[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+
Commit	Line	Data
265d3d9a SBS	1	#!/bin/bash
	2
	3	# Function to replace contractions
eb9061ae	4	replace_apostrophes() {
265d3d9a SBS	5	# Desc: Replace ' with ʼ in contractions
	6	# Note: In contractions of UTF-8 text file, replaces U+0027
	7	# APOSTROPHE with U+02BC MODIFIER LETTER APOSTROPHE
	8	# Input: stdin
	9	# arg1 file path
	10	# Output: stdout
eb9061ae	11	# Version: 1.0.0 (BK-2020-03)
265d3d9a SBS	12	# Depends: GNU sed 4.8
	13
	14	# Check input
	15	if [[ "$#" -gt 1 ]]; then
	16	echo "FATAL:Incorrect argument count:$#" 1>&2;
	17	return 1;
	18	fi;
	19
	20	if [[ -f "$1" ]]; then
	21	# Use specified file
	22	input="$1";
	23	else
	24	# Use standard input
	25	input="-";
	26	fi;
	27
	28	# Perform substitutions
7e653610 SBS	29	## Note: See https://en.wiktionary.org/wiki/Category:English_contractions
7e653610 SBS	30	## Note: Order of replacements sorted most-specific first.
265d3d9a	31	sed -E \
7e653610 SBS	32	-e "s/(you\|You\|YOU)'(ren\|REN\|ven\|VEN)'(t\|T)/\1ʼ\2ʼ\3/g" \
	33	-e "s/(you\|You\|YOU)'(dn\|DN)'(t\|T)'(ve\|VE)/\1ʼ\2ʼ\3ʼ\4/g" \
	34	-e "s/(you\|You\|YOU)'(d\|D\|ll\|LL)'(ve\|VE)/\1ʼ\2ʼ\3/g" \
	35	-e "s/(you\|You\|YOU)'(d\|D\|ll\|LL\|re\|RE\|ve\|VE)/\1ʼ\2/g" \
	36	-e "s/(y\|Y)'(all\|ALL)'(dn\|DN)'(t\|T)'(ve\|VE)/\1ʼ\2ʼ\3ʼ\4ʼ\5/g" \
	37	-e "s/(y\|Y)'(all\|ALL)'(d\|D)'(nt\|NT)'(ve\|VE)/\1ʼ\2ʼ\3ʼ\4ʼ\5/g" \
	38	-e "s/(y\|Y)'(all\|ALL)'(d\|D)'(ve\|VE)/\1ʼ\2ʼ\3ʼ\4/g" \
	39	-e "s/(y\|Y)'(all\|ALL)'(d\|D\|ll\|LL\|re\|RE\|ve\|VE)/\1ʼ\2ʼ\3/g" \
	40	-e "s/(y\|Y)'(all\|ALL)/\1ʼ\2/g" \
	41	-e "s/(y\|Y)'(ain\|AIN)'(t\|T)/\1ʼ\2ʼ\3/g" \
	42	-e "s/(wouldn\|Wouldn\|WOULDN)'(t\|T)'(ve\|VE)/\1ʼ\2ʼ\3/g" \
	43	-e "s/(wouldn\|Wouldn\|WOULDN)'(t\|T)/\1ʼ\2/g" \
	44	-e "s/(won\|Won\|WON)'(t\|T)'(ve\|VE)/\1ʼ\2ʼ\3/g" \
	45	-e "s/(won\|Won\|WON)'(t\|T)/\1ʼ\2/g" \
	46	-e "s/(who\|Who\|WHO)'(d\|D\|ll\|LL)'(ve\|VE)/\1ʼ\2ʼ\3/g" \
	47	-e "s/(who\|Who\|WHO)'(d\|D\|ll\|LL\|re\|RE\|s\|S\|ve\|VE)/\1ʼ\2/g" \
eb9061ae	48	-e "s/(where\|Where\|WHERE)'(d\|D\|s\|S)/\1ʼ\2/g" \
7e653610 SBS	49	-e "s/(what\|What\|WHAT)'(ll\|LL\|re\|RE\|s\|S\|ve\|VE)/\1ʼ\2/g" \
	50	-e "s/(weren\|Weren\|WEREN)'(t\|T)/\1ʼ\2/g" \
	51	-e "s/(we\|We\|WE)'(ven\|VEN)'(t\|T)/\1ʼ\2ʼ\3/g" \
	52	-e "s/(we\|We\|WE)'(ren\|REN)'(t\|T)/\1ʼ\2ʼ\3/g" \
	53	-e "s/(we\|We\|WE)'(d\|D\|ll\|LL)'(ve\|VE)/\1ʼ\2ʼ\3/g" \
	54	-e "s/(we\|We\|WE)'(d\|D\|ll\|LL\|re\|RE\|ve\|VE)/\1ʼ\2/g" \
eb9061ae	55	-e "s/(wasn\|Wasn\|WASN)'(t\|T)/\1ʼ\2/g" \
7e653610 SBS	56	-e "s/(they\|They\|THEY)'(d\|D\|ll\|LL)'(ve\|VE)/\1ʼ\2ʼ\3/g" \
	57	-e "s/(they\|They\|THEY)'(d\|D\|ll\|LL\|re\|RE\|ve\|VE)/\1ʼ\2/g" \
	58	-e "s/(there\|There\|THERE)'(ll\|LL)'(ve\|VE)/\1ʼ\2ʼ\3/g" \
	59	-e "s/(there\|There\|THERE)'(s\|S\|ve\|VE)/\1ʼ\2/g" \
	60	-e "s/(that\|That\|THAT)'(d\|D\|ll\|LL)'(ve\|VE)/\1ʼ\2ʼ\3/g" \
	61	-e "s/(that\|That\|THAT)'(d\|D\|ll\|LL\|s\|S)/\1ʼ\2/g" \
	62	-e "s/(shouldn\|Shouldn\|SHOULDN)'(t\|T)'(ve\|VE)/\1ʼ\2ʼ\3/g" \
	63	-e "s/(shouldn\|Shouldn\|SHOULDN)'(t\|T)/\1ʼ\2/g" \
	64	-e "s/(she\|She\|SHE)'(d\|D\|ll\|LL)'(ve\|VE)/\1ʼ\2ʼ\3/g" \
	65	-e "s/(she\|She\|SHE)'(d\|D\|ll\|LL\|s\|S)/\1ʼ\2/g" \
	66	-e "s/(shan\|Shan\|SHAN)'(t\|T)/\1ʼ\2/g" \
eb9061ae	67	-e "s/'(s\b)/ʼ\1/g" \
7e653610 SBS	68	-e "s/(oughtn\|Oughtn\|OUGHTN)'(t\|T)'(ve\|VE)/\1ʼ\2ʼ\3/g" \
7e653610 SBS	69	-e "s/(oughtn\|Oughtn\|OUGHTN)'(t\|T)/\1ʼ\2/g" \
eb9061ae	70	-e "s/(o\|O)'(clock\|CLOCK)/\1ʼ\2/g" \
7e653610 SBS	71	-e "s/(mustn\|Mustn\|MUSTN)'(t\|T)/\1ʼ\2/g" \
	72	-e "s/(mustn\|Mustn\|MUSTN)'(t\|T)'(ve\|VE)/\1ʼ\2ʼ\3/g" \
	73	-e "s/(mustn\|Mustn\|MUSTN)'(t\|T)/\1ʼ\2/g" \
	74	-e "s/(mightn\|Mightn\|MIGHTN)'(t\|T)'(ve\|VE)/\1ʼ\2ʼ\3/g" \
	75	-e "s/(mightn\|Mightn\|MIGHTN)'(t\|T)/\1ʼ\2/g" \
	76	-e "s/(might\|Might\|MIGHT)'(ve\|VE)/\1ʼ\2/g" \
	77	-e "s/(let\|Let\|LET)'(s\|S)/\1ʼ\2/g" \
	78	-e "s/(it\|It\|IT)'(d\|D\|ll\|LL)'(ve\|VE)/\1ʼ\2ʼ\3/g" \
	79	-e "s/(it\|It\|IT)'(d\|D\|ll\|LL\|s\|S)/\1ʼ\2/g" \
	80	-e "s/(isn\|Isn\|ISN)'(t\|T)/\1ʼ\2/g" \
	81	-e "s/(I\|i)'(dn\|DN)'(t\|T)'(ve\|VE)/\1ʼ\2ʼ\3ʼ\4/g" \
	82	-e "s/(I\|i)'(d\|D\|ll\|LL)'(ve\|VE)/\1ʼ\2ʼ\3/g" \
	83	-e "s/(I\|i)'(d\|D\|ll\|LL\|m\|M\|ve\|VE)/\1ʼ\2/g" \
eb9061ae	84	-e "s/(how\|How\|HOW)'(d\|D)/\1ʼ\2/g" \
7e653610 SBS	85	-e "s/(he\|He\|HE)'(d\|D\|ll\|LL)'(ve\|VE)/\1ʼ\2ʼ\3/g" \
	86	-e "s/(he\|He\|HE)'(d\|D\|ll\|LL\|s\|S)/\1ʼ\2/g" \
	87	-e "s/(haven\|Haven\|HAVEN)'(t\|T)/\1ʼ\2/g" \
	88	-e "s/(hasn\|Hasn\|HASN)'(t\|T)/\1ʼ\2/g" \
	89	-e "s/(hadn\|Hadn\|HADN)'(t\|T)'(ve\|VE)/\1ʼ\2ʼ\3/g" \
	90	-e "s/(hadn\|Hadn\|HADN)'(t\|T)/\1ʼ\2/g" \
eb9061ae	91	-e "s/'(em\b)/ʼ\1/g" \
7e653610 SBS	92	-e "s/(d\|D)'(ya\|YA\|you\|YOU)/\1ʼ\2/g" \
	93	-e "s/(don\|Don\|DON)'(t\|T)/\1ʼ\2/g" \
	94	-e "s/(doesn\|Doesn\|DOESN)'(t\|T)/\1ʼ\2/g" \
	95	-e "s/(didn\|Didn\|DIDN)'(t\|T)/\1ʼ\2/g" \
	96	-e "s/(could\|Could\|COULD)'(ve\|VE)/\1ʼ\2/g" \
	97	-e "s/(couldn\|Couldn\|COULDN)'(t\|T)'(ve\|VE)/\1ʼ\2ʼ\3/g" \
	98	-e "s/(couldn\|Couldn\|COULDN)'(t\|T)/\1ʼ\2/g" \
eb9061ae	99	-e "s/(c\|C)'(mere\|MERE)/\1ʼ\2/g" \
7e653610 SBS	100	-e "s/(can\|Can\|CAN)'(t\|T)'(ve\|VE)/\1ʼ\2ʼ\3/g" \
7e653610 SBS	101	-e "s/(can\|Can\|CAN)'(t\|T)/\1ʼ\2/g" \
eb9061ae SBS	102	-e "s/'(cause\|Cause\|CAUSE)/ʼ\1/g" \
eb9061ae SBS	103	-e "s/'(bout\|Bout\|BOUT)/ʼ\1/g" \
7e653610 SBS	104	-e "s/(aren\|Aren\|AREN)'(t\|T)/\1ʼ\2/g" \
	105	-e "s/(ate\|Ate\|ATE)'(nt\|NT)/\1ʼ\2/g" \
	106	-e "s/(ain\|Ain\|AIN)'(t\|T)/\1ʼ\2/g" \
	107	"$input"
265d3d9a	108	}; # replace ' with ʼ in contractions
4aa4ed1b SBS	109
	110	# Author: Steven Baltakatei Sandoval
	111	# License: GPLv3+