use publ_id
diff --git a/klk2eureco.sh b/klk2eureco.sh
index a01e077..7fe0386 100755
--- a/klk2eureco.sh
+++ b/klk2eureco.sh
@@ -13,8 +13,8 @@
TEI="$CORPUS/TEI" # output zipped tei files of the selection
#TEI="$TEST/TEI" # output zipped tei files of the selection
-#TEIFORMAT="tei";
-TEIFORMAT="i5";
+TEIFORMAT="tei";
+#TEIFORMAT="i5";
MASK="";
#MASK="-m";
@@ -41,48 +41,57 @@
# Generate corpus files by year and source in vrt
-# for VRT in $DAT/*.vrt # loop to be applied to yearly files
-for VRT in $DAT/klk_fi_v2_201[0-3].vrt
-# for VRT in "$DAT/klk_fi_v2_2015.vrt"
-# for VRT in "$DAT/klk_fi_v2_2019.vrt" "$DAT/klk_fi_v2_2020.vrt" "$DAT/klk_fi_v2_2021.vrt"
+#for VRT in $DAT/*.vrt # loop to be applied to yearly files
+for VRT in "$DAT/klk_fi_v2_2021.vrt"
do # (
BASENAME=`basename $VRT .vrt`
YEAR=`echo $BASENAME | gawk 'BEGIN {FS="_"} {print $4}'`
- echo "Generating proper XML files from $VRT in $XML/$YEAR by source..."
- gawk -v OUTDIR="$XML" -v YEAR="$YEAR" -f vrt2xml.awk $VRT
+ ##TMP echo "Generating proper XML files from $VRT in $XML/$YEAR by source..."
+ ##TMP gawk -v OUTDIR="$XML" -v YEAR="$YEAR" -f vrt2xml.awk $VRT
+
+
#--------------------------------------------------------
# Checking Wellformedness of the XML and generating TEI
#--------------------------------------------------------
- for s in $XML/*/
- # for s in $XML/Suomen_Kuvalehti/
- do ( # threading
+ ##TMP for s in $XML/*/
+ for s in $XML/Suomen_Kuvalehti/
+ do # ( # threading
SOURCE=`basename $s`
x=$s/$SOURCE$YEAR.xml
- # Some checks of the XML beforehand:
+ #---------------------------------------------------------------
+ # Some existence checks and removal of empty files beforehand:
+ #---------------------------------------------------------------
+
+ if [[ ! -f $x ]] ; then
+ echo "Warning: klk2eureco.sh: File $x does not exist; skipping";
+ break;
+ fi
+
+ if [[ -z `grep -l "<text " $x` ]] ; then
+ #echo "Warning: klk2eureco.sh: File $x does not contain any <text>s, skipping"
+ echo "Warning: klk2eureco.sh: File $x does not contain any <text>s,removing"
+ rm -fv $x;
+ break;
+ fi
- if [[ ! -f $x ]] ; then
- echo "Warning: klk2eureco.sh: File $x does not exist; skipping"
- break;
- fi
-
- if [[ -z `grep "<text " $x` ]] ; then
- echo "Warning: klk2eureco.sh: File $x does not contain any <text>s ; skipping"
- break;
- fi
-
+ #---------------------------------
+ # checking well-formedness of $x
+ #---------------------------------
+
echo " checking wellformedness of $x"
- xmllint --stream --noout $x
+ xmllint --stream --noout $x
if [[ $R != "" ]] ; then
echo "Error: xmllint error with error return code $R" >&2;
break;
fi
+
#--------------------
# calling vrt2tei.pl
#--------------------
@@ -98,6 +107,12 @@
# xmllint- format geht out of memory und --stream machen funktioniert nicht
ls -l $t
+ # if $t is empty, remove it:
+ if [ ! -s $i ]; then
+ rm -f $i
+ fi
+
+ # validating $t:
if [ "$TEIFORMAT" == "i5" ] ; then
echo " validating $t..."
xmllint --stream --noout --valid $t
@@ -106,13 +121,13 @@
xmllint --stream --noout --dtdvalid $TEIDTD $t # scheint so zu funktioneren - nicht kombinieren mit --format!
fi
- # echo " zipping $x..."
- # gzip -f $x
-
- # echo " zipping $t..."
- # gzip -f $t
-
- echo ) &
+###TMP # echo " zipping $x..."
+###TMP # gzip -f $x
+###TMP
+###TMP # echo " zipping $t..."
+###TMP # gzip -f $t
+
+ echo # ) &
done
wait # ) & # wait seems to have to be there for it to terminate