Extract Text from PDF in C# (100% .NET)

Introduction

This is a 100% .NET solution to extract text from PDF documents.

Background

Dan Letecky posted a nice code on how to extract text from PDF documents in C# based on PDFBox. Although his solution works well it has a drawback, the size of the required additional libraries is almost 16 MB. Using iTextSharp the size of required additional libraries is only 2.3 MB.

Using the Code

In order to use this solution in your projects, you need to do the following steps:

Add references to itextsharp.dll and SharpZiplib.dll
Add the PDFParser.cs class to your project

Then you can use the newly added class in the following way:

// create an instance of the pdfparser class
PDFParser pdfParser = new PDFParser();
   
// extract the text
String result = pdfParser.ExtractText(pdfFile);

I also created a small console application which uses the class and shows the progress of the conversion. Please keep in mind that if you try to extract text from big PDF files, keeping all the resultant text in memory is not the best solution, in these cases you should write the extracted text to the file after parsing every page.

How Is It Working?

My code is based on the algorithm in C ExtractPDFText. Using iTextSharp's PdfReader class to extract the deflated content of every page, I use a simple function ExtractTextFromPDFBytes to extract the text contents from the deflated page.

Further Improvements

Although the code worked well for me, I didn't find in Adobe's PDF reference how to parse special characters. So if someone knows how to do this, just post it and I will update the class.

History

20^th May, 2006: Initial post

License

This article, along with any associated source code and files, is licensed under The Code Project Open License (CPOL)

About the Author

Zollor

Web Developer

Romania

Member

Article Top

Poor

Excellent

Add a reason or comment to your vote: x
Votes of 3 or less require a comment

Comments and Discussions

You must Sign In to use this message board. (secure sign-in)

Search this forum
	Profile popups Noise Layout Per page

Found bug

MunissoR

1:57 24 Apr '12

In the foreach loop of CheckToken method you are trying access token[1], that is not there if the length of the token is 1 (e.g.: you are checking for ' and ").
Sign In·View Thread·Permalink	5.00/5 (1 vote)

Great Post, Works Great!

Member 2022645

5:13 16 Apr '12

You did everyone a big favor documenting the text parsing; great job!!! People should have done this for iTextSharp long ago. Thank you!!! Nevin House Programmer/Developer
Sign In·View Thread·Permalink

Great post!

Eric Castellon

9:35 9 Apr '12

Great, my vote of 5. In fact i solved a problem with this. Thanks!!!
Sign In·View Thread·Permalink

My vote of 5

brinda roy

23:41 20 Feb '12

good
Sign In·View Thread·Permalink

My vote of 1

mjkhan786

21:35 20 Jan '12

code is not working
Sign In·View Thread·Permalink

how to export data from excel to PDF ?

nimolZero

6:10 28 Aug '11

Dear all, anybody pls point me the way how can i export data from excel to pdf. or export data from datagridview to pdf that support unicode string.. feel is very complicated serveral days ago but can't find any solution.. pls help me
Sign In·View Thread·Permalink

not work

cutithongtin

14:45 1 Aug '11

very simple feedback : it does not work !
Sign In·View Thread·Permalink	2.00/5 (1 vote)

Dosn't work.

sasirekam

19:13 29 Jun '11

I couldn't extract text from pdf. I have just download the coding and include the dll file and PDFSharper file to my project but i didn't get any string in output file. Anybody please help me. Thanks in advance..
Sign In·View Thread·Permalink	2.00/5 (1 vote)

Alternate Solution

kaaskop

3:44 7 May '11

The iTextSharp.dll that is included with this project bombed when I ran the program on a test file that I created with Acrobat X. The latest version of iTextSharp works better. The program itself works sort of with PDF files created with ABBYY, but it does not interpret all the tokens correctly. The result is unwanted spaces within the text. While looking for an explanation of the tokens that are embedded in the stream, I came arcross http://www.java2s.com/Open-Source/CSharp/PDF/iTextSharp/iTextSharp/text/pdf/parser/Catalogparser.htm. It has the source that compiles to a program that not only extracts the text, but also lists the dictionary and content stream. The only drawback is that you have to copy and paste 26 files into Visual Studio since I have not been able to find a download link, but it does what I needed it to do and more.

Re: Alternate Solution

Wizdave05

9:04 2 Feb '12

This code is actually in the latest version of iTextSharp (5.1.3.0) and is much simpler to use than the code in this project (and works more reliably too). Here's a simple class that uses iTextSharp:

using System;
using iTextSharp.text.pdf;
using iTextSharp.text.pdf.parser;
 
public class PdfTextParser
{
    public string ExtractTextFromPDFPage(string pdfFile, int pageNumber)
    {
        PdfReader reader = new PdfReader(pdfFile);
        string text = PdfTextExtractor.GetTextFromPage(reader, pageNumber);
        try { reader.Close(); }
        catch {}
        return text;
    }
}

Re: Alternate Solution

Member 8641242

15:45 13 Feb '12

Wow, thank you! This works perfectly and requires quite a bit less code.
Sign In·View Thread·Permalink

(Solved) Error when reading some document (page missing)

Lord TaGoH

0:13 8 Apr '11

Thanks you very much for your CODE! Thumbs Up | :thumbsup:

you saved my ass on my current project when PDFBox fail to extract the text! Cool | :cool:

I encounter some problem reading some pages of some document anyway
because in the code (method ExtractTextFromPDFBytes)you call:

if (CheckToken(new string[] {"'", "T*", "\""}, previousCharacters))
{
         resultString += "\n";
}

But the CheckToken take for granted that ALL tokens are 2 character long at least

if ((recent[_numberOfCharsToKeep - 3] == token[0]) &&
   (recent[_numberOfCharsToKeep - 2] == token[1]) &&
...

checking if the token is long 1 character or more solve the problem.

you need to change the CheckToken method with this one:

private bool CheckToken(string[] tokens, char[] recent)
        {
            foreach(string token in tokens)
            {
                if (token.Length > 1)
                {
                    if ((recent[_numberOfCharsToKeep - 3] == token[0]) &&
                        (recent[_numberOfCharsToKeep - 2] == token[1]) &&
                        ((recent[_numberOfCharsToKeep - 1] == ' ') ||
                        (recent[_numberOfCharsToKeep - 1] == 0x0d) ||
                        (recent[_numberOfCharsToKeep - 1] == 0x0a)) &&
                        ((recent[_numberOfCharsToKeep - 4] == ' ') ||
                        (recent[_numberOfCharsToKeep - 4] == 0x0d) ||
                        (recent[_numberOfCharsToKeep - 4] == 0x0a))
                        )
                    {
                        return true;
                    }
                }
                else
                {
                    if ((recent[_numberOfCharsToKeep - 2] == token[0]) &&
                        ((recent[_numberOfCharsToKeep - 1] == ' ') ||
                        (recent[_numberOfCharsToKeep - 1] == 0x0d) ||
                        (recent[_numberOfCharsToKeep - 1] == 0x0a)) &&
                        ((recent[_numberOfCharsToKeep - 4] == ' ') ||
                        (recent[_numberOfCharsToKeep - 4] == 0x0d) ||
                        (recent[_numberOfCharsToKeep - 4] == 0x0a))
                        )
                    {
                        return true;
                    }
                }
            }
            return false;
        }

Re: (Solved) Error when reading some document (page missing) [modified]

JBress

7:40 22 Jun '11

Thanks for pointing this out

The CheckToken method seems to find [whitespace][token_chars][whitespace] at the end of recent

So when tokens[i] contains 2 characters (from -4 to -1) :

	(
		(recent[_numberOfCharsToKeep - 4] == ' ')
		|| (recent[_numberOfCharsToKeep - 4] == 0x0d)
		|| (recent[_numberOfCharsToKeep - 4] == 0x0a)
	)
	&& (recent[_numberOfCharsToKeep - 3] == token[0])
	&& (recent[_numberOfCharsToKeep - 2] == token[1])
	&& (
		(recent[_numberOfCharsToKeep - 1] == ' ')
		|| (recent[_numberOfCharsToKeep - 1] == 0x0d)
		|| (recent[_numberOfCharsToKeep - 1] == 0x0a)
	)

But when tokens[i] contains 1 character, it should be (from -3 to -1) :

	(
		(recent[_numberOfCharsToKeep - 3] == ' ')
		|| (recent[_numberOfCharsToKeep - 3] == 0x0d)
		|| (recent[_numberOfCharsToKeep - 3] == 0x0a)
	)
	&& (recent[_numberOfCharsToKeep - 2] == token[0])
	&& (
		(recent[_numberOfCharsToKeep - 1] == ' ')
		|| (recent[_numberOfCharsToKeep - 1] == 0x0d)
		|| (recent[_numberOfCharsToKeep - 1] == 0x0a)
	)

Right ?

modified on Wednesday, June 22, 2011 1:47 PM

New line problem

dejan19dejan19

3:53 4 Jan '11

Hi all, is there any way to know when the new line comes from the pdf file. I gоt txt from pdf file, but I need to know when the lines end. I tried to compare, is this char c = (char) input [i]; is new line, but without success. Is there a way to do it. BR, Dejan
Sign In·View Thread·Permalink

not working on scaned image pdf file

gaurav.ipec

18:52 16 Dec '10

this solution is not working on scand image PDF file ....output is blank text file.
Sign In·View Thread·Permalink	1.50/5 (4 votes)

Re: not working on scaned image pdf file

dr_csci

4:07 10 Jan '11

Indeed it should return a blank file since your scanned pdf document is really an image. There will not be any text data in the PDF for this program to extract. What you are really looking for is OCR: http://en.wikipedia.org/wiki/Optical_character_recognition
Sign In·View Thread·Permalink	3.50/5 (2 votes)

My vote of 5

stefan_lahnor

22:54 25 Nov '10

works great for me
Sign In·View Thread·Permalink	1.00/5 (1 vote)

My vote of 1

ajc

7:16 27 Oct '10

The code segment provided does NOT match the pdfParser class supplied in the source code.
Sign In·View Thread·Permalink	5.00/5 (1 vote)

Support for Unicode strings [modified]

Vasiliy Zverev

6:25 29 Sep '10

PDF stores ASCII strings (one byte per character) as is in parentheses: (string). But Unicode strings (two bytes per character) are stored as hex string in angle brackets: <0073007400720069006e0067>. Each 4 hex digits are one Unicode character.
I have added decoding of such hex strings into ExtractTextFromPDFBytes(). I have added new utility function GetCharFromHex() which converts 4 hex digits to Unicode char. I prepare result string in StringBuilder instead of String to improve performance. And I included support of ANSI characters encoded as octal code (thanks to MrVeloso). I've also removed try/catch in ExtractTextFromPDFBytes() because it masks any program/logic problems and can lead to losing a text without any warning.

#region GetCharFromHex
/// <summary>
/// convert 4 (or less) hex digits to unicode character.
/// Hex digits are stored at the end of previousCharacters.
/// </summary>
private char GetCharFromHex(char[] previousCharacters, int hexDigits)
{
    short code = 0;
    code = Convert.ToInt16(new string(previousCharacters, previousCharacters.Length - hexDigits, hexDigits), 16);
    // if there are less digits than 4, we add zeros to the right (e.g. 'ab' -> 'ab00'). This is done according to PDF specification.
    while (hexDigits < 4)
    {
        code <<= 4;
        hexDigits++;
    }
    return (char)code;
}
#endregion
 
#region ExtractTextFromPDFBytes
/// <summary>
/// This method processes an uncompressed Adobe (text) object 
/// and extracts text.
/// </summary>
/// <param name="input">uncompressed</param>
/// <returns></returns>
public string ExtractTextFromPDFBytes(byte[] input)
{
    if (input == null || input.Length == 0) return "";
 
    StringBuilder resultString = new StringBuilder(4096);
 
    // Flag showing if we are we currently inside a text object
    bool inTextObject = false;
 
    // Flag showing if the next character is literal 
    // e.g. '\\' to get a '\' character or '\(' to get '('
    bool nextLiteral = false;
 
    // () Bracket nesting level. Text appears inside ()
    int bracketDepth = 0;
    // is in hex string: <1234>
    bool inHexString = false;
    // number of hex digits read while in hex string.
    int hexDigits = 0;
 
    // Keep previous chars to get extract numbers etc.:
    char[] previousCharacters = new char[_numberOfCharsToKeep];
    for (int j = 0; j < _numberOfCharsToKeep; j++) previousCharacters[j] = ' ';
 
    for (int i = 0; i < input.Length; i++)
    {
        char c = (char)input[i];
 
        if (inTextObject)
        {
            // Position the text
            if (bracketDepth == 0 && !inHexString)
            {
                if (CheckToken(new string[] { "TD", "Td" }, previousCharacters))
                    resultString.Append("\n\r");
                else
                {
                    if (CheckToken(new string[] { "'", "T*", "\"" /*"*/ /*fix code highlighting on codeproject*/ }, previousCharacters))
                        resultString.Append("\n");
                    else
                    {
                        if (CheckToken(new string[] { "Tj" }, previousCharacters))
                            resultString.Append(' ');
                    }
                }
            }
 
            // End of a text object, also go to a new line.
            if (bracketDepth == 0 && !inHexString &&
                CheckToken(new string[] { "ET" }, previousCharacters))
            {
 
                inTextObject = false;
                if (resultString.Length > 0 && resultString[resultString.Length - 1] != ' ')
                    resultString.Append(' ');
            }
            else
            {
                // Start outputting text
                if ((c == '(') && (bracketDepth == 0) && (!nextLiteral))
                {
                    bracketDepth = 1;
                }
                else
                {
                    // Stop outputting text
                    if ((c == ')') && (bracketDepth == 1) && (!nextLiteral))
                    {
                        bracketDepth = 0;
                    }
                    else
                    {
                        if (c == '<' && !inHexString && bracketDepth == 0)
                        { // Start of hex string
                            inHexString = true;
                            hexDigits = 0;
                        }
                        else
                        {
                            if ((c == '<' || c == '>') && inHexString) // end of hex string e.g. <1234>
                                // or this is not hex string but dictionary e.g. <<dict>>
                            {
                                if (hexDigits > 0) // convert last char
                                    resultString.Append(GetCharFromHex(previousCharacters, hexDigits));
                                inHexString = false;
                            }
                            else
                                // Just a normal text character:
                                if (bracketDepth == 1)
                                {
                                    if (c == '\\' && !nextLiteral)
                                    { // start of escaped character
                                        nextLiteral = true;
                                    }
                                    else
                                    {
                                        if (nextLiteral && c >= '0' && c <= '9')
                                        { // character is encoded in octal code
                                            char[] octalCode = {(char)input[i], (char)input[++i], (char)input[++i]};
                                            c = (char)Convert.ToInt32(new string(octalCode), 8);
                                        }
                                        
                                        if (((c >= ' ') && (c <= '~')) ||
                                            ((c >= 128) && (c < 255)))
                                        {
                                            resultString.Append(c);
                                        }
 
                                        nextLiteral = false;
                                    }
                                }
                                else
                                    if (inHexString)
                                    {
                                        if (hexDigits == 4)
                                        { // ready to extract next unicode character (4 hex digits)
                                            resultString.Append(GetCharFromHex(previousCharacters, hexDigits));
                                            hexDigits = 0;
                                        }
                                        hexDigits++; // new hex digit c is not added to previousCharacters yet, but it will be added below,
                                            // so I increment hexDigits after the check for 4.
                                    }
                        }
                    }
                }
            }
        }
 
        // Store the recent characters for 
        // when we have to go back for a checking
        for (int j = 0; j < _numberOfCharsToKeep - 1; j++)
        {
            previousCharacters[j] = previousCharacters[j + 1];
        }
        previousCharacters[_numberOfCharsToKeep - 1] = c;
 
        // Start of a text object
        if (!inTextObject && CheckToken(new string[] { "BT" }, previousCharacters))
        {
            inTextObject = true;
        }
    }
    return resultString.ToString();
}
#endregion

modified on Wednesday, September 29, 2010 12:47 PM

Re: Support for Unicode strings

gulak

23:19 28 Jan '11

I can't get this work with my PDF documents.

For example PDF which has content like below return very strange characters. Could this be character set problem or something?

BT
308.8 521.8 Td /F1 10 Tf[<1110>7<10>-3<0E>4<12>]TJ
ET
Q
q 0 0 0 rg
BT
344.8 510.4 Td /F1 10 Tf[<2E>4<08>3<01>-3<08>3<16>1<04>2<0B17>-4<0C>3<04>2<18>-5<0C>-17<15>1<15>1<1B>-1<27>-2<02>9<2B>-5<1C>]TJ
ET
Q
q 0 0 0 rg
BT
92.9 489.8 Td /F1 14 Tf[<29>1<0C>-3<07>3<18>-1<04>-1<24>-4<04>-1<04>-1<18>-1<18>-1<06>3<15>-2<27>1<2F>1<02>-2<1B>]TJ
ET
Q
q 0 0 0 rg
BT
92.9 461.7 Td /F1 10 Tf[<0D>3<0E>4<0F>-4<10>-3<1112>]TJ

Re: Support for Unicode strings

Vasiliy Zverev

9:02 29 Jan '11

The project from this article cannot parse TJ operator (do not mix up with Tj). Does it return anything for you?
Sign In·View Thread·Permalink

Re: Support for Unicode strings

gulak

3:47 31 Jan '11

It returns very strange characters like hebrew or something.. Do you know how the code should be modified that it could work with these type of PDF's also?
Sign In·View Thread·Permalink

text position

elinfo1

1:51 9 Sep '10

how to get the correct position (X,Y in pixels) of each 'word' or 'char' of the extracted text ?
Sign In·View Thread·Permalink	1.00/5 (1 vote)

char " - ascii (34)

doomelo

23:54 30 Aug '10

This sample have problem with extract text from pdf when exist char " - ascii (34) in this file. Any suggestions ?
Sign In·View Thread·Permalink	2.00/5 (1 vote)

[My vote of 2] Techinchal mistake

Umair Aslam Bhatti

21:52 30 Aug '10

String result = pdfParser.ExtractText (path); The line is totally wrong as your method pdfparser.extracttext return boolean , and there it is written console.writeline Doesn't work for me.
Sign In·View Thread·Permalink	5.00/5 (1 vote)