#!/usr/bin/perl -w
# Script to create a collation file for use in 
# International Components for Unicode (ICU)
#
# Keith Stribley <devel@thanlwinsoft.org>
# Distributable under the same terms as either the ICU license or the CLDR.
#
# This collation implements Myanmar Spelling Book Order.
# Note: Pali order would be much easier to implement, but
# is no longer used. 
# This collation is very inefficent because it uses
# a huge number of collating elements to get the right
# sort order 
# This is the only way that I can see of doing it within the 
# limitations of the locale file format. However, if you know
# a better way please let me know.
#
# Installation: 
# type 
# perl genMyCollateICU.pl > my.txt
# Copy my.txt into the source/data/coll directory of ICU, add my.txt to 
# colfiles.mk in the same directory and build ICU

local $consonants;
local $vowels;
local $tones;
local $yrwh;
local $killed;

# Myanmar sorting is done by clusters
# The sorting order is:
# 1. Base consonant
# 2. YRWH medials
# 3. killed consonant
# 4. vowel
# 5. tone
# Most of these are easy because the sorting order is the same as the 
# storage order for 1,2 and 5. However, 4 is stored before 3 so all
# permutations of 3 and 4 need to be defined as collating elements
# and listed between 2 and 5.
#

# Burmese consonants that may take U+103A
@consonants = (
("\\u1000"),#KA
("\\u1001"),#KHA,
("\\u1002"),#GA
"\\u1003",#GHA
("\\u1004",),#NGA
("\\u1005"),#CA
"\\u1006",#CHA
("\\u1007"),#JA,
("\\u1008"),#JHA,
("\\u1009"),#NYA
"\\u100A",#NNYA
"\\u100B",#TTA
"\\u100C",#TTHA
"\\u100D",#DDA
"\\u100E",#DDHA
("\\u100F"),#NNA
"\\u1010",#TA
"\\u1011",#THA
("\\u1012"),#DA
"\\u1013",#DHA
("\\u1014"),#NA
"\\u1015",#PA
("\\u1016"),#PHA
("\\u1017"),#BA
"\\u1018",#BHA
"\\u1019",#MA
"\\u101A",#YA
"\\u101B",#RA
"\\u101C",#LA
"\\u101D",#WA
("\\u101E"),#SA,
("\\u101F"),#HA
("\\u1020"),#LLA
("\\u1021")
);

@yrwh = (
"",
"\\u105E",#mon medial na
"\\u105F",#mon medial ma
"\\u103B",
"\\u103C",
"\\u1060",#mon medial la
"\\u103D",
"\\u1082",# shan medial wa
"\\u103E",
"\\u103B\\u103D",
"\\u103C\\u103D",
"\\u103B\\u103E",
"\\u103C\\u103E",
"\\u103D\\u103E",
"\\u103B\\u103D\\u103E",
"\\u103C\\u103D\\u103E"
);

@vowels = (
"",
"\\u102C",#aa
"\\u1083",#shan aa
"\\u1072",#kayah oe
"\\u102D",#i
"\\u1071",#Geba Karen i
"\\u102E",#ii
"\\u1033",#mon ii
"\\u1067",#w pwo eu
"\\u1068",#w pwo ue
"\\u102F",#u
"\\u1073",#kayah u
"\\u1062",#Sgaw eu
"\\u1074",#kayah ee
"\\u1030",#uu
"\\u1056",#vocalic r
"\\u1057",#vocalic rr
"\\u1058",#vocalic l 
"\\u1059",#vocalic ll
"\\u1031",#e
"\\u1084",#shan e
"\\u1035",# e above
"\\u1085",# shan e above
"\\u1032",#ai
"\\u1031\\u102C",# aw High tone
"\\u1031\\u102C\\u103A",#aw low tone
"\\u1034",#mon o
"\\u1036",# an
"\\u102D\\u102F",#o
"\\u1063",#Sgaw Hathi
"\\u1086"#shan final y
);

@vowelsTakeFinal = (
"",
"\\u102C",#1
"\\u102D",#2
"\\u102F",#4
"\\u1031",#6 e.g. foreign words
"\\u1031\\u102C",#8 High tone
"\\u102D\\u102F"#13
);


@killed = (
"\\u1000",
"\\u1001",
"\\u1002",
"\\u1003",
"\\u1004",
"\\u1005",
"\\u1006",
"\\u1007",
"\\u1009",
"\\u100A",
"\\u100B",
"\\u100C",
"\\u100D",
"\\u100E",
"\\u100F",
"\\u1010",
"\\u1011",
"\\u1012",
"\\u1013",
"\\u1014",
"\\u1015",
"\\u1016",
"\\u1017",
"\\u1018",
"\\u1019",
"\\u101A",
"\\u101B",
"\\u101C",
"\\u101E",
"\\u101F",
""
);

# the following includes tones from several scripts, the order chosen here
# groups tones of the same number together
@tones = (
"",
"\\u1069",#w-pwo-1
"\\u1037",#dot-below
"\\u1087",#shan-2
"\\u108B",#shan-council-2
"\\u106A",#w-pwo-2
"\\u1062\\u103A",#sgaw karen
"\\u1088",#shan-3
"\\u108C",#shan-council-3
"\\u106B",#w-pwo-3
"\\u102C\\u103A",#sgaw karen
"\\u1038",#visarga
"\\u106C",#w-pwo-4
"\\u1037\\u1038",#sgaw karen
"\\u1089",#shan-5
"\\u106D",#w-pwo-5
"\\u1063\\u103A",#sgaw karen
"\\u108A",#shan-6
"\\u108D",#shan-council-emphatic
"\\u108F",#palaung tone 6
"\\u1064"#sgaw karen
);

# start file
print << 'EOT';
// Myanmar Collation for ICU
// Developed by www.ThanLwinSoft.org
// Distributable under the same terms as the ICU license.
// This file is generated by genMyCollateICU.pl - do not hand edit
my {
    Version { "1.0" }
    collations{
        standard{
            Sequence{
[normalization on]
EOT

# define collating elements

printf("/* Tones */\n");
# Tones are just secondary, tertiary in UCA, but this gives the wrong results
# when you have a following syllable e.g. consider
# 1000 102C
# 1000 102C 1000
# 1000 102C 1038
# Where should the Burmese tones be inserted? 
# 108C is consistent with Shan which also uses 1038 (see http://std.dkuug.dk/jtc1/sc2/wg2/docs/n3277.pdf page 4)
# but it isn't very important if you only collate Burmese so long as the 
# \u1037, \u1038 relative order is present
print << 'EOT';
"&\u108C"
"<\u1037"
"<\u1038"
"<\u1037\u1038"
/* Vowels */
"&\u102C"
"<<\u102B"
"&\u1032"
"<\u1031\u102C"
"<<\u1031\u102B"
"<\u1031\u102C\u103A"
"<<\u1031\u102B\u103A"
"&\u1034"
"<\u1036"
"<\u102D\u102F"
EOT

printf("/* Vowels with finals */\n");
# The final takes precedence over the vowel, but occurs after it, hence they 
# all need to be listed explicitly.
$iKilled = 0;

# vowels with killed consonant
for ($c = 0; $c<= $#consonants; $c++)
{
  $consonant = $consonants[$c];
  if ($killed[$iKilled] eq $consonant)
  {
     #printf STDERR "$consonant\n";
     $iKilled++;
     if ($consonant eq "\\u1004") # kinzi
     {
       printf("\"<%s\\u103A\\u1039\"\n",$consonant);
     }
     else
     {
       printf("\"<%s\\u1039\"\n",$consonant);
     }
     printf("\"<<%s\\u103A\"\n",$consonant);
     for ($v = 1; $v<=$#vowelsTakeFinal; $v++)
     {
         $theVowel =  $vowelsTakeFinal[$v];
         if ($consonant eq "\\u1004") # kinzi
         {
           printf("\"<%s%s\\u103A\\u1039\"\n",$vowelsTakeFinal[$v],$consonant);
         }
         else
         {
           printf("\"<%s%s\\u1039\"\n",$vowelsTakeFinal[$v],$consonant);
         }
         printf("\"<<%s%s\\u103A\"\n",$vowelsTakeFinal[$v],$consonant);
         if ($theVowel =~ s/\\u102C/\\u102B/)
         {
            if ($consonant eq "\\u1004") # kinzi
            {
               printf("\"<<%s%s\\u103A\\u1039\"\n",$theVowel,$consonant);
            }
            else
            {
               printf("\"<<%s%s\\u1039\"\n",$theVowel,$consonant);
            }
           printf("\"<<%s%s\\u103A\"\n",$theVowel,$consonant);
         }
     }
  }
  else
  {
     printf("\"<%s\\u103A\"\n",$consonant);
     for ($v = 1; $v<=$#vowelsTakeFinal; $v++)
     {
         $theVowel =  $vowelsTakeFinal[$v];
         printf("\"<%s%s\\u103A\"\n",$vowelsTakeFinal[$v],$consonant);
         if ($theVowel =~ s/\\u102C/\\u102B/)
         {
           printf("\"<<%s%s\\u103A\"\n",$theVowel,$consonant);
         }
     }
  }
}

printf("/* Medials */\n");
for ($y = 1; $y<=$#yrwh; $y++)
{
    if ($y == 1)
    {
        printf("\"<%s\"\n", $yrwh[$y]);
    }
    else
    {
        printf("\"<%s\"\n", $yrwh[$y]);
    }
}

printf("\n"); 

# It is debatable whether to make the \u102D\u1036 and \u102F\u1036 difference 
# from the \u1019 equivalent primary or secondary
# primary gives a closer match with the Myanmar/English dictionary
# secondary gives a better match with the preface to the Myanmar spelling dictionary
# It is hard to produce a better match because you need to include 1037, 1038,
# which distorts the other rules, probably due to triggering unwanted expansions.
print << 'EOT';

/* Independent vowels */
"&\u1021\u102d<<<\u1023"

"&\u1021\u102e<<<\u1024"

"&\u1021\u102f<<<\u1025"

"&\u1021\u1030<<<\u1026=\u1025\u102e"

"&\u1021\u1031<<<\u1027<<<\u1028"

"&\u1021\u1031\u102c<<<\u1029"

"&\u1021\u1031\u102c\u103A<<<\u102A"

"&\u102D\u1019\u103A<\u102D\u1036<\u102D\u1019\u103A\u1037<\u102D\u1036\u1037"
"<\u102D\u1019\u103A\u1038<\u102D\u1036\u1038"
"&\u102F\u1019\u103A<\u102F\u1036<\u102F\u1019\u103A\u1037<\u102F\u1036\u1037"
"<\u102F\u1019\u103A\u1038<\u102F\u1036\u1038"

"&\u1021\u102F\u1036<<<\u1025\u102F\u1036"

/* Independent vowels with finals */
EOT

# combinations of finals with independent vowels / finals
for ($k = 0; $k<$#killed; $k++)
{
	printf("\"&\\u1021\\u102d%s\\u1039<<<\\u1023%s\\u1039\"\n", $killed[$k], $killed[$k]);
	printf("\"&\\u1021\\u102d%s\\u103A<<<\\u1023%s\\u103A\"\n", $killed[$k], $killed[$k]);
	printf("\"&\\u1021\\u102f%s\\u1039<<<\\u1025%s\\u1039\"\n", $killed[$k], $killed[$k]);
	printf("\"&\\u1021\\u102f%s\\u103A<<<\\u1025%s\\u103A\"\n", $killed[$k], $killed[$k]);
	printf("\"&\\u1021\\u1031%s\\u1039<<<\\u1027%s\\u1039\"\n", $killed[$k], $killed[$k]);
	printf("\"&\\u1021\\u1031%s\\u103A<<<\\u1027%s\\u103A\"\n", $killed[$k], $killed[$k]);
	printf("\"&\\u1021\\u1031\\u102C%s\\u1039<<<\\u1029%s\\u1039\"\n", $killed[$k], $killed[$k]);
	printf("\"&\\u1021\\u1031\\u102C%s\\u103A<<<\\u1029%s\\u103A\"\n", $killed[$k], $killed[$k]);
# I think the next one only occurs in the word ဩောင်း
	printf("\"&\\u1021\\u1031\\u102C%s\\u1039<<<\\u1029\\u1031\\u102C%s\\u1039\"\n", $killed[$k], $killed[$k]);
	printf("\"&\\u1021\\u1031\\u102C%s\\u103A<<<\\u1029\\u1031\\u102C%s\\u103A\"\n", $killed[$k], $killed[$k]);
}

print << 'EOT';

EOT

  print << 'EOT';

/* Contractions */
"&\u1031\u102C\u1000\u103A\u1000\u103B=\u1031\u102C\u1000\u103A\u103B"/* (suffix of) man */
"&\u1014\u103A\u1014\u102F\u1015\u103A=\u1014\u103A\u102F\u1015\u103A" /* (suffix of) I */
/* Great Sa */
"&\u1031\u101E\u1039\u101E=\u1031\u103F"
"&\u102D\u101E\u1039\u101E=\u102D\u103F"
"&\u102F\u101E\u1039\u101E=\u102F\u103F"
"&\u1021\u102D\u101E\u1039\u101E<<<\u1023\u103F"
"&\u1021\u102F\u101E\u1039\u101E<<<\u1025\u103F"
"&\u101E\u1039\u101E=\u103F"

/* Symbols - collate as long form */
"&\u1014\u103E\u102D\u102F\u1000\u103A<<\u104C"
"&\u101B\u103D\u1031\u1037<<\u104D"
"&\u101C\u100A\u103A\u1038\u1000\u1031\u102C\u1004\u103A\u1038<<\u104E\u1004\u103A\u1038"
"&\u1021\u102D<<\u104F"

/* Short Forms may need to be added here */
"&\u101C\u1000\u103A\u101A\u102C=\u101C\u1000\u103A\u103B\u102C" /* right hand side */
"&\u101E\u1019\u102E=\u101E\u1039\u1019\u102E" /* daughter */
"&\u1011\u1019\u1004\u103A\u1038=\u1011\u1039\u1019\u1004\u103A\u1038" /* cooked rice */
"&\u101C\u1000\u103A\u1018\u1000\u103A=\u101C\u1039\u1018\u1000\u103A" /*tea*/

      }
      Version{"1.0"}
    }
  }
}
EOT