运行NUnit时，第三方Pdf库明显慢

Question

I am evaluating Winnovative's PdfToText library and have run into something that concerns me.

Everything runs fine and I am able to extract the text content from a small 20k or less pdf immediately if I am running a console application. However, if I call the same code from the NUnit gui running it takes 15-25 seconds (I've verified it's PdfToText by putting a breakpoint on the line that extracts the text and hitting F10 to see how long it takes to advance to the next line).

This concerns me because I'm not sure where to lay blame since I don't know the cause. Is there a problem with NUnit or PdfToText? All I want to do is extract the text from a pdf, but 20 seconds is completely unreasonable if I'm going to see this behavior under certain conditions. If it's just when running NUnit, that's acceptable, but otherwise I'll have to look elsewhere.

It's easier to demonstrate the problem using a complete VS Solution (2010), so here's the link to make it easier to setup and run (no need to download NUnit or PdfToText or even a sample pdf): http://dl.dropbox.com/u/273037/PdfToTextProblem.zip (You may have to change the reference to PdfToText to use the x86 dll if you're running on a 32-bit machine).

Just hit F5 and the NUnit Gui runner will load.

I'm not tied to this library, if you have suggestions, I've tried iTextSharp (way too expensive for 2 lines of code), and looked at Aspose (I didn't try it, but the SaaS license is $11k). But they either lack the required functionality or are way too expensive.

Answer 1

(comment turned into answer)

How complex are your PDFs? The 4.1.6 version of iText allows for a closed sourced solution. Although 4.1.6 doesn't directly have a text extractor it isn't too terribly hard to write one using the PdfReader and GetPageContent().

Answer 2

Below is the code I used to extract the text from the PDF using iTextSharp v4.1.6 . If it seems overly verbose, it's related to how I'm using it and the flexibility required.

using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using iTextSharp.text.pdf;

namespace ClassLibrary1
{
    public class PdfToken
    {
        private PdfToken(int type, string value)
        {
            Type = type;
            Value = value;
        }

        public static PdfToken Create(PRTokeniser tokenizer)
        {
            return new PdfToken(tokenizer.TokenType, tokenizer.StringValue);
        }

        public int Type { get; private set; }
        public string Value { get; private set; }
        public bool IsOperand
        {
            get
            {
                return Type == PRTokeniser.TK_OTHER;
            }
        }
    }

    public class PdfOperation
    {
        public PdfOperation(PdfToken operationToken, IEnumerable<PdfToken> arguments)
        {
            Name = operationToken.Value;
            Arguments = arguments;
        }

        public string Name { get; private set; }
        public IEnumerable<PdfToken> Arguments { get; private set; }
    }

    public interface IPdfParsingStrategy
    {
        void Execute(PdfOperation op);
    }

    public class PlainTextParsingStrategy : IPdfParsingStrategy
    {
        StringBuilder text = new StringBuilder();

        public PlainTextParsingStrategy()
        {

        }

        public String GetText()
        {
            return text.ToString();
        }

        #region IPdfParsingStrategy Members

        public void Execute(PdfOperation op)
        {
            // see Adobe PDF specs for additional operations
            switch (op.Name)
            {
                case "TJ":
                    PrintText(op);
                    break;
                case "Tm":
                    SetMatrix(op);
                    break;
                case "Tf":
                    SetFont(op);
                    break;
                case "S":
                    PrintSection(op);
                    break;
                case "G":
                case "g":
                case "rg":
                    SetColor(op);
                    break;
            }
        }

        #endregion

        bool newSection = false;

        private void PrintSection(PdfOperation op)
        {
            text.AppendLine("------------------------------------------------------------");
            newSection = true;
        }

        private void PrintNewline(PdfOperation op)
        {
            text.AppendLine();
        }

        private void PrintText(PdfOperation op)
        {
            if (newSection)
            {
                newSection = false;
                StringBuilder header = new StringBuilder();
                PrintText(op, header);
            }

            PrintText(op, text);
        }

        private static void PrintText(PdfOperation op, StringBuilder text)
        {
            foreach (PdfToken t in op.Arguments)
            {
                switch (t.Type)
                {
                    case PRTokeniser.TK_STRING:
                        text.Append(t.Value);
                        break;
                    case PRTokeniser.TK_NUMBER:
                        text.Append(" ");
                        break;
                }
            }
        }

        String lastFont = String.Empty;
        String lastFontSize = String.Empty;

        private void SetFont(PdfOperation op)
        {
            var args = op.Arguments.ToList();
            string font = args[0].Value;
            string size = args[1].Value;

            //if (font != lastFont || size != lastFontSize)
            //    text.AppendLine();

            lastFont = font;
            lastFontSize = size;
        }

        String lastX = String.Empty;
        String lastY = String.Empty;

        private void SetMatrix(PdfOperation op)
        {
            var args = op.Arguments.ToList();
            string x = args[4].Value;
            string y = args[5].Value;

            if (lastY != y)
                text.AppendLine();
            else if (lastX != x)
                text.Append(" ");

            lastX = x;
            lastY = y;
        }

        String lastColor = String.Empty;

        private void SetColor(PdfOperation op)
        {
            lastColor = PrintCommand(op).Replace(" ", "_");
        }

        private static string PrintCommand(PdfOperation op)
        {
            StringBuilder text = new StringBuilder();
            foreach (PdfToken t in op.Arguments)
                text.AppendFormat("{0} ", t.Value);
            text.Append(op.Name);
            return text.ToString();
        }

    }
}

And here's how I call it:

using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using System.IO;
using iTextSharp.text.pdf;

namespace ClassLibrary1
{
    public class PdfExtractor
    {
        public static string GetText(byte[] pdfBuffer)
        {
            PlainTextParsingStrategy strategy = new PlainTextParsingStrategy();
            ParsePdf(pdfBuffer, strategy);
            return strategy.GetText();
        }

        private static void ParsePdf(byte[] pdf, IPdfParsingStrategy strategy)
        {
            PdfReader reader = new PdfReader(pdf);

            for (int i = 1; i <= reader.NumberOfPages; i++)
            {
                byte[] page = reader.GetPageContent(i);
                if (page != null)
                {
                    PRTokeniser tokenizer = new PRTokeniser(page);
                    List<PdfToken> parameters = new List<PdfToken>();

                    while (tokenizer.NextToken())
                    {
                        var token = PdfToken.Create(tokenizer);
                        if (token.IsOperand)
                        {
                            strategy.Execute(new PdfOperation(token, parameters));
                            parameters.Clear();
                        }
                        else
                        {
                            parameters.Add(token);
                        }
                    }
                }
            }

        }
    }
}