sd -- a drop-in replacement for `cd'

# count linefeed separated lines in string `buffer'.
# we append a trailing \n (implying `buffer' is assumed
# to not contain a trailing empty line) via `sdprint'
# since `wc -l' actually counts the number of \n.
# if buffer is empty we return 0
#-------------------------------------------------------
   typeset -i lc=0    #using `-i' suppresses spurious blanks in `wc -l' output.
   [[ -z $1 ]] || lc=$($sdprint "$1" | wc -l)
   $sdprint $lc
}

function sdlogread {
   sdlog=$(<$dirv)
   sdmagicline=${sdlog%%$'\n'*}
   sdlog=$($sdprint "$sdlog" | grep -v "^$sdmagicstring")
   sdlist=$(sdstack)
   sdnew=""
}

function sdlogwrite {
   $sdprint $sdmagicstring-- time stamp: $(date)$'\n'"$sdlog" >| $dirv
   sdlist=$(sdstack)
................................................................................
   lines=$(sdlinecount "$sdlog")

   if (( lines > sdmax )); then
      typeset -i limit

      ((limit = sdmax*9/10))
      ((limit = sdlines > limit ? sdlines:limit))
      sdlog=$($sdprint "$sdlog" | tail -$limit)
      mv -f $dirv ${dirv}.bak
      sdlogwrite
   fi
}

function sdremove {
#----------------------------------------
................................................................................
   typeset dname check
   typeset IFS=$'\n'

   sdl=$sdlines
   sdirs -l $sdmax #look at whole database
   sdlines=$sdl

   dname=$($sdprint "$sdlist" | $awk -F"\t" '$NF ~ /'"$pat"'/ {print $NF}')
   sdirs -l $sdlines  #important to reset before possibly returning...
   [[ -n $dname ]] || { $sdprint no match; return 1; }

   $sdprint "$dname"
   $sdprint -n "remove these entries from the logfile? [y/N] "
   read check
   check=${check:-"n"}
................................................................................
   fi

   # note to self: the '|' quoting '|' is not really required here.
   # its done just to silence `ksh -n'..
   pat="${dname//$IFS/\|}"
   pat="^(${pat//\//\\/})\$"

   sdlog=$($sdprint "$sdlog"|egrep -v "$pat")

   sdlogwrite
}

function sdclean {
#----------------------------------------------
#remove stale entries (and `/') from sdlog/dirv
#----------------------------------------------
   typeset IFS=$'\n'
   typeset -a entries stale
   typeset -i i j


   # note to self: mksh is really bad at `+=' string concatenation
   # when the string becomes large (as is the case here with `entries'
   # and, possibly, `stale'. for that reason we now use indexed
   # arrays for buffering and paste the result together in the end.
   # in ksh/bash it doesn't matter: both approaches are equally
   # fast (and both these shells remain faster than mksh).
   for dname in $sdlog; do
      [[ -d ${dname/#\~/$HOME} ]] && {
         entries[i++]=$dname
      } || stale[j++]=$dname
   done
   [[ -n $stale ]] && {
      stale=$($sdprint -n "${stale[*]}" | sort | uniq )
      $sdprint "$stale"
      $sdprint -n "eliminate these stale directories from the logfile? [y/N] "
      read check
      check=${check:-"n"}
      if [[ "$check" == "y" ]]; then
         sdlog=${entries[*]}
         sdlogwrite
................................................................................
      stacksize=$(sdlinecount "$sdlist")
      mostlen=${#most}
      i=0
      while ((i < indent0+mostlen)); do
         indent+=" "
         ((i++))
      done
      top=$($sdprint "$sdlist" | head -10 | $awk -F'\t' '{print $(NF-1) "\t" $NF}')
      top=${top//$'\n'/$'\n'$indent}
   } || stacksize=0

   vdir=${dirv/#$HOME/\~}


   report="
................................................................................
# given unique entry/directory is used as the resulting weight for
# that entry when sorting. note, that the visit frequencies are used
# as a secondary sort key since in extreme cases (very large `sdpower'
# values) weights can become exactly zero at the bottom of the sorted
# stack and the primary key does not suffice for unambiguous sorting.
# generally this will not happen, though.
#-----------------------------------------------------------------------------
   $sdprint "$sdlog" | tail -$sdlines |\
   $awk '
      BEGIN {
         OFS     = "\t"
         sdlines = '$sdlines'
         sdpower = '$sdpower'
      }
      {
................................................................................
      if [[ $pat == "$sdpattern" ]]; then
         ((sdcount++))
      else
         sdcount=1
      fi
      sdpattern=$pat
      ((sdcount > 1)) && {
         rank=$($sdprint "$sdlist" |\
            $awk -F"\t" "$awkpat"' {
               if (++count == '$sdcount') {
                  print $(NF-1)
                  exit
               }
            }'
         )
         [[ -n $rank ]] && awkpat="NR == $rank" || { sdcount=1; $sdprint starting over ... >&2 ;}
      }
   fi

   matches=$($sdprint "$sdlist" |\
      $awk -F"\t" "$awkpat"' {
         sub(/^~/, "'"$HOME"'", $NF)
         print $NF
      }')


   # select first actually existing directory
   typeset IFS=$'\n'
   for match in $matches; do
      [[ -d $match ]] && {
         dname=$match
         break
................................................................................
               break
            } || {
               [[ -d "$dname" ]] && {
                  # see comment at start of block for explanation of this:
                  command cd "$dname"
                  return $?
               } || {
                  echo "No matching directory"
                  return 1
               }
            }
         } || {
            # we already _know_ that the cd fails _and_ the stack is
            # already maximized. so just let the user see the error
            # message (should be "no such file or directory").
................................................................................
}

function sdchoose { ## dname
   # possibly called by `sdirs' to allow cd to directory selected from
   # list of matches. note to self: sort alphabetically?
   typeset dname="$1"
   typeset -i hits num
   $sdprint "$dname" | $awk -F"\t" '
      BEGIN {
         print "rank\tindex\tname"
      }
      {
         print $(NF-1) "\t" NR "\t" $NF
      }' >&2
   hits=$($sdprint "$dname"|wc -l)
   $sdprint -n "choose index (<CR> = 1; CTRL-C = abort): " >&2
   read -r num
   ((num == 0)) && num=1
   ((num > 0 && num <= hits)) && {
      dname=$($sdprint "$dname" | $awk -F"\t" 'NR == '$num' {sub(/^~/,"'$HOME'",$NF); print $NF}')
      cd "$dname"
      # care is taken that a call to `sdirs' does not send anything to stdout elswhere
      # and here only if we believe that we are operating under tcsh (indicated
      # by `sdistcsh == 1'). this output is then needed by the `sdirs' executable
      # with tcsh.
      ((sdistcsh == 1)) && $sdprint $dname
   }
................................................................................

   if [[ $# == 0 ]]; then
      [[ -n $sdlog ]] || return 1
      dname=$sdlist
   else
      pat="$@"
      pat="${pat//\//\\/}"
      dname=$($sdprint "$sdlist" | $awk -F"\t" '$NF ~ /'"$pat"'/')
      [[ -n $dname ]] || return 1
   fi

   dname=$($sdprint "$dname" | $dump)
   if ((sdselect == 0)); then
      $sdprint $'frec\tfreq\trank\tname'
      $sdprint "$dname"
   else
      sdchoose "$dname"
   fi
}

# count linefeed separated lines in string `buffer'.
# we append a trailing \n (implying `buffer' is assumed
# to not contain a trailing empty line) via `sdprint'
# since `wc -l' actually counts the number of \n.
# if buffer is empty we return 0
#-------------------------------------------------------
   typeset -i lc=0    #using `-i' suppresses spurious blanks in `wc -l' output.
   [[ -z $1 ]] || lc=$(wc -l <<< "$1")
   $sdprint $lc
}

function sdlogread {
   sdlog=$(<$dirv)
   sdmagicline=${sdlog%%$'\n'*}
   sdlog=$(grep -v "^$sdmagicstring" <<< "$sdlog")
   sdlist=$(sdstack)
   sdnew=""
}

function sdlogwrite {
   $sdprint $sdmagicstring-- time stamp: $(date)$'\n'"$sdlog" >| $dirv
   sdlist=$(sdstack)
................................................................................
   lines=$(sdlinecount "$sdlog")

   if (( lines > sdmax )); then
      typeset -i limit

      ((limit = sdmax*9/10))
      ((limit = sdlines > limit ? sdlines:limit))
      sdlog=$(tail -$limit <<< "$sdlog")
      mv -f $dirv ${dirv}.bak
      sdlogwrite
   fi
}

function sdremove {
#----------------------------------------
................................................................................
   typeset dname check
   typeset IFS=$'\n'

   sdl=$sdlines
   sdirs -l $sdmax #look at whole database
   sdlines=$sdl

   dname=$($awk -F"\t" '$NF ~ /'"$pat"'/ {print $NF}' <<< "$sdlist")
   sdirs -l $sdlines  #important to reset before possibly returning...
   [[ -n $dname ]] || { $sdprint no match; return 1; }

   $sdprint "$dname"
   $sdprint -n "remove these entries from the logfile? [y/N] "
   read check
   check=${check:-"n"}
................................................................................
   fi

   # note to self: the '|' quoting '|' is not really required here.
   # its done just to silence `ksh -n'..
   pat="${dname//$IFS/\|}"
   pat="^(${pat//\//\\/})\$"

   sdlog=$(egrep -v "$pat" <<< "$sdlog")

   sdlogwrite
}

function sdclean {
#----------------------------------------------
#remove stale entries (and `/') from sdlog/dirv
#----------------------------------------------
   typeset IFS=$'\n'
   typeset -a entries stale
   typeset -i i j
   typeset dname check

   # note to self: mksh is really bad at `+=' string concatenation
   # when the string becomes large (as is the case here with `entries'
   # and, possibly, `stale'. for that reason we now use indexed
   # arrays for buffering and paste the result together in the end.
   # in ksh/bash it doesn't matter: both approaches are equally
   # fast (and both these shells remain faster than mksh).
   for dname in $sdlog; do
      [[ -d ${dname/#\~/$HOME} ]] && {
         entries[i++]=$dname
      } || stale[j++]=$dname
   done
   [[ -n $stale ]] && {
      stale=$(sort <<< "${stale[*]}" | uniq)
      $sdprint "$stale"
      $sdprint -n "eliminate these stale directories from the logfile? [y/N] "
      read check
      check=${check:-"n"}
      if [[ "$check" == "y" ]]; then
         sdlog=${entries[*]}
         sdlogwrite
................................................................................
      stacksize=$(sdlinecount "$sdlist")
      mostlen=${#most}
      i=0
      while ((i < indent0+mostlen)); do
         indent+=" "
         ((i++))
      done
      top=$($awk -F'\t' 'NR <= 10 {print $(NF-1) "\t" $NF}' <<< "$sdlist" )
      top=${top//$'\n'/$'\n'$indent}
   } || stacksize=0

   vdir=${dirv/#$HOME/\~}


   report="
................................................................................
# given unique entry/directory is used as the resulting weight for
# that entry when sorting. note, that the visit frequencies are used
# as a secondary sort key since in extreme cases (very large `sdpower'
# values) weights can become exactly zero at the bottom of the sorted
# stack and the primary key does not suffice for unambiguous sorting.
# generally this will not happen, though.
#-----------------------------------------------------------------------------
   tail -$sdlines <<< "$sdlog" |\
   $awk '
      BEGIN {
         OFS     = "\t"
         sdlines = '$sdlines'
         sdpower = '$sdpower'
      }
      {
................................................................................
      if [[ $pat == "$sdpattern" ]]; then
         ((sdcount++))
      else
         sdcount=1
      fi
      sdpattern=$pat
      ((sdcount > 1)) && {
         rank=$(
            $awk -F"\t" "$awkpat"' {
               if (++count == '$sdcount') {
                  print $(NF-1)
                  exit
               }
            }' <<< "$sdlist"  
         )
         [[ -n $rank ]] && awkpat="NR == $rank" || { sdcount=1; $sdprint starting over ... >&2 ;}
      }
   fi

   matches=$(
      $awk -F"\t" "$awkpat"' {
         sub(/^~/, "'"$HOME"'", $NF)
         print $NF
      }' <<< "$sdlist"
   )

   # select first actually existing directory
   typeset IFS=$'\n'
   for match in $matches; do
      [[ -d $match ]] && {
         dname=$match
         break
................................................................................
               break
            } || {
               [[ -d "$dname" ]] && {
                  # see comment at start of block for explanation of this:
                  command cd "$dname"
                  return $?
               } || {
                  $sdprint "No matching directory"
                  return 1
               }
            }
         } || {
            # we already _know_ that the cd fails _and_ the stack is
            # already maximized. so just let the user see the error
            # message (should be "no such file or directory").
................................................................................
}

function sdchoose { ## dname
   # possibly called by `sdirs' to allow cd to directory selected from
   # list of matches. note to self: sort alphabetically?
   typeset dname="$1"
   typeset -i hits num
   $awk -F"\t" '
      BEGIN {
         print "rank\tindex\tname"
      }
      {
         print $(NF-1) "\t" NR "\t" $NF
      }' <<< "$dname" >&2
   hits=$(wc -l <<< "$dname")
   $sdprint -n "choose index (<CR> = 1; CTRL-C = abort): " >&2
   read -r num
   ((num == 0)) && num=1
   ((num > 0 && num <= hits)) && {
      dname=$($awk -F"\t" 'NR == '$num' {sub(/^~/,"'$HOME'",$NF); print $NF}' <<< "$dname")
      cd "$dname"
      # care is taken that a call to `sdirs' does not send anything to stdout elswhere
      # and here only if we believe that we are operating under tcsh (indicated
      # by `sdistcsh == 1'). this output is then needed by the `sdirs' executable
      # with tcsh.
      ((sdistcsh == 1)) && $sdprint $dname
   }
................................................................................

   if [[ $# == 0 ]]; then
      [[ -n $sdlog ]] || return 1
      dname=$sdlist
   else
      pat="$@"
      pat="${pat//\//\\/}"
      dname=$($awk -F"\t" '$NF ~ /'"$pat"'/' <<< "$sdlist")
      [[ -n $dname ]] || return 1
   fi

   dname=$($dump <<< "$dname")
   if ((sdselect == 0)); then
      $sdprint $'frec\tfreq\trank\tname'
      $sdprint "$dname"
   else
      sdchoose "$dname"
   fi
}