I am using the trial version of telerik for xamarin for .net core which was released last year, to convert pdf to text. Our service is hosted in Azure. The text which I get back, in certain cases the spaces are missing (say for example instead of 'I [am] here', it displays 'I[am]here'. This happens randomly. 

The code which we have used is as follows -

byte[] pdfBinary = Convert.FromBase64String(inputString);

 TextFormatProviderSettings textFormatProviderSettings = new TextFormatProviderSettings("\r\n", " ");
         
 var textFormatProvider = new TextFormatProvider();
 var pdfFormatProvider = new PdfFormatProvider();

 RadFixedDocument document = pdfFormatProvider.Import(pdfBinary);
 result = textFormatProvider.Export(document, textFormatProviderSettings);

A KeyNotFoundException is thrown when trying to open a PDF containing specific Type1 font.

The values of rotated widgets on a rotated page are invisible after exporting them. The value can be seen only while editing a field.

Add support for documents with an invalid stream cross-reference table

Exception when the endstream is no on a new row 

When merging files that contain the "159 '\u009f'" char, ArgumentException("The encoding is not supported.") is thrown.

Workaround:

Use PdfFormatProvider.

When the bookmarks are declared as objects instead of a string they are not visible.

When a TrueType font is defined, the mapping of character codes to glyph indices depends on the built-in cmap table mappings defined in the font and the Encoding property defined in the PDF dictionary.

However, the current implementation maps all characters with cmap tables for Microsoft Symbolic and Macintosh Roman, which causes incorrect mapping results, e.g. space characters are mapped to an Ê glyph.

The issue is also described in the following public item: TryGetCharCode for OpenTypeFont uses wrong cmap and returns wrong charcode.

Workaround: Change the font  of the TextBoxField's widget appearance:
foreach (var widget in field.Widgets)
{
    widget.TextProperties.Font = FontsRepository.Helvetica;
}

When the content of a page is rotated, each glyph is exported on a new line.

Importing document containing TextBox widgets with NormalContentSource defining text properties, such as FontSize and Position, are exported with incorrect appearance when the value of the field is modified.

Some documents result in glyphs with wrong widths after import-export. Widths are mixed making some of them to overlap and others to have bigger distance.

After signing a document using the PdfProcessing library and the document is loaded in Adobe Acrobat an error is shown when the signature widget is clicked:
Error during signature verification.  

Adobe Acrobat error.
Expected a dict object.

By specification, the InteractiveForm fields can have the same name if they are descendants of a common ancestor. Such fields are different representations of the same underlying field; they should differ only in properties that specify their visual appearance. When such a document is imported an ArgumentException: 'An item with the same key has already been added.' is thrown.

When importing color space defined as an external resource name, a NotSupportedColorSpaceException: 'CS0 color space is not supported.' is thrown.

For example, importing URI Action with invalid mailto URL scheme - mailto:***@***.**(E-mail), throws UriFormatException: 'Invalid URI: The hostname could not be parsed.'

By specification the explicit destinations define a required page property. However, some documents contain definition for destinations with null page value, which causes an ArgumentNullException on import.

When a document is exported with a cross-reference stream, the stream is missing the required according to the specification Length property. Although most of the viewers open the document without any issues, some cannot. Such documents cannot be opened in MS Edge.

The case is invalid according to the specification but all readers manage to open the document and allow editing its fields. The document can be also opened in RadPdfViewer but without the fields and the widgets.

The appearance dictionary specifies how the annotation is presented visually on the page. If the appearance missing or is empty in PDF document NullReferenceException is thrown on import. 

When importing document with predefined ToUnicode CMaps (e.g. Identity-H), an InvalidCastException is thrown with cause: 

Unable to cast object of type 'Telerik.Windows.Documents.Fixed.FormatProviders.Pdf.Model.Types.PdfName' to type 'Telerik.Windows.Documents.Fixed.FormatProviders.Pdf.Model.Elements.CMaps.ToUnicodeCMap'.