document memory cache support.

Certain multipage file formats can be slow to load using the default implementation of DocumentFactory. This is especially true for document file formats such as DOCX/DOC, XLSX/XLS RTF, PDF, and TXT.

The Using the OptimizedLoad Functions to Speed Loading Large Files topic contains detailed explanation of optimization techniques used by LEADTOOLS to speed up parsing the images, SVG, and text data of the pages of these types of files.

As explained in the topic, CodecsOptimizedLoadData can contain managed or unmanaged memory, depending on the file format. If a format supports managed data (such as TXT and AFP), the DocumentFactory stores this data into the cache itself and re-uses it whenever a document is loaded from the cache.

If a format can only support unmanaged data (such as DOCX and XSLX), the DocumentFactory is unable to store this type of data in a cache for re-use. The default behavior is for the factory to re-create this data from scratch anytime a document is loaded from the cache, which can degrade performance.

Performance is generally not a problem for desktop-based applications using the document library, such as the Windows Document Viewer or Document Converter. Here, the workflow is to load the document once from a file or URL, use the resulting LEADDocument, and only dispose of it when the user has closed the document.

This, however, can become a problem for applications that require saving the document into the cache and then re-loading it again (such as the Document Service). Here, the workflow is to load the document initially, generate its unmanaged optimized data, and then save it into the cache and dispose the object. Subsequent calls to get the images, SVG, and text data of the pages will have to first load the document from the cache, and thus regenerate the data with every call.

DocumentMemoryCache can be used by the factory to greatly enhance the performance of such documents by keeping the unmanaged memory alive in memory, independent of the document. Subsequent calls to load the document from cache will then use the memory-cached data instead of re-generating it from scratch. The result is quicker loading and parsing of the data of the pages. Naturally, keeping unmanaged data in memory increases the application's resource usage, and so careful consideration must be used when deciding how and when to store this data.

Document Service Workflow 1

A typical client/server application such as the LEADTOOLS Document Viewer/Document Service works as follows:

1. The Document Viewer loads or uploads a document from an external resource, and eventually calls DocumentFactory.LoadFromUri or DocumentFactory.LoadFromFile on the Document Service.

2. DocumentFactory calls RasterCodecs.StartOptimizedLoad and RasterCodecs.GetInformation to obtain information about the source document, such as format and number of pages. This can take some time if the document is very large and complex and is especially true for file formats such as XLSX/XLS.

3. A CodecsOptimizedLoadData object can be obtained at this point. If obtained, the factory stores it internally inside the LEADDocument object.

4. The service then saves this document information in the cache. If CodecsOptimizedLoadData supports managed data, it is saved into the cache as well. This is true for file formats such as TXT and AFP.

5. The service disposes the document with all its data, including the CodecsOptimizedLoadData, if available.

6. The document ID in the cache is returned to the viewer.

7. The viewer builds the skeleton of empty pages and thumbnails.

8. The viewer sends one or more requests asynchronously to the service to obtain pages images, SVG, or text data.

9. Each one of these requests results in loading the document information from the cache using DocumentFactory.LoadFromCache.

10. CodecsOptimizedLoadData is created for the document. If its managed data were stored in the cache, the managed data is loaded and used (the parsing of the complex document structure is not performed).

11. If CodecsOptimizedLoadData supports only unmanaged data, then the data was not stored in the cache. Parsing of the complex document structure must be performed again.

Document Service Workflow 2

DocumentFactory supports caching of the unmanaged data described above in memory. When enabled, the workflow is modified as follows:

1. The Document Viewer loads or uploads a document from an external resource, eventually calls DocumentFactory.LoadFromUri or DocumentFactory.LoadFromFile on the Document Service.

2. DocumentFactory calls RasterCodecs.StartOptimizedLoad and RasterCodecs.GetInformation to obtain information on the source document, such as format and number of pages. This operation may take some time if the document is very large and complex and is especially true for file formats such as XLSX/XLS.

   2.1 New behavior: The previous operation is timed and stored in LEADDocument.LoadDuration.

3. A CodecsOptimizedLoadData object may be obtained at this point, and the factory stores it internally inside the LEADDocument object.

4. The service then saves this document information in the cache. If CodecsOptimizedLoadData supports managed data, it is saved into the cache as well. This is true for file formats such as TXT and AFP.

   4.1 New behavior: If CodecsOptimizedLoadData supports unmanaged data only such as DOCX and XLSX/XLS, then LoadDuration is compared against DocumentMemoryCacheStartOptions.MinimumLoadDuration and if greater, the unmanaged data is stored inside an internal memory cache associated with the document.

5. The service disposes of the document with all its data, including the CodecsOptimizedLoadData if available.

6. The document ID in the cache is returned to the viewer.

7. The viewer builds the skeleton of empty pages and thumbnails.

8. The viewer sends one or more requests asynchronously to the service to obtain pages, images, SVG, or text data.

9. Each one of these requests results in loading the document information from the cache using DocumentFactory.LoadFromCache.

10. CodecsOptimizedLoadData is created for the document. If its managed data were stored in the cache, it is loaded and used (parsing of the complex document structure is not performed.

    10.1 New behavior: Otherwise, the internal memory cache is queries for a CodecsOptimizedLoadData associated with this document, and used if found. Parsing of the complex document structure is not performed again. Otherwise,

11. If CodecsOptimizedLoadData supports only unmanaged data, then it was not stored in the cache. Parsing of the complex document structure must be performed again.

    11.1 New behavior: If step 11 occurred then Step 4.1 is repeated and this CodecsOptimizedLoadData is compared and potentially added to the internal memory cache. Thus, subsequent calls to LoadFromCache will use this data and the increase of speed is obtained again. This scenario happens when the data stored in the memory cache has expired after a specified time of inactivity.

Therefore, using the document memory cache will increase the performance of loading and parsing pages from complex and large document formats at the expense of keeping the unmanaged data in memory at the server side. LEADTOOLS internal testing showed an increase in speed of up to 50 times when loading very large XLSX files in the JavaScript document viewer if document memory cache is used.

Usage

DocumentFactory contains the static property DocumentFactory.DocumentMemoryCache that controls the usage of this feature in the document toolkit.

DocumentMemoryCache usage can be enabled as follows:

• Create an instance of DocumentMemoryCacheStartOptions with the options to use.

• Call DocumentFactory.DocumentMemoryCache.Start.

This will initialize the internal memory cache and clean up timer and all subsequent LoadFromUri, LoadFromFile, and LoadFromCache will switch to "Document Service Workflow 2" described above.

The internal cache contains CodecsOptimizedLoadData objects associated with document IDs. Each of these items also contains a timestamp to mark its last usage. The timestamps are updated whenever the document is "touched". If a specific amount of time passes without any activity on the saved data of a document, it is considered expired and is removed from the cache.

The engine will automatically keep the data of a document alive between calls to LoadFromCache and dispose, thus preventing long-running operations (such as converting a large document to a different format), from triggering the expiry on the data.

The engine will also automatically purge the data of a document when it is deleted from the cache.

DocumentMemoryCacheStartOptions contains the following options:

DocumentMemoryCache contains the following members:

This example will simulate a client loading a small and then a large document from URI and then shows the times used. This sample will produce results similar to the following:

Using memory cache is False
Initial load from leadtools.pdf took ~0 seconds
Is using memory cache is False
Multi-threaded load of all pages took ~0 seconds
Initial load from complex.xlsx took ~10 seconds
Is using memory cache is False 1
Multi-threaded load of all pages took ~13 1 seconds
Using memory cache is True
Initial load from leadtools.pdf took ~0 seconds
Is using memory cache is False
Multi-threaded load of all pages took ~0 seconds
Initial load from complex.xlsx took ~10 seconds
Is using memory cache is True 2
Multi-threaded load of all pages took ~0 2 seconds

Notice how the time it took to load all the pages of complex.xlsx in a multi-threaded code decreased from 13 to almost 0 seconds when memory cache is used.

using Leadtools; 
using Leadtools.Codecs; 
using Leadtools.Document.Writer; 
 
using Leadtools.Document; 
using Leadtools.Caching; 
using Leadtools.Annotations.Engine; 
using Leadtools.Ocr; 
using Leadtools.Barcode; 
using Leadtools.Document.Converter; 
 
public void DocumentMemoryCacheExample() 
{ 
	// The cache we are using 
	FileCache cache = new FileCache(); 
	cache.PolicySerializationMode = CacheSerializationMode.Json; 
	cache.DataSerializationMode = CacheSerializationMode.Json; 
	cache.CacheDirectory = @"c:\cache-dir"; 
 
	// The document files we are using 
	string[] documentFiles = 
	{ 
		// PDF files are very fast to load and will not use memory cache 
		@"C:\LEADTOOLS21\Resources\Images\leadtools.pdf", 
		// Large Excel files are complex and loading may take some time, they could use memory cache 
		@"C:\LEADTOOLS21\Resources\Images\complex.xlsx", 
	}; 
 
	// First process without memory cache and obtain some times 
	LoadAndProcessDocument(documentFiles, cache); 
 
	// Then process with memory cache 
 
	// Document memory cache options to use 
	var documentMemoryCacheStartOptions = new DocumentMemoryCacheStartOptions 
	{ 
		// Use for documents that take more than 2 seconds to load initially 
		MinimumLoadDuration = TimeSpan.FromSeconds(2), 
		// No maximum limit on the number of cache items to keep in memory 
		MaximumItems = 0, 
		// Purse items from the cache if not touched for 60 seconds 
		SlidingExpiration = TimeSpan.FromSeconds(60), 
		// Check for expired items every 60 seconds 
		TimerInterval = TimeSpan.FromSeconds(60) 
	}; 
 
	// Use it 
	DocumentFactory.DocumentMemoryCache.Start(documentMemoryCacheStartOptions); 
 
	// Run again 
	// For the first document, times should be very close to the since this is a PDF document and is very fast (less than MinimumLoadDuration) 
	// For the second document, initial times should be the same, but loading all pages should be much faster 
	LoadAndProcessDocument(documentFiles, cache); 
 
	// Clean up 
	DocumentFactory.DocumentMemoryCache.Stop(); 
} 
 
private static void LoadAndProcessDocument(string[] documentFiles, ObjectCache cache) 
{ 
	Console.WriteLine($"Using memory cache is {DocumentFactory.DocumentMemoryCache.IsStarted}"); 
	string[] documentIds = new string[documentFiles.Length]; 
 
	var stopwatch = new Stopwatch(); 
	TimeSpan elapsed; 
 
	for (var i = 0; i < documentFiles.Length; i++) 
	{ 
		string documentFile = documentFiles[i]; 
		int pageCount; 
 
		// First try without memory cache and obtain some times 
		stopwatch.Restart(); 
		using (LEADDocument document = DocumentFactory.LoadFromFile( 
		   documentFile, 
		   new LoadDocumentOptions 
		   { 
			   Cache = cache 
		   })) 
		{ 
			document.Images.MaximumImagePixelSize = 2048; 
			document.AutoSaveToCache = false; 
			document.AutoDeleteFromCache = false; 
			document.SaveToCache(); 
			documentIds[i] = document.DocumentId; 
			pageCount = document.Pages.Count; 
		} 
		elapsed = stopwatch.Elapsed; 
		Console.WriteLine($"Initial load from {Path.GetFileName(documentFile)} took ~{(int)elapsed.TotalSeconds} seconds"); 
		// Check if it's in the cache 
		Console.WriteLine($"Is using memory cache is {DocumentFactory.DocumentMemoryCache.HasDocument(documentIds[i], false)}"); 
 
		// Next call LoadFromCache and process a page in multiple threads 
		stopwatch.Restart(); 
		LoadAllPagesInThreads(documentIds[i], pageCount, cache); 
		elapsed = stopwatch.Elapsed; 
		Console.WriteLine($"Multi-threaded load of all pages took ~{(int)elapsed.TotalSeconds} seconds"); 
	} 
 
	// Clean up 
	DeleteDocumentsFromCache(documentIds, cache); 
} 
 
private static void LoadAllPagesInThreads(string documentId, int pageCount, ObjectCache cache) 
{ 
	System.Threading.Tasks.Parallel.For( 
	   1, 
	   pageCount + 1, 
	   new System.Threading.Tasks.ParallelOptions { MaxDegreeOfParallelism = 4 }, 
	   (int pageNumber) => 
	   { 
		// Load the document from the cache 
		using (LEADDocument document = DocumentFactory.LoadFromCache( 
		   new LoadFromCacheOptions 
			 { 
				 Cache = cache, 
				 DocumentId = documentId 
			 })) 
		   { 
			// Simulates processing of the page 
			DocumentPage documentPage = document.Pages[pageNumber - 1]; 
			   using (RasterImage image = documentPage.GetImage()) 
			   { 
 
			   } 
		   } 
	   }); 
} 
 
private static void DeleteDocumentsFromCache(string[] documentIds, ObjectCache cache) 
{ 
	foreach (string documentId in documentIds) 
	{ 
		DocumentFactory.DeleteFromCache(new LoadFromCacheOptions 
		{ 
			Cache = cache, 
			DocumentId = documentId, 
		}); 
	} 
} 

Imports Leadtools 
Imports Leadtools.Codecs 
Imports Leadtools.Document.Writer 
Imports Leadtools.Svg 
Imports Leadtools.Document 
Imports Leadtools.Caching 
Imports Leadtools.Annotations.Engine 
Imports Leadtools.Barcode 
Imports Leadtools.Ocr 
Imports LeadtoolsDocumentExamples.LeadtoolsExamples.Common 
Imports Leadtools.Document.Converter 
 
Public Shared Sub DocumentMemoryCacheExample() 
   ' The cache we are using 
   Dim Cache As New FileCache() 
   Cache.PolicySerializationMode = CacheSerializationMode.Json 
   Cache.DataSerializationMode = CacheSerializationMode.Json 
   Cache.CacheDirectory = "c:\cache-dir" 
 
   ' The document files we are using 
   Dim documentFiles() As String = 
   { 
      "C:\LEADTOOLS21\Resources\Images\leadtools.pdf", ' PDF files are very fast to load And will Not use memory cache 
      "C:\LEADTOOLS21\Resources\Images\complex.xlsx" ' Large Excel files are complex And loading may take some time, they could use memory cache 
   } 
 
   ' First process without memory cache And obtain some times 
   LoadAndProcessDocument(documentFiles, Cache) 
 
   ' Then process with memory cache 
 
   ' Document memory cache options to use 
   Dim DocumentMemoryCacheStartOptions As New DocumentMemoryCacheStartOptions With 
   { 
      .MinimumLoadDuration = TimeSpan.FromSeconds(2), ' Use for documents that take more than 2 seconds to load initially 
      .MaximumItems = 0, ' No maximum limit on the number of cache items to keep in memory 
      .SlidingExpiration = TimeSpan.FromSeconds(60), ' Purse items from the cache if Not touched for 60 seconds 
      .TimerInterval = TimeSpan.FromSeconds(60) ' Check for expired items every 60 seconds 
   } 
 
   ' Use it 
   DocumentFactory.DocumentMemoryCache.Start(DocumentMemoryCacheStartOptions) 
 
   ' Run again 
   ' For the first document, times should be very close to the since this Is a PDF document And Is very fast (less than MinimumLoadDuration) 
   ' For the second document, initial times should be the same, but loading all pages should be much faster 
   LoadAndProcessDocument(documentFiles, Cache) 
 
   ' Clean up 
   DocumentFactory.DocumentMemoryCache.Stop() 
End Sub 
 
Private Shared Sub LoadAndProcessDocument(documentFiles() As String, cache As ObjectCache) 
   Console.WriteLine($"Using memory cache is {DocumentFactory.DocumentMemoryCache.IsStarted}") 
   Dim documentIds(documentFiles.Length) As String 
 
   Dim stopwatch As New Stopwatch() 
   Dim elapsed As TimeSpan 
 
   For i As Integer = 0 To documentFiles.Length - 1 
      Dim documentFile As String = documentFiles(i) 
      Dim pageCount As Integer 
 
      ' First try without memory cache and obtain some times 
      stopwatch.Restart() 
      Using document As LEADDocument = DocumentFactory.LoadFromFile( 
         documentFile, 
         New LoadDocumentOptions With 
         { 
            .Cache = cache 
         }) 
         document.Images.MaximumImagePixelSize = 2048 
         document.AutoSaveToCache = False 
         document.AutoDeleteFromCache = False 
         document.SaveToCache() 
         documentIds(i) = document.DocumentId 
         pageCount = document.Pages.Count 
      End Using 
      elapsed = stopwatch.Elapsed 
      Console.WriteLine($"Initial load from {Path.GetFileName(documentFile)} took ~{CInt(elapsed.TotalSeconds)} seconds") 
      ' Check if its in the cache 
      Console.WriteLine($"Is using memory cache is {DocumentFactory.DocumentMemoryCache.HasDocument(documentIds(i), False)}") 
 
      ' Next call LoadFromCache and process a page in multiple threads 
      stopwatch.Restart() 
      LoadAllPagesInThreads(documentIds(i), pageCount, cache) 
      elapsed = stopwatch.Elapsed 
      Console.WriteLine($"Multi-threaded load of all pages took ~{CInt(elapsed.TotalSeconds)} seconds") 
   Next 
 
   ' Clean up 
   DeleteDocumentsFromCache(documentIds, cache) 
End Sub 
 
Private Shared Sub LoadAllPagesInThreads(documentId As String, pageCount As Integer, cache As ObjectCache) 
   System.Threading.Tasks.Parallel.For( 
      1, 
      pageCount + 1, 
      New System.Threading.Tasks.ParallelOptions With {.MaxDegreeOfParallelism = 4}, 
      Sub(pageNumber As Integer) 
         ' Load the document from the cache 
         Using document As LEADDocument = DocumentFactory.LoadFromCache( 
            New LoadFromCacheOptions With 
            { 
               .Cache = cache, 
               .DocumentId = documentId 
            }) 
            ' Simulates processing of the page 
            Dim DocumentPage As DocumentPage = document.Pages(pageNumber - 1) 
            Using image As RasterImage = DocumentPage.GetImage() 
            End Using 
         End Using 
      End Sub 
      ) 
End Sub 
 
Private Shared Sub DeleteDocumentsFromCache(documentIds() As String, cache As ObjectCache) 
   For Each documentId As String In documentIds 
      DocumentFactory.DeleteFromCache(New LoadFromCacheOptions With 
      { 
         .Cache = cache, 
         .DocumentId = documentId 
      }) 
   Next 
End Sub 

Loading Using LEADTOOLS Document Library

Using the OptimizedLoad Functions to Speed Loading Large Files

Reference

DocumentMemoryCache Members

Leadtools.Document Namespace

Loading Using LEADTOOLS Document Library