viewvc.cgi/safari/get_book.sh

#!/bin/sh

# proxy settings (same as in firebird)
fping -q proxy && export http_proxy=http://proxy:8080 ||
fping -q proxy.lan && export http_proxy=http://proxy.lan:8080
# user agent (same as in firebird)
ua="Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.6) Gecko/20040506 Firefox/0.8"
cookie_file="cookie.txt"

# wait between pages
export wait=120

if [ ! -z "$http_proxy" ] ; then
        echo "Using proxy $http_proxy"
fi

if [ -z "$1" ] ; then
        echo "Usage: $0 ISBN"
        exit 1
fi

if [ -e orig ] ; then
        echo "orig directory found. Resume download? [Y/n]"
        read ans
        if [ "$ans" = "n" ] ; then
                exit 1;
        fi
        mv orig/* .
        rm -Rf orig

        grep -l 'promo.asp' * | xargs -i rm {}
        grep -l 'This is only a preview of the full book' * | xargs -i rm {}
fi

isbn=`echo $1 | sed 's/-//g' | tr '[a-z]' '[A-Z]'`

function mirror() {

        url="$1"

        file=`echo $url | sed -e s,http://[^?]*?,index.html?, -e s,#.*$,, -e s,/,%2F,g`
        if [ -e "$file" ] ; then
#               echo "skip $url"
                echo -n "."
                return
        fi

        cookie=""
        if echo $url | grep '/index' >/dev/null ; then
                echo -n "no login (index) "
        elif echo $url | grep 'mode=toc' >/dev/null ; then
                echo -n "no login (toc) "
        else
                if [ ! -e $cookie_file ] ; then
                        echo "cookies file $cookie_file doesn't exits! Please create it."
                        echo "It should be in format:"
                        echo "Cookie: Site=UICode=&Portal=oreilly&GUID=..."
                        exit 1
                fi
                read cookie < $cookie_file
                if [ -z "$cookie" ] ; then
                        echo "Empty cookie file $cookie_file !"
                        exit 1
                fi
        fi

        if [ -z "$cookie" ] ; then
                echo "$url [no cookie]"
        else
                echo "$url [with cookie]"
        fi

        wget -q -p -nH -nc -k -t 1 -U "$ua" --cookies=off --header="$cookie" $url
        perl -e '$t=rand($ENV{wait} || 120);print "sleep for $t sec.\n"; sleep($t);'

}

function geturl() {
        hindent -s $1 | grep -i href | grep mode=[st][eo]c | \
        sed -e 's/^.*<a.*href="//i' \
                -e 's/".*//' -e 's/amp;//g' \
                -e 's,^[^\?]*\?,http://safari.oreilly.com/,' \
                -e 's/#.*$//' \
                -e 's/\&srchText=//' \
                -e 's/open=false/open=true/' | \
                grep '&s=1&b=1&f=1&t=1&c=1&u=1&r=&o=1' | \
                grep -i "xmlid=[0-9A-Za-z\-][0-9A-Za-z\-]*" | \
                grep -vi "xmlid=[0-9A-Za-z\-]*/[0-9][0-9][0-9][0-9][0-9][0-9][0-9]" | \
                sort -u >> in
}

function uniqurl() {
        mv in in.tmp
        grep -v 'view=[A-Z]' in.tmp | sort -u > in
        grep 'view=[A-Z].*/index' in.tmp | sort -u >> in
        links=`wc -l in | cut -d" " -f1`
        echo "found $links unique links"
}

function mirror_in() {
        cat in | while read url ; do
                mirror "$url"
                #sleep $wait

                if grep 'promo.asp' `ls -t index.html* | head -3` >/dev/null ; then
                        echo "WARNING: safari seems to logout you as user. Aborting."
                        exit 1
                fi

                if grep -i '>Account locked<' `ls -t index.html* | head -3` >/dev/null ; then
                        echo "WARNING: your safari account is locked. Aborting."
                        exit 1
                fi

                if grep -i 'session disabled' `ls -t index.html* | head -3` >/dev/null ; then
                        echo "WARNING: your safari session is disabled. Aborting."
                        exit 1
                fi

                if grep -i 'This is only a preview of the full book' `ls -t index.html* | head -3` >/dev/null ; then
                        echo "WARNING: you didn't add this book to your bookshelf!"
                        exit 1
                fi
        done
        echo
}

echo -n > in
mirror "http://safari.oreilly.com/$isbn"
echo

file=`ls *index.html?XmlId=*`
isbn=`echo $file | cut -d= -f2`
echo "extract URLs from first page $file... [$isbn]"
geturl $file $isbn
uniqurl

mirror_in

echo -n "extracting URLs [1]"
ls index.html* | while read file ; do
        echo -n "."
        geturl $file $isbn
done
echo

uniqurl

mirror_in

echo -n > in
echo -n "extracting URLs [2]"
ls index.html* | while read file ; do
        echo -n "."
        geturl $file $isbn
done

uniqurl

mirror_in

# convert links in html
bn=`basename $0`
dir=`echo $0 | sed "s/$bn$//"`
ls index.html* | xargs -i $dir/filter.pl {}
mkdir orig
mv index.html* orig/

1	dpavlin	1.1	#!/bin/sh
2
3	dpavlin	1.5	# proxy settings (same as in firebird)
4	dpavlin	1.11	fping -q proxy && export http_proxy=http://proxy:8080 \|\|
5			fping -q proxy.lan && export http_proxy=http://proxy.lan:8080
6	dpavlin	1.5	# user agent (same as in firebird)
7	dpavlin	1.9	ua="Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.6) Gecko/20040506 Firefox/0.8"
8			cookie_file="cookie.txt"
9	dpavlin	1.5
10			# wait between pages
11			export wait=120
12
13	dpavlin	1.11	if [ ! -z "$http_proxy" ] ; then
14			echo "Using proxy $http_proxy"
15			fi
16	dpavlin	1.1
17	dpavlin	1.2	if [ -z "$1" ] ; then
18			echo "Usage: $0 ISBN"
19			exit 1
20			fi
21
22	dpavlin	1.5	if [ -e orig ] ; then
23			echo "orig directory found. Resume download? [Y/n]"
24			read ans
25			if [ "$ans" = "n" ] ; then
26			exit 1;
27			fi
28			mv orig/* .
29			rm -Rf orig
30
31	dpavlin	1.7	grep -l 'promo.asp' * \| xargs -i rm {}
32			grep -l 'This is only a preview of the full book' * \| xargs -i rm {}
33	dpavlin	1.5	fi
34
35	dpavlin	1.11	isbn=`echo $1 \| sed 's/-//g' \| tr '[a-z]' '[A-Z]'`
36	dpavlin	1.1
37			function mirror() {
38	dpavlin	1.5
39			url="$1"
40
41			file=`echo $url \| sed -e s,http://[^?]?,index.html?, -e s,#.$,, -e s,/,%2F,g`
42			if [ -e "$file" ] ; then
43			# echo "skip $url"
44	dpavlin	1.8	echo -n "."
45	dpavlin	1.5	return
46			fi
47
48	dpavlin	1.11	cookie=""
49	dpavlin	1.5	if echo $url \| grep '/index' >/dev/null ; then
50			echo -n "no login (index) "
51			elif echo $url \| grep 'mode=toc' >/dev/null ; then
52			echo -n "no login (toc) "
53			else
54	dpavlin	1.9	if [ ! -e $cookie_file ] ; then
55			echo "cookies file $cookie_file doesn't exits! Please create it."
56			echo "It should be in format:"
57			echo "Cookie: Site=UICode=&Portal=oreilly&GUID=..."
58			exit 1
59			fi
60			read cookie < $cookie_file
61			if [ -z "$cookie" ] ; then
62			echo "Empty cookie file $cookie_file !"
63			exit 1
64			fi
65	dpavlin	1.5	fi
66
67	dpavlin	1.9	if [ -z "$cookie" ] ; then
68			echo "$url [no cookie]"
69			else
70			echo "$url [with cookie]"
71			fi
72	dpavlin	1.1
73	dpavlin	1.9	wget -q -p -nH -nc -k -t 1 -U "$ua" --cookies=off --header="$cookie" $url
74	dpavlin	1.11	perl -e '$t=rand($ENV{wait} \|\| 120);print "sleep for $t sec.\n"; sleep($t);'
75	dpavlin	1.5
76	dpavlin	1.1	}
77
78			function geturl() {
79	dpavlin	1.2	hindent -s $1 \| grep -i href \| grep mode=[st][eo]c \| \
80	dpavlin	1.1	sed -e 's/^.<a.href="//i' \
81			-e 's/".*//' -e 's/amp;//g' \
82			-e 's,^[^\?]*\?,http://safari.oreilly.com/,' \
83	dpavlin	1.8	-e 's/#.*$//' \
84	dpavlin	1.2	-e 's/\&srchText=//' \
85			-e 's/open=false/open=true/' \| \
86	dpavlin	1.1	grep '&s=1&b=1&f=1&t=1&c=1&u=1&r=&o=1' \| \
87	dpavlin	1.11	grep -i "xmlid=[0-9A-Za-z\-][0-9A-Za-z\-]*" \| \
88			grep -vi "xmlid=[0-9A-Za-z\-]*/[0-9][0-9][0-9][0-9][0-9][0-9][0-9]" \| \
89	dpavlin	1.1	sort -u >> in
90			}
91
92	dpavlin	1.2	function uniqurl() {
93			mv in in.tmp
94			grep -v 'view=[A-Z]' in.tmp \| sort -u > in
95			grep 'view=[A-Z].*/index' in.tmp \| sort -u >> in
96	dpavlin	1.8	links=`wc -l in \| cut -d" " -f1`
97			echo "found $links unique links"
98	dpavlin	1.2	}
99
100	dpavlin	1.4	function mirror_in() {
101			cat in \| while read url ; do
102			mirror "$url"
103	dpavlin	1.5	#sleep $wait
104	dpavlin	1.4
105			if grep 'promo.asp' `ls -t index.html* \| head -3` >/dev/null ; then
106	dpavlin	1.5	echo "WARNING: safari seems to logout you as user. Aborting."
107			exit 1
108			fi
109
110			if grep -i '>Account locked<' `ls -t index.html* \| head -3` >/dev/null ; then
111			echo "WARNING: your safari account is locked. Aborting."
112			exit 1
113			fi
114
115			if grep -i 'session disabled' `ls -t index.html* \| head -3` >/dev/null ; then
116			echo "WARNING: your safari session is disabled. Aborting."
117	dpavlin	1.4	exit 1
118			fi
119	dpavlin	1.7
120			if grep -i 'This is only a preview of the full book' `ls -t index.html* \| head -3` >/dev/null ; then
121			echo "WARNING: you didn't add this book to your bookshelf!"
122			exit 1
123			fi
124	dpavlin	1.4	done
125	dpavlin	1.8	echo
126	dpavlin	1.3	}
127
128	dpavlin	1.4	echo -n > in
129	dpavlin	1.11	mirror "http://safari.oreilly.com/$isbn"
130	dpavlin	1.8	echo
131	dpavlin	1.1
132	dpavlin	1.11	file=`ls index.html?XmlId=`
133			isbn=`echo $file \| cut -d= -f2`
134			echo "extract URLs from first page $file... [$isbn]"
135			geturl $file $isbn
136	dpavlin	1.2	uniqurl
137	dpavlin	1.1
138	dpavlin	1.4	mirror_in
139	dpavlin	1.1
140			echo -n "extracting URLs [1]"
141			ls index.html* \| while read file ; do
142			echo -n "."
143			geturl $file $isbn
144			done
145			echo
146
147	dpavlin	1.2	uniqurl
148	dpavlin	1.1
149	dpavlin	1.4	mirror_in
150	dpavlin	1.1
151	dpavlin	1.4	echo -n > in
152	dpavlin	1.1	echo -n "extracting URLs [2]"
153			ls index.html* \| while read file ; do
154			echo -n "."
155			geturl $file $isbn
156			done
157
158	dpavlin	1.2	uniqurl
159
160	dpavlin	1.4	mirror_in
161	dpavlin	1.2
162			# convert links in html
163			bn=`basename $0`
164			dir=`echo $0 \| sed "s/$bn$//"`
165			ls index.html* \| xargs -i $dir/filter.pl {}
166			mkdir orig
167			mv index.html* orig/
168	dpavlin	1.1