Filters/FilterPdf.cs

   1 //
   2 // FilterPdf.cs: Very simplistic PDF filter
   3 //
   4 // Author:
   5 //   Christopher Orr <dashboard@protactin.co.uk>
   6 //
   7 // Copyright 2004 by Christopher Orr
   8 //
   9
  10 using System;
  11 using System.IO;
  12 using System.Diagnostics;
  13
  14 using Beagle.Util;
  15 using Beagle.Daemon;
  16
  17 namespace Beagle.Filters {
  18
  19         public class FilterPdf : Beagle.Daemon.Filter {
  20
  21                 public FilterPdf ()
  22                 {
  23                         AddSupportedFlavor (FilterFlavor.NewFromMimeType ("application/pdf"));
  24                         SnippetMode = true;
  25                 }
  26
  27                 // FIXME: we should have a reasonable failure mode if pdftotext is
  28                 // not installed.
  29
  30                 protected override void DoPullProperties ()
  31                 {
  32                         // create new external process
  33                         Process pc = new Process ();
  34                         pc.StartInfo.FileName = "pdfinfo";
  35                         // FIXME: We probably need to quote special chars in the path
  36                         pc.StartInfo.Arguments = String.Format (" \"{0}\"", FileInfo.FullName);
  37                         pc.StartInfo.RedirectStandardInput = false;
  38                         pc.StartInfo.RedirectStandardOutput = true;
  39                         pc.StartInfo.UseShellExecute = false;
  40                         try {
  41                                 pc.Start ();
  42                         } catch (System.ComponentModel.Win32Exception) {
  43                                 Logger.Log.Warn ("Unable to find pdfinfo in path; PDF file not indexed.");
  44                                 Finished ();
  45                                 return;
  46                         }
  47
  48                         // add pdfinfo's output to pool
  49                         StreamReader pout = pc.StandardOutput;
  50                         string str = null;
  51                         string[] tokens = null;
  52                         string strMetaTag = null;
  53                         bool bKeyword = false;
  54
  55                         while ((str = pout.ReadLine ()) != null) {
  56                                 bKeyword = false;
  57                                 strMetaTag = null;
  58                                 tokens = str.Split (':');
  59                                 if (tokens.Length > 1) {
  60                                         switch (tokens[0]) {
  61                                         case "Title":
  62                                                 strMetaTag = "dc:title";
  63                                                 break;
  64                                         case "Author":
  65                                                 strMetaTag = "dc:author";
  66                                                 break;
  67                                         case "Pages":
  68                                                 strMetaTag = "fixme:page-count";
  69                                                 bKeyword = true;
  70                                                 break;
  71                                         case "Creator":
  72                                                 strMetaTag = "dc:creator";
  73                                                 break;
  74                                         case "Producer":
  75                                                 strMetaTag = "dc:appname";
  76                                                 break;
  77                                         }
  78                                         if (strMetaTag != null) {
  79                                                 if (bKeyword)
  80                                                         AddProperty (Beagle.Property.NewKeyword (strMetaTag,
  81                                                                                                  tokens[1].Trim()));
  82                                                 else
  83                                                         AddProperty (Beagle.Property.New (strMetaTag,
  84                                                                                           tokens[1].Trim()));
  85                                         }
  86
  87                                 }
  88                         }
  89                         pout.Close ();
  90                         pc.WaitForExit ();
  91                         pc.Close ();
  92                 }
  93
  94                 protected override void DoPull ()
  95                 {
  96                         // create new external process
  97                         Process pc = new Process ();
  98                         pc.StartInfo.FileName = "pdftotext";
  99                         // FIXME: We probably need to quote special chars in the path
 100                         pc.StartInfo.Arguments = String.Format ("-nopgbrk -enc UTF-8 \"{0}\" -", FileInfo.FullName);
 101                         pc.StartInfo.RedirectStandardInput = false;
 102                         pc.StartInfo.RedirectStandardOutput = true;
 103                         pc.StartInfo.UseShellExecute = false;
 104                         try {
 105                                 pc.Start ();
 106                         } catch (System.ComponentModel.Win32Exception) {
 107                                 Logger.Log.Warn ("Unable to find pdftotext in path; PDF file not indexed.");
 108                                 Finished ();
 109                                 return;
 110                         }
 111
 112                         // Nice the process so we don't monopolize the CPU as much; we might want to even set this to idle
 113                         pc.PriorityClass = ProcessPriorityClass.BelowNormal;
 114
 115                         // add pdftotext's output to pool
 116                         StreamReader pout = pc.StandardOutput;
 117
 118                         // FIXME:  I don't think this is really required
 119                         // Line by line parsing, however, we have to make
 120                         // sure, that "pdftotext" doesn't output any "New-lines".
 121                         string str;
 122                         while ((str = pout.ReadLine()) != null) {
 123                                 AppendText (str);
 124                                 AppendStructuralBreak ();
 125                         }
 126                         pout.Close ();
 127                         pc.WaitForExit ();
 128                         pc.Close ();
 129                         Finished ();
 130                 }
 131         }
 132 }
 133