text outputs
diff --git a/klk2eureco.sh b/klk2eureco.sh
index 47a67b6..8b0ddbc 100755
--- a/klk2eureco.sh
+++ b/klk2eureco.sh
@@ -1,15 +1,18 @@
#/bin/bash
EURECO="/scratch/project_2010889/eureco"
+TEST="/scratch/project_2010889/TEST"
CORPUS="$EURECO/klk-fi-v2-vrt"
TEIDTD="$EURECO/tei/tei_all.dtd"
DAT="$CORPUS/data" # input vrt files of klk-fi by year
-XML="$CORPUS/XML" # output zipped vrtxml files of selection of papers
-TEI="$CORPUS/TEI" # output zipped tei files of the selection
+XML="$CORPUS/XML" # output zipped vrtxml files of selection of newspapers
+#XML="$TEST/XML" # output zipped vrtxml files of selection of newspapers
+#TEI="$CORPUS/TEI" # output zipped tei files of the selection
+TEI="$TEST/TEI" # output zipped tei files of the selection
-
+mkdir -p $TEST
mkdir -p $XML
mkdir -p $TEI
@@ -36,13 +39,14 @@
BASENAME=`basename $VRT .vrt`
YY=`echo $BASENAME | gawk 'BEGIN {FS="_"} {print $4}'`
- echo "Generating proper XML files from $VRT in $XML/$YY by source..."
+ #echo "Generating proper XML files from $VRT in $XML/$YY by source..."
#gawk -v OUTDIR="$XML" -v YEAR="$YY" -f vrt2xml.awk $VRT # will generate corpus files for different sources in the YEAR dir
# Checking Wellformedness of the XML and generating TEI
# for s in $XML/*/
- # for s in $XML/Suomen_Kuvalehti/
- for s in $XML/Aamulehti/ $XML/Etela-Suomen_Sanomat/ $XML/Hameen_Sanomat/ $XML/Helsingin_Sanomat/ $XML/Ilkka-Pohjalainen/ $XML/Ilta-Sanomat/ $XML/Kaleva/ $XML/Keskipohjanmaa/ $XML/Satakunnan_Kansa/ $XML/Savon_Sanomat/ $XML/Turun_Sanomat/
+ for s in $XML/Suomen_Kuvalehti/
+ # for s in $XML/Helsingin_Sanomat/
+ # for s in $XML/Aamulehti/ $XML/Etela-Suomen_Sanomat/ $XML/Hameen_Sanomat/ $XML/Helsingin_Sanomat/ $XML/Ilkka-Pohjalainen/ $XML/Ilta-Sanomat/ $XML/Kaleva/ $XML/Keskipohjanmaa/ $XML/Satakunnan_Kansa/ $XML/Savon_Sanomat/ $XML/Turun_Sanomat/
do # ( # threading
SOURCE=`basename $s`
x=$s/$SOURCE$YY.xml
@@ -59,12 +63,10 @@
t="$TEI/$SOURCE/$BASENAME.tei.xml"
t0="$TEI/$SOURCE/$BASENAME.tei.0.xml"
- echo " generating $t0 using vrt2tei.pl"
- ./vrt2tei.pl $x > $t0
+ echo " generating $t using vrt2tei.pl, and prettifying..."
+ ./vrt2tei.pl $x | xml_pp > $t # xml_pp works but takes ages
+ # xmllint --format $t0 > $t # geht out of memory und --stream machen funktioniert nicht
- echo " prettifying..."
- #xmllint --format $t0 > $t # geht out of memory und --stream machen funktioniert nicht
- xml_pp $t0 > $t # works but takes ages
ls -l $t
echo " validating..."