#! /bin/sh
# wwwoffle-chunks -- chop a wwwoffle index into smaller chunks that stick around
# Copyright       : http://www.fsf.org/copyleft/gpl.html
# Author          : Dan Jacobson -- http://jidanni.org/comp/wwwoffle/
# Created On      : Aug 2001
# Last Modified On: Mon Jan 17 09:49:38 2011
# Update Count    : 481

#Try wwwoffle-chunks --help
this_prog=`basename $0`
test $# -gt 1 -o "x$1" = x--help && exec cat <<\EOF
Concept: There's no way we are going to obediently read all the stuff
in the "lasttime" index before the "thirst for info" drives us back
online.  Prevtime indexes are a help, but better yet, each time when
we go offline, let's chop up lasttime into oh, screen sized separate
parts that stick around until you are finished with them."

You end up with (default=)32 entry chunks of "lasttime" that pile up
until you are confident you have finished savoring each entry, at
which time you leave your browser to do "wwwoffle-chunks -1" which
zaps the last chunk you read, and (if $BROWSER is set) hopefully then
sends you back to read some other chunk.

Usage:
You are a wwwoffle user who is now offline.
$ mv wwwoffle-chunks ~/bin/ #put this program in your bin dir. chmod +x ...
$ wwwoffle-chunks #run this program
$ $BROWSER ~/wwwoffle-chunks/index.html #use your favorite browser
When you are finished with (all the links in) a chunk, you will probably want
to delete it:
$ wwwoffle-chunks -u
#interactively delete chunks in order of last access (ls -tu)

[Feb. 2010: Due to mount(8) nostrictatime being so common these days,
all options that relate to ls -tu have been disabled, even though
still blabbed about here.]

(^C to quit), or
$ wwwoffle-chunks -t
#interactively delete chunks in order of content time (ls -t)
or
$ wwwoffle-chunks -1 #delete last chunk accessed, no questions asked.
Both of which call wwwoffle-chunks -i to remake the index automatically.
Note, we said last chunks accessed, not last chunks created.
wwwoffle-chunks -1 would be nice to be able to do from inside the browser,
but we don't know how.  Wait: one can copy "wwwoffle-chunks -1
" (newline included) once with the mouse and then each time: select the shell
window, paste, and return back to the browser, all with the mouse!
However, be careful that after a while you don't accidentally cut and send
other text to the shell!

Then you will probably want to automate this after every call.
The script I hang up with is:
poff -a;(su -c 'm=~/bin/wwwoffle-chunks;$m&&$m -i' jidanni)&

Also make a bookmark in your browser to ~/wwwoffle-chunks/index.html
-c gives date comparison of chunks vs. last/prevtime indexes
--help prints this help.
Do wade through the source code to see any further customizations possible.
E.g. to cut the prevtime3 into chunks 44 lines long:
$ size=44 wwwoffle-chunks prevtime3
or:
$ export WWWOFFLE_CHUNKS='size=44' #best done in .bash_profile
$ wwwoffle-chunks prevtime3

-e edits the last accessed chunk.
Additional option flags can be found in by digging in the source.
EOF

export LC_ALL=C LANG=C
DIR=${HOME?}/wwwoffle-chunks
: ${wwwoffle_cache=/var/spool/wwwoffle} #if not otherwise set by the user
case "$*" in
-*[1elpsux]*)
	echo 1>&2 "Due to probable mount(8) nostrictatime, this option is often unsafe."
	exit 88;;
-e) #edit last accessed chunk
	cd $DIR && ${EDITOR?} `ls -tu|sed '/^index\.html$/d;q'`; exit;;
-c) #compare dates of chunks and last/prevtime indexes
     ls -dtog $DIR/* $wwwoffle_cache/*time*/.timestamp|
     perl -anlwe 'print "@F[3..6]"'; exit;;
-r)echo $this_prog: try -u nowadays 1>&2; exit 56;;
-p) #secret plucker mess option
    set -eu; cd $DIR; O=-M\ 2 pluckme `ls -tu|sed '/index\.html/d;q'`; exit;;
-x) #secret exporter option, uses my wwwoffle-cpio, includes the chunk index
    set -eu; cd $DIR; set -- `ls -tu|sed '/index\.html/d;q'`
    o=/tmp/$(basename $1 .html).cpio
    echo $PWD/$1|cpio -ovF $o
    lynx -dump -listonly $1|perl -nwe 's/^\s+\d+\.\s//&&print'|
    wwwoffle-cpio -ovAF $o
    bzip2 $o; ls -og $o*; exit;;
-s) #secret spider  option, but doesn't touch the U* files!
    set -eu; cd $DIR
    wget --spider --force-html -i `ls -tu|sed '/index\.html/d;q'`; exit;;
-l) #secret link report option
    set -eu; cd $DIR
    lynx -listonly -dump `ls -tu|sed '/index\.html/d;q'`|
    perl -nwe 's/^\s+\d+\.\s+//&&print'; exit;;
-[tu1]) #simple chunk removal
    #no simple way to put a button at the bottom of a chunk
    # to remove it and show another chunk, so:
    set -e
    cd $DIR
    if test $1 = -1
	then candidate=`ls -tu|sed '/index\.html/d;q'` atime='3 minutes ago'
	test $candidate || { echo 2>&1 $this_prog: nothing to remove; exit 87;}
	if test `stat -c %X $candidate` -lt `date +%s -d "$atime"`
	    then echo Sir, $candidate was last accessed more than $atime.
	    echo Therefore, I assume you made a mistake.; exit 44; fi
        mv -v $candidate /tmp
	cd `pwd` && sleep 1 && $this_prog -i
	if test $BROWSER && pidof $BROWSER > /dev/null
	    then $BROWSER index.html
	fi
	exit
    fi
    trap "echo && echo ... && cd `pwd` && sleep 1 && $this_prog -i" 0
    echo Times read of first 10:
    ls $1tog|head
    find `ls $1tu` -maxdepth 0 -type f ! -name index.html -ok rm -f {} \;
    false
esac
doctype='<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
"http://www.w3.org/TR/html4/strict.dtd">'
charset='<META http-equiv="Content-Type" content="text/html; charset=utf-8">'
case "$*" in -i) #refresh index
  set -e
  cd $DIR
  i=index.html
 if test -f $i -a ! $i -ot .
  then
      echo index need not be remade
  else
      echo making index
      cat<<EOF > $i
$doctype
<HEAD>
$charset
<TITLE>WCI-WWWOFFLE Chunks Index</TITLE>
</HEAD>
<BODY><PRE>
WWWOFFLE Chunks Index
EOF
#not all ls' have ls -l --time-style=iso, nor all find have -printf, so:
#        ls -lr $DIR|awk '$NF=="index.html"{next}
#        NF>7{if($6$7$8!=n){printf"--"}else{printf"  "};n=$6$7$8;
#        printf "%s %2s %s <A href=\"%s\">%s</A>\n", $6,$7,$8,$9,$9}' >> $i

#4/2004 above becomes unordered if we ever edit a chunk, so
      ls -r|perl -wne 'use POSIX qw(strftime);
        next if /^index\.html$/; chomp; $f=$_;
        s/(^[[:xdigit:]]{8})-.*/strftime "%a %b %e %H:%M %Y ",
          localtime(hex $1)/e && print;
        print "<A href=\"$f\">$f<\/A>\n";'>> $i
      echo '</PRE></BODY></HTML>' >> $i
  fi
  exit
esac
test -d $wwwoffle_cache||{
    echo $this_prog: $wwwoffle_cache: invalid wwwoffle_cache directory 1>&2
    exit 24;} #actually we don't really need it below, but it's nice
index=${1-lasttime}
test -d $DIR||mkdir $DIR
cd $DIR||exit 22
set -e #over paranoid
indexdir=$wwwoffle_cache/$index
ts=$indexdir/.timestamp
if test -r $ts
then
    set -- $(stat -c '%Y %y' $ts)
    datestr=$(printf %x $1)
    #trying hard to order the directory no matter prevtime, etc.
    #for lynx, netscape, now ls, ls -t will give same order... that of $ts date
    oIFS=$IFS; IFS=-.$oIFS; set -- $@; IFS=$oIFS
    nice_date=$4.$3.$2\ $5 #want guts first for skinny icons
else #too bad
    nice_date="[`date` saved]"
    datestr=$(printf %x $(date +%s))
fi
prefix=$datestr-$index- #don't mention $index else confusing...
eval $WWWOFFLE_CHUNKS
: ${size:=32}
#now back to wwwoffle-ls, as wwwoffle -o won't print user:passwd@site 2003.11
#but then Chinese just becomes %..%..
: ${sort:=alpha}; wwwoffle -o http://localhost:8080/index/$index?sort=$sort|
sed -n 's/^<li>&nbsp;//p'|
split -$size - $prefix
set -- $prefix??
if test -f $1
    then tmp=($prefix??)
    else echo $this_prog: Seems there were no files fetched for $index! 1>&2; exit 34
fi
shortrefs=(${tmp[@]##*-}) #??: better than *
for i in ${tmp[*]}; do mv -vf $i $i.html; done #-f: handy when testing etc.
refs=($prefix??.html)

i=0; while test ${refs[$i]}
do
	links[$i]="<a href=\"${refs[$i]}\">${shortrefs[$i]}</a>"
	: $((i++))
done

echo making ${refs[*]}|xargs -n 1 #neater
m=0; for i in ${refs[*]}
do
	: $((m++))
	oldlink=${links[$m-1]}; links[$m-1]=[${shortrefs[$m-1]}]" "
	test "${refs[$m]}"||sorry="(current=last page)"
	#well, maybe they read from last to first...
	frag=${i##*-}; frag=${frag%.*}
	ed -s $i <<!
1i
$doctype
<HEAD>
$charset
<TITLE>$nice_date $frag WWWOFFLE chunk</TITLE>
</HEAD><BODY>
<pre>WWWOFFLE chunk $i $nice_date${sort:+ sort=$sort}
.
\$a
 -END chunk $i $nice_date${sort:+ sort=$sort}-
</pre></BODY></HTML>
<!-- Local Variables: -->
<!-- truncate-lines: 1 -->
<!-- make-backup-files: nil -->
<!-- End: -->
.
w
q
!
#<p>[
#<a href="index.html">(up)</a> ${links[@]}
#${refs[$m]+ . . . . . . Next:} <a href="${refs[$m]}">${shortrefs[$m]}</a>$sorry
#]</p>
links[$m-1]=$oldlink
#I wanted to use align=center, but I also want the strict DTD...

#with a mouse it feels better when "Next" is near the middle of the
#page.  With lynx, it being at the beginning of line is better, or hit ^E

#Dan Mercer, comp.unix.shell: You can also use:  : "${a[@]}"; echo $_
#I should add the standard footer WWWOFFLE links, but wait, I'm trying
#to correct a size & slowness problem, so i should make this as clean
#as possible.

#because WWWOFFLE 2.6a at least produces illegal html with raw & in URLs
#tidy -q $i | tr -d \\240 > /tmp/WWW$$; mv /tmp/WWW$$ $i
#also tidy causes \240 where nbsp was, this eats the next char, at
#least with lynx...

test -r $ts && touch -r $ts $i
done
#OK, now perhaps browse the index
#and after a while clean up the old *-chunk-* files by hand
#wish I could put a button at the bottom of each chunk to remove
#that chunk with just a click...

#I have put
#m=wwwoffle-chunks; $m && $m -i
#in a script that is run when my modem is hung up.
:||<<\EOF #more usage stuff, parked here
After going offline, run this program to break a too-big WWWOFFLE index
into chunks to save wear and tear and waiting each time one hits the
BACK key in their favorite browser, or if they don't read their Last
Time indexes before they drop off the back of the [1] [2] [3] ... list.
Then one points their browser to $DIR/...  optional argument is what
index to get, default is "$defaultindex".  One needn't be root to run
this program.
#Thanks to Felix Karpfen, only know other user, for tips.
EOF