chore(unitproc/bkt-replace_apostrophes)

[BK-2020-03.git] / unitproc / bkt-replace_apostrophes

#!/bin/bash

# Function to replace contractions
replace_apostrophes() {
    # Desc: Replace ' with ʼ in contractions
    # Note: In contractions of UTF-8 text file, replaces U+0027
    #   APOSTROPHE with U+02BC MODIFIER LETTER APOSTROPHE
    # Input: stdin
    #        arg1   file path
    # Output: stdout
    # Version: 1.0.0 (BK-2020-03)
    # Depends: GNU sed 4.8

    # Check input
    if [[ "$#" -gt 1 ]]; then
        echo "FATAL:Incorrect argument count:$#" 1>&2;
        return 1;
    fi;

    if [[ -f "$1" ]]; then
        # Use specified file
        input="$1";
    else
        # Use standard input
        input="-";
    fi;

    # Perform substitutions
    ## Note: See https://en.wiktionary.org/wiki/Category:English_contractions
    ## Note: Order of replacements sorted most-specific first.
    sed -E \
        -e "s/(you|You|YOU)'(ren|REN|ven|VEN)'(t|T)/\1ʼ\2ʼ\3/g" \
        -e "s/(you|You|YOU)'(dn|DN)'(t|T)'(ve|VE)/\1ʼ\2ʼ\3ʼ\4/g" \
        -e "s/(you|You|YOU)'(d|D|ll|LL)'(ve|VE)/\1ʼ\2ʼ\3/g" \
        -e "s/(you|You|YOU)'(d|D|ll|LL|re|RE|ve|VE)/\1ʼ\2/g" \
        -e "s/(y|Y)'(all|ALL)'(dn|DN)'(t|T)'(ve|VE)/\1ʼ\2ʼ\3ʼ\4ʼ\5/g" \
        -e "s/(y|Y)'(all|ALL)'(d|D)'(nt|NT)'(ve|VE)/\1ʼ\2ʼ\3ʼ\4ʼ\5/g" \
        -e "s/(y|Y)'(all|ALL)'(d|D)'(ve|VE)/\1ʼ\2ʼ\3ʼ\4/g" \
        -e "s/(y|Y)'(all|ALL)'(d|D|ll|LL|re|RE|ve|VE)/\1ʼ\2ʼ\3/g" \
        -e "s/(y|Y)'(all|ALL)/\1ʼ\2/g" \
        -e "s/(y|Y)'(ain|AIN)'(t|T)/\1ʼ\2ʼ\3/g" \
        -e "s/(wouldn|Wouldn|WOULDN)'(t|T)'(ve|VE)/\1ʼ\2ʼ\3/g" \
        -e "s/(wouldn|Wouldn|WOULDN)'(t|T)/\1ʼ\2/g" \
        -e "s/(won|Won|WON)'(t|T)'(ve|VE)/\1ʼ\2ʼ\3/g" \
        -e "s/(won|Won|WON)'(t|T)/\1ʼ\2/g" \
        -e "s/(who|Who|WHO)'(d|D|ll|LL)'(ve|VE)/\1ʼ\2ʼ\3/g" \
        -e "s/(who|Who|WHO)'(d|D|ll|LL|re|RE|s|S|ve|VE)/\1ʼ\2/g" \
        -e "s/(where|Where|WHERE)'(d|D|s|S)/\1ʼ\2/g" \
        -e "s/(what|What|WHAT)'(ll|LL|re|RE|s|S|ve|VE)/\1ʼ\2/g" \
        -e "s/(weren|Weren|WEREN)'(t|T)/\1ʼ\2/g" \
        -e "s/(we|We|WE)'(ven|VEN)'(t|T)/\1ʼ\2ʼ\3/g" \
        -e "s/(we|We|WE)'(ren|REN)'(t|T)/\1ʼ\2ʼ\3/g" \
        -e "s/(we|We|WE)'(d|D|ll|LL)'(ve|VE)/\1ʼ\2ʼ\3/g" \
        -e "s/(we|We|WE)'(d|D|ll|LL|re|RE|ve|VE)/\1ʼ\2/g" \
        -e "s/(wasn|Wasn|WASN)'(t|T)/\1ʼ\2/g" \
        -e "s/(they|They|THEY)'(d|D|ll|LL)'(ve|VE)/\1ʼ\2ʼ\3/g" \
        -e "s/(they|They|THEY)'(d|D|ll|LL|re|RE|ve|VE)/\1ʼ\2/g" \
        -e "s/(there|There|THERE)'(ll|LL)'(ve|VE)/\1ʼ\2ʼ\3/g" \
        -e "s/(there|There|THERE)'(s|S|ve|VE)/\1ʼ\2/g" \
        -e "s/(that|That|THAT)'(d|D|ll|LL)'(ve|VE)/\1ʼ\2ʼ\3/g" \
        -e "s/(that|That|THAT)'(d|D|ll|LL|s|S)/\1ʼ\2/g" \
        -e "s/(shouldn|Shouldn|SHOULDN)'(t|T)'(ve|VE)/\1ʼ\2ʼ\3/g" \
        -e "s/(shouldn|Shouldn|SHOULDN)'(t|T)/\1ʼ\2/g" \
        -e "s/(she|She|SHE)'(d|D|ll|LL)'(ve|VE)/\1ʼ\2ʼ\3/g" \
        -e "s/(she|She|SHE)'(d|D|ll|LL|s|S)/\1ʼ\2/g" \
        -e "s/(shan|Shan|SHAN)'(t|T)/\1ʼ\2/g" \
        -e "s/'(s\b)/ʼ\1/g" \
        -e "s/(oughtn|Oughtn|OUGHTN)'(t|T)'(ve|VE)/\1ʼ\2ʼ\3/g" \
        -e "s/(oughtn|Oughtn|OUGHTN)'(t|T)/\1ʼ\2/g" \
        -e "s/(o|O)'(clock|CLOCK)/\1ʼ\2/g" \
        -e "s/(mustn|Mustn|MUSTN)'(t|T)/\1ʼ\2/g" \
        -e "s/(mustn|Mustn|MUSTN)'(t|T)'(ve|VE)/\1ʼ\2ʼ\3/g" \
        -e "s/(mustn|Mustn|MUSTN)'(t|T)/\1ʼ\2/g" \
        -e "s/(mightn|Mightn|MIGHTN)'(t|T)'(ve|VE)/\1ʼ\2ʼ\3/g" \
        -e "s/(mightn|Mightn|MIGHTN)'(t|T)/\1ʼ\2/g" \
        -e "s/(might|Might|MIGHT)'(ve|VE)/\1ʼ\2/g" \
        -e "s/(let|Let|LET)'(s|S)/\1ʼ\2/g" \
        -e "s/(it|It|IT)'(d|D|ll|LL)'(ve|VE)/\1ʼ\2ʼ\3/g" \
        -e "s/(it|It|IT)'(d|D|ll|LL|s|S)/\1ʼ\2/g" \
        -e "s/(isn|Isn|ISN)'(t|T)/\1ʼ\2/g" \
        -e "s/(I|i)'(dn|DN)'(t|T)'(ve|VE)/\1ʼ\2ʼ\3ʼ\4/g" \
        -e "s/(I|i)'(d|D|ll|LL)'(ve|VE)/\1ʼ\2ʼ\3/g" \
        -e "s/(I|i)'(d|D|ll|LL|m|M|ve|VE)/\1ʼ\2/g" \
        -e "s/(how|How|HOW)'(d|D)/\1ʼ\2/g" \
        -e "s/(he|He|HE)'(d|D|ll|LL)'(ve|VE)/\1ʼ\2ʼ\3/g" \
        -e "s/(he|He|HE)'(d|D|ll|LL|s|S)/\1ʼ\2/g" \
        -e "s/(haven|Haven|HAVEN)'(t|T)/\1ʼ\2/g" \
        -e "s/(hasn|Hasn|HASN)'(t|T)/\1ʼ\2/g" \
        -e "s/(hadn|Hadn|HADN)'(t|T)'(ve|VE)/\1ʼ\2ʼ\3/g" \
        -e "s/(hadn|Hadn|HADN)'(t|T)/\1ʼ\2/g" \
        -e "s/'(em\b)/ʼ\1/g" \
        -e "s/(d|D)'(ya|YA|you|YOU)/\1ʼ\2/g" \
        -e "s/(don|Don|DON)'(t|T)/\1ʼ\2/g" \
        -e "s/(doesn|Doesn|DOESN)'(t|T)/\1ʼ\2/g" \
        -e "s/(didn|Didn|DIDN)'(t|T)/\1ʼ\2/g" \
        -e "s/(could|Could|COULD)'(ve|VE)/\1ʼ\2/g" \
        -e "s/(couldn|Couldn|COULDN)'(t|T)'(ve|VE)/\1ʼ\2ʼ\3/g" \
        -e "s/(couldn|Couldn|COULDN)'(t|T)/\1ʼ\2/g" \
        -e "s/(c|C)'(mere|MERE)/\1ʼ\2/g" \
        -e "s/(can|Can|CAN)'(t|T)'(ve|VE)/\1ʼ\2ʼ\3/g" \
        -e "s/(can|Can|CAN)'(t|T)/\1ʼ\2/g" \
        -e "s/'(cause|Cause|CAUSE)/ʼ\1/g" \
        -e "s/'(bout|Bout|BOUT)/ʼ\1/g" \
        -e "s/(aren|Aren|AREN)'(t|T)/\1ʼ\2/g" \
        -e "s/(ate|Ate|ATE)'(nt|NT)/\1ʼ\2/g" \
        -e "s/(ain|Ain|AIN)'(t|T)/\1ʼ\2/g" \
        "$input"
}; # replace ' with ʼ in contractions

# Author: Steven Baltakatei Sandoval
# License: GPLv3+