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                         SafeProcess pc = new SafeProcess ();
  34                         pc.Arguments = new string [] { "pdfinfo", FileInfo.FullName };
  35                         pc.RedirectStandardOutput = true;
  36                         pc.RedirectStandardError = true;
  37
  38                         try {
  39                                 pc.Start ();
  40                         } catch (SafeProcessException e) {
  41                                 Log.Warn (e.Message);
  42                                 Error ();
  43                                 return;
  44                         }
  45
  46                         // add pdfinfo's output to pool
  47                         StreamReader pout = new StreamReader (pc.StandardOutput);
  48                         string str = null;
  49                         string[] tokens = null;
  50                         string strMetaTag = null;
  51                         bool bKeyword = false;
  52
  53                         while ((str = pout.ReadLine ()) != null) {
  54                                 bKeyword = false;
  55                                 strMetaTag = null;
  56                                 tokens = str.Split (':');
  57                                 if (tokens.Length > 1) {
  58                                         switch (tokens[0]) {
  59                                         case "Title":
  60                                                 strMetaTag = "dc:title";
  61                                                 break;
  62                                         case "Author":
  63                                                 strMetaTag = "dc:author";
  64                                                 break;
  65                                         case "Pages":
  66                                                 strMetaTag = "fixme:page-count";
  67                                                 bKeyword = true;
  68                                                 break;
  69                                         case "Creator":
  70                                                 strMetaTag = "dc:creator";
  71                                                 break;
  72                                         case "Producer":
  73                                                 strMetaTag = "dc:appname";
  74                                                 break;
  75                                         }
  76                                         if (strMetaTag != null) {
  77                                                 if (bKeyword)
  78                                                         AddProperty (Beagle.Property.NewUnsearched (strMetaTag,
  79                                                                                                  tokens[1].Trim()));
  80                                                 else
  81                                                         AddProperty (Beagle.Property.New (strMetaTag,
  82                                                                                           tokens[1].Trim()));
  83                                         }
  84
  85                                 }
  86                         }
  87                         pout.Close ();
  88
  89                         // Log any errors or warnings from stderr
  90                         pout = new StreamReader (pc.StandardError);
  91                         while ((str = pout.ReadLine ()) != null)
  92                                 Log.Warn ("pdfinfo [{0}]: {1}", Uri, str);
  93
  94                         pout.Close ();
  95                         pc.Close ();
  96                 }
  97
  98                 protected override void DoPull ()
  99                 {
 100                         // create new external process
 101                         SafeProcess pc = new SafeProcess ();
 102                         pc.Arguments = new string [] { "pdftotext", "-q", "-nopgbrk", "-enc", "UTF-8", FileInfo.FullName, "-" };
 103                         pc.RedirectStandardOutput = true;
 104                         pc.RedirectStandardError = true;
 105
 106                         try {
 107                                 pc.Start ();
 108                         } catch (SafeProcessException e) {
 109                                 Log.Warn (e.Message);
 110                                 Error ();
 111                                 return;
 112                         }
 113
 114                         // add pdftotext's output to pool
 115                         StreamReader pout = new StreamReader (pc.StandardOutput);
 116
 117                         // FIXME:  I don't think this is really required
 118                         // Line by line parsing, however, we have to make
 119                         // sure, that "pdftotext" doesn't output any "New-lines".
 120                         string str;
 121                         while ((str = pout.ReadLine()) != null) {
 122                                 AppendText (str);
 123                                 AppendStructuralBreak ();
 124                                 if (! AllowMoreWords ())
 125                                         break;
 126                         }
 127                         pout.Close ();
 128
 129                         pout = new StreamReader (pc.StandardError);
 130                         while ((str = pout.ReadLine ()) != null)
 131                                 Log.Warn ("pdftotext [{0}]: {1}", Uri, str);
 132
 133                         pout.Close ();
 134                         pc.Close ();
 135                         Finished ();
 136                 }
 137         }
 138 }
 139