view update.sh @ 75:3d9e42477367

More improvements in the parsing and XML output.
author Matti Hamalainen <ccr@tnsp.org>
date Wed, 24 Oct 2012 07:33:28 +0300
parents 2cfb0a7eac9b
children d1b65d9903ab
line wrap: on
line source

#!/bin/sh
URLSUFFIX=".htm"

function parse()
{
# Create cache directories, if they do not exist
OLDCACHEDIR="${CACHEDIR}old/"
if test ! -d "${OLDCACHEDIR}"; then
	mkdir -p "${OLDCACHEDIR}"
fi

if test ! -d "${CACHEDIR}"; then
	mkdir -p "${CACHEDIR}"
fi

URLPREFIX="$1"
CLASSFILE="$2"
LISTFILE="$2.tmp"
PATPREFIX="$3"

if wget -q -O "$LISTFILE" "$URLPREFIX"; then
	perl -ne "if (/<a href=\"${PATPREFIX}([A-Z]{3}\d\S+)${URLSUFFIX}\">/) { print \"\$1\n\"; }" < "$LISTFILE" > "$CLASSFILE"
	echo -n "* Fetched classfile $CLASSFILE: "
	cat "$CLASSFILE" | wc -l
fi

if test -e "$CLASSFILE"; then
cat "$CLASSFILE" | while read i; do
	parse=no
	INFILE="${CACHEDIR}${i}.html"
	ONFILE="${OLDCACHEDIR}${i}.html"
	DATAFILE="${i}.data"
	wget -q -O "${INFILE}.new" "${URLPREFIX}${PATPREFIX}${i}${URLSUFFIX}"

	if test -e "${INFILE}.new"; then
		# New data fetched, does old file exist?
		if test -e "$INFILE"; then
			# Yes, do a diff
			if ! diff -u "$INFILE" "$INFILE.new" > "$INFILE.diff"; then
				# There were differences, do a parse
				parse=yes
				mv "$INFILE" "$ONFILE" && \
				mv "$INFILE.new" "$INFILE"
			else
				# No changes, apparently .. remove the new one
				rm -f "$INFILE.new" "$INFILE.diff"
			fi
		else
			# No old file, parse new data
			mv "$INFILE.new" "$INFILE"
			parse=yes
		fi
	fi
	# No new file fetched, does datafile exist?
	if test ! -e "${CACHEDIR}${DATAFILE}"; then
		# No, try to parse it if old file input exists
		parse=yes
	fi

	# Parsing of old data requested?
	if test "x$parse" = "xyes" -a -e "$INFILE"; then
		if test -e "${CACHEDIR}${DATAFILE}"; then
			mv "${CACHEDIR}${DATAFILE}" "${OLDCACHEDIR}${DATAFILE}"
		fi
		echo "Parsing $i"
		perl parsedata.pl -php "$INFILE" -o "${CACHEDIR}${DATAFILE}"
		perl parsedata.pl -xml "$INFILE" -o "${CACHEDIR}${i}.xml"
	fi
done
fi
}

CACHEDIR="cache/"
parse "http://www.oamk.fi/tyojarjestykset/otek/luokat/" "classes.txt" "OR_"
#parse "http://www.oamk.fi/~heikkim/riihi2/Oppilaat/" "classes.txt" "Ryh._"

CACHEDIR="cache-next/"
parse "http://www.oamk.fi/~heikkim/riihi2/Oppilaat/" "classes_next.txt" "Ryh._"