new TextExtractor()
TextExtractor is used to analyze a PDF page and extract words and logical structure within a given region. The resulting list of lines and words can be traversed element by element or accessed as a string buffer. The class also includes utility methods to extract PDF text as HTML or XML.
Possible use case scenarios for TextExtractor include:
- Converting PDF pages to text or XML for content repurposing.
- Searching PDF pages for specific words or keywords.
- Indexing large PDF repositories for indexing or content retrieval purposes (i.e. implementing a PDF search engine).
- Classifying or summarizing PDF documents based on their text content.
- Finding specific words for content editing purposes (such as splitting pages based on keywords etc).
The main task of TextExtractor is to interpret PDF pages and offer a simple to use API to:
- Normalize all text content to Unicode.
- Extract inferred logical structure (word by word, line by line, or paragraph by paragraph).
- Extract positioning information for every line, word, or a glyph.
- Extract style information (such as information about the font, font size, font styles, etc) for every line, word, or a glyph.
- Control the content analysis process. A number of options (such as removal of text obscured by images) is available to let the user direct the flow of content recognition algorithms that will meet their requirements.
- Offer utility methods to convert PDF page content to text, XML, or HTML.
Note: TextExtractor is analyzing only textual content of the page. This means that the rasterized (e.g. in scanned pages) or vectorized text (where glyphs are converted to path outlines) will not be recognized as text. Please note that it is still possible to extract this content using pdftron.PDF.ElementReader interface.
In some cases TextExtractor may extract text that does not appear to be on the visible page (e.g. when text is obscured by an image or a rectangle). In these situations it is possible to use processing flags such as 'e_remove_hidden_text' and 'e_no_invisible_text' to remove hidden text.
A sample use case (in C++):
... Initialize PDFNet ... PDFDoc doc(filein); doc.InitSecurityHandler(); Page page = *doc.PageBegin(); TextExtractor txt; txt.Begin(page, 0, TextExtractor::e_remove_hidden_text); UString text; txt.GetAsText(text); // or traverse words one by one... TextExtractor::Line line = txt.GetFirstLine(), lend; TextExtractor::Word word, wend; for (; line!=lend; line=line.GetNextLine()) { for (word=line.GetFirstWord(); word!=wend; word=word.GetNextWord()) { text.Assign(word.GetString(), word.GetStringLen()); cout << text << '\n'; } }
A sample use case (in C#):
... Initialize PDFNet ... PDFDoc doc = new PDFDoc(filein); doc.InitSecurityHandler(); Page page = doc.PageBegin().Current(); TextExtractor txt = new TextExtractor(); txt.Begin(page, 0, TextExtractor.ProcessingFlags.e_remove_hidden_text); string text = txt.GetAsText(); // or traverse words one by one... TextExtractor.Word word; for (TextExtractor.Line line = txt.GetFirstLine(); line.IsValid(); line=line.GetNextLine()) { for (word=line.GetFirstWord(); word.IsValid(); word=word.GetNextWord()) { Console.WriteLine(word.GetString()); } }
For full sample code, please take a look at TextExtract sample project.
Methods
-
<static> create()
-
Constructor and destructor
Returns:
A promise that resolves to an object of type: "TextExtractor"
- Type
- PDFNet.TextExtractor
-
begin(page, clip_ptr, flags)
-
Start reading the page.
Parameters:
Name Type Description page
PDFNet.Page Page to read.
clip_ptr
PDFNet.rect A pointer to the optional clipping rectangle. This parameter can be used to selectively read text from a given rectangle.
flags
number A list of ProcessingFlags used to control text extraction algorithm.
-
destroy()
-
Frees the native memory of the object.
-
getAsText(dehyphen)
-
get all words in the current selection as a single string.
Parameters:
Name Type Description dehyphen
boolean If true, finds and removes hyphens that split words across two lines. Hyphens are often used a the end of lines as an indicator that a word spans two lines. Hyphen detection enables removal of hyphen character and merging of text runs to form a single word. This option has no effect on Tagged PDF files.
Returns:
A promise that resolves to an object of type: "string"
- Type
- string
-
getAsXML(xml_output_flags)
-
get text content in a form of an XML string.
Parameters:
Name Type Description xml_output_flags
number flags controlling XML output. For more information, please see TextExtract::XMLOutputFlags.
XML output will be encoded in UTF-8 and will have the following structure:
PDFNet SDK is ...The above XML output was generated by passing the following union of flags in the call to GetAsXML(): (TextExtractor::e_words_as_elements | TextExtractor::e_output_bbox | TextExtractor::e_output_style_info)
In case 'xml_output_flags' was not specified, the default XML output would look as follows:
<Page num="1 crop_box="0, 0, 612, 792" media_box="0, 0, 612, 792" rotate="0">
PDFNet SDK is an amazingly comprehensive, high-quality PDF developer toolkit... levels. Using the PDFNet PDF library, ... ...
Returns:
A promise that resolves to an object of type: "string"
- Type
- string
-
getFirstLine()
-
Returns:
A promise that resolves to the first line of text on the selected page.
- Type
- PDFNet.textextractorline
-
getNumLines()
-
Returns:
A promise that resolves to the number of lines of text on the selected page.
- Type
- number
-
getQuads(mtx, quads, quads_size)
-
[CURRENTLY BUGGED]
Parameters:
Name Type Description mtx
PDFNet.matrix2d The quadrilateral representing a tight bounding box
quads
number n
quads_size
number n for this word (in unrotated page coordinates).
-
getRightToLeftLanguage()
-
Returns:
A promise that resolves to the directionality of text extractor.
- Type
- boolean
-
getTextUnderAnnot(annot)
-
Get all the characters that intersect an annotation.
Parameters:
Name Type Description annot
PDFNet.Annot The annotation to intersect with.
Returns:
A promise that resolves to an object of type: "string"
- Type
- string
-
getWordCount()
-
Returns:
A promise that resolves to the number of words on the page.
- Type
- number
-
setRightToLeftLanguage(rtl)
-
Sets the directionality of text extractor. Must be called before the processing of a page started.
Parameters:
Name Type Description rtl
boolean mode reverses the directionality of TextExtractor algorithm.