basis/unicode/collation/collation.factor

   1 ! Copyright (C) 2008 Daniel Ehrenberg.
   2 ! See http://factorcode.org/license.txt for BSD license.
   3 USING: accessors arrays assocs combinators
   4 combinators.short-circuit combinators.smart fry kernel locals make
   5 math math.order math.parser namespaces sequences
   6 simple-flat-file splitting strings unicode.data ;
   7 IN: unicode.collation
   8
   9 <PRIVATE
  10
  11 SYMBOL: ducet
  12
  13 TUPLE: weight-levels primary secondary tertiary ignorable? ;
  14 : <weight-levels> ( primary secondary tertiary -- weight-levels> )
  15     weight-levels new
  16         swap >>tertiary
  17         swap >>secondary
  18         swap >>primary ; inline
  19
  20 : parse-weight ( string -- weight )
  21     "]" split but-last [
  22         weight-levels new swap rest unclip CHAR: * = swapd >>ignorable?
  23         swap "." split first3 [ hex> ] tri@
  24         [ >>primary ] [ >>secondary ] [ >>tertiary ] tri*
  25     ] map ;
  26
  27 : parse-keys ( string -- chars )
  28     " " split [ hex> ] "" map-as ;
  29
  30 : parse-ducet ( file -- ducet )
  31     load-data-file [ [ parse-keys ] [ parse-weight ] bi* ] H{ } assoc-map-as ;
  32
  33 "vocab:unicode/UCA/allkeys.txt" parse-ducet ducet set-global
  34
  35 ! Fix up table for long contractions
  36 : help-one ( assoc key -- )
  37     ! Need to be more general? Not for DUCET, apparently
  38     2 head 2dup swap key? [ 2drop ] [
  39         [ [ 1string of ] with { } map-as concat ]
  40         [ swap set-at ] 2bi
  41     ] if ;
  42
  43 : fixup-ducet ( -- )
  44     {
  45         {
  46             { 0x0FB2 0x0F71 } ! CE(0FB2) CE(0F71)
  47             {
  48                 T{ weight-levels
  49                     { primary 12719 }
  50                     { secondary 32 }
  51                     { tertiary 2 }
  52                 }
  53                 T{ weight-levels
  54                     { primary 12741 }
  55                     { secondary 32 }
  56                     { tertiary 2 }
  57                 }
  58             }
  59         }
  60         {
  61             { 0x0FB3 0x0F71 } ! CE(0FB3) CE(0F71)
  62             {
  63                 T{ weight-levels
  64                     { primary 12720 }
  65                     { secondary 32 }
  66                     { tertiary 2 }
  67                 }
  68                 T{ weight-levels
  69                     { primary 12741 }
  70                     { secondary 32 }
  71                     { tertiary 2 }
  72                 }
  73             }
  74         }
  75
  76         ! FIXME: WRONG WEIGHTS
  77         {
  78             { 0x0FB2 0x0F71 0x0F72 } ! CE(0FB2) CE(0F71 0F72)
  79             {
  80                 T{ weight-levels
  81                     { primary 12719 }
  82                     { secondary 32 }
  83                     { tertiary 2 }
  84                 }
  85                 T{ weight-levels
  86                     { primary 12741 }
  87                     { secondary 32 }
  88                     { tertiary 2 }
  89                 }
  90             }
  91         }
  92         {
  93             { 0x0FB2 0x0F73        } ! CE(0FB2) CE(0F71 0F72)
  94             {
  95                 T{ weight-levels
  96                     { primary 12719 }
  97                     { secondary 32 }
  98                     { tertiary 2 }
  99                 }
 100                 T{ weight-levels
 101                     { primary 12741 }
 102                     { secondary 32 }
 103                     { tertiary 2 }
 104                 }
 105             }
 106         }
 107         {
 108             { 0x0FB2 0x0F71 0x0F74 } ! CE(0FB2) CE(0F71 0F74)
 109             {
 110                 T{ weight-levels
 111                     { primary 12719 }
 112                     { secondary 32 }
 113                     { tertiary 2 }
 114                 }
 115                 T{ weight-levels
 116                     { primary 12741 }
 117                     { secondary 32 }
 118                     { tertiary 2 }
 119                 }
 120             }
 121         }
 122         {
 123             { 0x0FB2 0x0F75        } ! CE(0FB2) CE(0F71 0F74)
 124             {
 125                 T{ weight-levels
 126                     { primary 12719 }
 127                     { secondary 32 }
 128                     { tertiary 2 }
 129                 }
 130                 T{ weight-levels
 131                     { primary 12741 }
 132                     { secondary 32 }
 133                     { tertiary 2 }
 134                 }
 135             }
 136         }
 137         {
 138             { 0x0FB3 0x0F71 0x0F72 } ! CE(0FB3) CE(0F71 0F72)
 139             {
 140                 T{ weight-levels
 141                     { primary 12719 }
 142                     { secondary 32 }
 143                     { tertiary 2 }
 144                 }
 145                 T{ weight-levels
 146                     { primary 12741 }
 147                     { secondary 32 }
 148                     { tertiary 2 }
 149                 }
 150             }
 151         }
 152         {
 153             { 0x0FB3 0x0F73        } ! CE(0FB3) CE(0F71 0F72)
 154             {
 155                 T{ weight-levels
 156                     { primary 12719 }
 157                     { secondary 32 }
 158                     { tertiary 2 }
 159                 }
 160                 T{ weight-levels
 161                     { primary 12741 }
 162                     { secondary 32 }
 163                     { tertiary 2 }
 164                 }
 165             }
 166         }
 167         {
 168             { 0x0FB3 0x0F71 0x0F74 } ! CE(0FB3) CE(0F71 0F74)
 169             {
 170                 T{ weight-levels
 171                     { primary 12719 }
 172                     { secondary 32 }
 173                     { tertiary 2 }
 174                 }
 175                 T{ weight-levels
 176                     { primary 12741 }
 177                     { secondary 32 }
 178                     { tertiary 2 }
 179                 }
 180             }
 181         }
 182         {
 183             { 0x0FB3 0x0F75        } ! CE(0FB3) CE(0F71 0F74)
 184             {
 185                 T{ weight-levels
 186                     { primary 12719 }
 187                     { secondary 32 }
 188                     { tertiary 2 }
 189                 }
 190                 T{ weight-levels
 191                     { primary 12741 }
 192                     { secondary 32 }
 193                     { tertiary 2 }
 194                 }
 195             }
 196         }
 197     } ducet get-global '[ swap >string _ set-at ] assoc-each ;
 198
 199 ! Add a few missing ducet values
 200 fixup-ducet
 201
 202 : tangut-block? ( char -- ? )
 203     ! Tangut Block, Tangut Components Block
 204     { [ 0x17000 0x187FF between? ] [ 0x18800 0x18AFF between? ] } 1|| ; inline
 205
 206 ! https://wiki.computercraft.cc/Module:Unicode_data
 207 ! Unicode TR10 - Computing Implicit Weights
 208 : base ( char -- base )
 209     {
 210         { [ dup 0x03400 0x04DB5 between? ] [ drop 0xFB80 ] } ! Extension A
 211         { [ dup 0x20000 0x2A6D6 between? ] [ drop 0xFB80 ] } ! Extension B
 212         { [ dup 0x2A700 0x2B734 between? ] [ drop 0xFB80 ] } ! Extension C
 213         { [ dup 0x2B740 0x2B81D between? ] [ drop 0xFB80 ] } ! Extension D
 214         { [ dup 0x2B820 0x2CEA1 between? ] [ drop 0xFB80 ] } ! Extension E
 215         { [ dup 0x2CEB0 0x2EBE0 between? ] [ drop 0xFB80 ] } ! Extension F
 216         { [ dup 0x04E00 0x09FD5 between? ] [ drop 0xFB40 ] } ! CJK
 217         [ drop 0xFBC0 ] ! Other
 218     } cond ;
 219
 220 : tangut-AAAA ( char -- weight-levels )
 221     drop 0xfb00 0x0020 0x0002 <weight-levels> ; inline
 222
 223 : tangut-BBBB ( char -- weight-levels )
 224     0x17000 - 0x8000 bitor 0 0 <weight-levels> ; inline
 225
 226 : AAAA ( char -- weight-levels )
 227     [ base ] [ -15 shift ] bi + 0x0020 0x0002 <weight-levels> ; inline
 228
 229 : BBBB ( char -- weight-levels )
 230     0x7FFF bitand 0x8000 bitor 0 0 <weight-levels> ; inline
 231
 232 : derive-weight ( 1string -- weight-levels-pair )
 233     first
 234     dup tangut-block? [
 235         [ tangut-AAAA ] [ tangut-BBBB ] bi 2array
 236     ] [
 237         [ AAAA ] [ BBBB ] bi 2array
 238     ] if ;
 239
 240 : building-last ( -- char )
 241     building get [ 0 ] [ last last ] if-empty ;
 242
 243 : blocked? ( char -- ? )
 244     combining-class dup { 0 f } member?
 245     [ drop building-last non-starter? ]
 246     [ building-last combining-class = ] if ;
 247
 248 : possible-bases ( -- slice-of-building )
 249     building get dup [ first non-starter? not ] find-last
 250     drop [ 0 ] unless* tail-slice ;
 251
 252 :: ?combine ( char slice i -- ? )
 253     i slice nth char suffix :> str
 254     str ducet get-global key? dup
 255     [ str i slice set-nth ] when ;
 256
 257 : add ( char -- )
 258     dup blocked? [ 1string , ] [
 259         dup possible-bases dup length <iota>
 260         [ ?combine ] 2with any?
 261         [ drop ] [ 1string , ] if
 262     ] if ;
 263
 264 : string>graphemes ( string -- graphemes )
 265     [ [ add ] each ] { } make ;
 266
 267 : char>weight-levels ( 1string -- weight-levels )
 268     ducet get-global ?at [ derive-weight ] unless ; inline
 269
 270 : graphemes>weights ( graphemes -- weights )
 271     [
 272         dup weight-levels?
 273         [ 1array ] ! From tailoring
 274         [ char>weight-levels ] if
 275     ] { } map-as concat ;
 276
 277 : append-weights ( weight-levels quot -- seq )
 278     [ [ ignorable?>> ] reject ] dip
 279     map [ zero? ] reject ; inline
 280
 281 : variable-weight ( weight-levels -- obj )
 282     dup ignorable?>> [ primary>> ] [ drop 0xFFFF ] if ;
 283
 284 : weights>bytes ( weights -- array )
 285     [
 286         {
 287             [ [ primary>> ] append-weights { 0 } ]
 288             [ [ secondary>> ] append-weights { 0 } ]
 289             [ [ tertiary>> ] append-weights { 0 } ]
 290             [ [ [ secondary>> ] [ tertiary>> ] bi [ zero? ] bi@ and not ] filter [ variable-weight ] map ]
 291         } cleave
 292     ] { } append-outputs-as ;
 293
 294 PRIVATE>
 295
 296 : completely-ignorable? ( weight -- ? )
 297     {
 298         [ primary>> zero? ]
 299         [ secondary>> zero? ]
 300         [ tertiary>> zero? ]
 301     } 1&& ;
 302
 303 : filter-ignorable ( weights -- weights' )
 304     f swap [
 305         [ nip ] [ primary>> zero? and ] 2bi
 306         [ swap ignorable?>> or ]
 307         [ swap completely-ignorable? or not ] 2bi
 308     ] filter nip ;