tris/recoll
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Improve rcldoc filter and switch back to using it for indexing instead of direct antiword exec. This is slightly slower but it does catch a number of .doc files which would not be indexed otherwise

Browse Source
This commit is contained in:
Jean-Francois Dockes 2011-05-10 09:03:13 +02:00
parent ce607032fa
commit dd8f42253c
5 changed files with 49 additions and 17 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

43
src/filters/rcldoc
View File

@ -102,15 +102,18 @@ checkcmds awk antiword iconv
$decoder "$infile" |
awk 'BEGIN'\
' {
print "<html><head><title></title>"
print "<meta http-equiv=\"Content-Type\" content=\"text/html;charset=UTF-8\">"
print "</head>\n<body>\n<p>"
cont = ""
gotdata = 0
}
{
if (!($0 ~ /^[ ]*$/))
if (!($0 ~ /^[ ]*$/)) {
gotdata = 1
if (gotdata == 0) {
print "<html><head><title></title>"
print "<meta http-equiv=\"Content-Type\" content=\"text/html;charset=UTF-8\">"
print "</head>\n<body>\n<p>"
}
}
$0 = cont $0
cont = ""
@ -139,12 +142,32 @@ END {
print "</p></body></html>"
}'
# wvWare is much slower and we don't use it by default, but it handles
# some files that antiword won't, so use it as a last resort. Yes the
# html document will have a second header section, but this doesn't
# seem to be an issue with our brave html input handler...
# Antiword rarely fails, we try to catch the most common reasons:
if test $? -eq 1 ; then
if iscmd wvWare ; then
wvWare --nographics --charset=utf-8 "$infile"
# Check actual document type
mtype=`file -b -i "$infile" | awk '{sub(";", "", $1);print $1}'`
if test X"$mtype" = Xtext/rtf; then
# RTF document disguising as msword either because it has a .doc
# extension or because it's an attachment with a wrong mime.
exec `dirname $0`/rclrtf $infile
fi
if test X"$mtype" = Xtext/plain; then
# Someone gave a .doc ext to their texts. Happens...
exec `dirname $0`/rcltext $infile
fi
if test X"$mtype" = Xapplication/msword; then
# Actually application/msword: try wvWare, which is much
# slower and we don't use it by default, but it handles some
# files that antiword won't, so use it as a last resort.
if iscmd wvWare ; then
exec wvWare --nographics --charset=utf-8 "$infile"
fi
fi
# else let the error be...
exit 1
fi

12
src/sampleconf/mimeconf
View File

@ -29,10 +29,14 @@ application/x-bzip2 = uncompress rcluncomp bunzip2 %f %t
# each filter, see the exemples below (ie: msword)
[index]
# MSWORD: install wvware and restore the rcldoc version to fix the "text
# stream of this file is too small to handle" error if it's a problem for you
# application/msword = exec rcldoc
application/msword = exec antiword -t -i 1 -m UTF-8;mimetype=text/plain;charset=utf-8
# MSWORD: the rcldoc script handles a number of marginal case that raw
# antiword won't:
# - with wvWare: "text stream of this file is too small to handle"
# - with unrtf: rtf files disguising as doc files.
# The default is now again to use rcldoc. Use raw antiword if speed is more
# important for you than catching all data,
application/msword = exec rcldoc
#application/msword = exec antiword -t -i 1 -m UTF-8;mimetype=text/plain;charset=utf-8
# You can also use wvware directly but it's much slower.
# application/msword = exec wvWare --charset=utf-8 --nographics

6
tests/msword/msword.sh
View File

@ -5,8 +5,10 @@ topdir=`dirname $0`/..
initvariables $0
recollq '"Evenements et programme 2006"' 2> $mystderr |
egrep -v '^Recoll query: ' > $mystdout
(
recollq '"Evenements et programme 2006"'
recollq 'pcx11 manuel de programmation iamactuallyanrtf'
) 2> $mystderr | egrep -v '^Recoll query: ' > $mystdout
diff -w ${myname}.txt $mystdout > $mydiffs 2>&1

2
tests/msword/msword.txt
View File

@ -1,3 +1,5 @@
2 results
application/msword [file:///home/dockes/projets/fulltext/testrecoll/msword/programme.doc] [programme.doc] 58880 bytes
application/msword [file:///home/dockes/projets/fulltext/testrecoll/zip/misc.zip] [misc.zip] 168155 bytes
1 results
application/msword [file:///home/dockes/projets/fulltext/testrecoll/msword/IAmActuallyAnRTF.DOC] [DOC PROGRAMMEUR PCX9] 85381 bytes

3
tests/runtests.sh
View File

@ -13,8 +13,9 @@ makeindex() {
echo "Indexing"
recollindex -z
}
if test x$noindex = x ; then
makeindex
fi
# Yes, we could/should use the $toptmp from shared.sh here, but what if
# this is unset ?