translit.sh

   1 #!/bin/sh
   2
   3 # assumes utf8 locale..
   4 # remove nonascii from the output of pdftotext -layout standard.pdf
   5
   6 sed '
   7 s/\f/(newpage)/g
   8 # utf8 fixes
   9 s/ﬁ/fi/g
  10 s/ﬂ/fl/g
  11 s/§/!S/g
  12 s/©/(C)/g
  13 s/—/--/g
  14 s/−/-/g
  15 s/∗/*/g
  16 s/ˆ/^/g
  17 s/〈/</g
  18 s/〉/>/g
  19 s/⎡/[^/g
  20 s/⎤/^]/g
  21 s/⎣/[_/g
  22 s/⎦/_]/g
  23 s/⎢/[ /g
  24 s/⎥/ ]/g
  25 s/⎧/{/g
  26 s/⎨/{/g
  27 s/⎩/{/g
  28 s/±/(+-)/g
  29 s/≤/<=/g
  30 s/≥/>=/g
  31 s/≠/!=/g
  32 s/Σ/(Sum)/g
  33 s/√/sqrt:/g
  34 s/π/pi/g
  35 s/∞/(inf)/g
  36 s/ƒ/fl./g
  37 s/∫/(integral)/g
  38 s/Γ/(Gamma)/g
  39 s/×/x/g
  40 s/•/o/g
  41 s/⎯/-/g
  42 s/↑/(uparrow)/g
  43 s/↓/(downarrow)/g
  44 s/↔/(<->)/g
  45 s/→/(->)/g
  46 s/‘/'\''/g
  47 s/’/'\''/g
  48 # pdftotext layout fixes
  49 s/_ _/__/g
  50 s/\([0-9]\). \([0-9]\)/\1.\2/g
  51 ' | LC_ALL=C tr -c '\n-~' '?' | awk '
  52 BEGIN {
  53         getline
  54         last=$0
  55         side=0
  56 }
  57 /^\(newpage\)/ {
  58         n=split(last,a)
  59         if(side)
  60                 p=a[1]
  61         else
  62                 p=a[n]
  63         side=!side
  64         print "[page " p "]"
  65         getline
  66         getline
  67         last=$0
  68         next
  69 }
  70 {
  71         print last
  72         last=$0
  73 }
  74 '