version 1.13, 2005/01/05 22:13:40
|
version 1.32, 2007/10/17 16:05:22
|
Line 1
|
Line 1
|
\ UTF-8 handling 12dec04py |
\ UTF-8 handling 12dec04py |
|
|
\ Copyright (C) 2004 Free Software Foundation, Inc. |
\ Copyright (C) 2004,2005,2006 Free Software Foundation, Inc. |
|
|
\ This file is part of Gforth. |
\ This file is part of Gforth. |
|
|
Line 27 $80 Value max-single-byte
|
Line 27 $80 Value max-single-byte
|
: u8len ( u8 -- n ) |
: u8len ( u8 -- n ) |
dup max-single-byte u< IF drop 1 EXIT THEN \ special case ASCII |
dup max-single-byte u< IF drop 1 EXIT THEN \ special case ASCII |
$800 2 >r |
$800 2 >r |
BEGIN 2dup u>= WHILE 5 lshift r> 1+ >r REPEAT |
BEGIN 2dup u>= WHILE 5 lshift r> 1+ >r dup 0= UNTIL THEN |
2drop r> ; |
2drop r> ; |
|
|
: u8@+ ( u8addr -- u8addr' u ) |
: u8@+ ( u8addr -- u8addr' u ) |
count dup max-single-byte u< ?EXIT \ special case ASCII |
count dup max-single-byte u< ?EXIT \ special case ASCII |
|
dup $C2 u< IF UTF-8-err throw THEN \ malformed character |
$7F and $40 >r |
$7F and $40 >r |
BEGIN dup r@ and WHILE r@ xor |
BEGIN dup r@ and WHILE r@ xor |
6 lshift r> 5 lshift >r >r count |
6 lshift r> 5 lshift >r >r count |
Line 56 $80 Value max-single-byte
|
Line 57 $80 Value max-single-byte
|
|
|
:noname ( addr u -- char addr' u' ) |
:noname ( addr u -- char addr' u' ) |
\ !! the if here seems to work around some breakage, but not |
\ !! the if here seems to work around some breakage, but not |
\ entirely; e.g., try '็' with LANG=C. |
\ entirely; e.g., try 'รง' with LANG=C. |
dup 1 u<= IF defers char@ EXIT THEN |
dup 1 u<= IF defers char@ EXIT THEN |
over + >r u8@+ swap r> over - ; IS char@ |
over + >r u8@+ swap r> over - ; IS char@ |
|
|
Line 70 $80 Value max-single-byte
|
Line 71 $80 Value max-single-byte
|
|
|
\ utf key and emit |
\ utf key and emit |
|
|
|
Defer check-xy ' noop IS check-xy |
|
|
: u8key ( -- u ) |
: u8key ( -- u ) |
defers key dup max-single-byte u< ?EXIT \ special case ASCII |
defers key dup max-single-byte u< ?EXIT \ special case ASCII |
|
dup $FF = ?EXIT \ special resize character |
|
dup $C2 u< IF UTF-8-err throw THEN \ malformed character |
$7F and $40 >r |
$7F and $40 >r |
BEGIN dup r@ and WHILE r@ xor |
BEGIN dup r@ and WHILE r@ xor |
6 lshift r> 5 lshift >r >r defers key |
6 lshift r> 5 lshift >r >r defers key |
Line 89 $80 Value max-single-byte
|
Line 94 $80 Value max-single-byte
|
|
|
\ utf-8 stuff for xchars |
\ utf-8 stuff for xchars |
|
|
: +u8/string ( c-addr1 u1 -- c-addr2 u2 ) |
: +u8/string ( xc-addr1 u1 -- xc-addr2 u2 ) |
over dup u8>> swap - /string ; |
over dup u8>> swap - /string ; |
|
: u8\string- ( xcaddr u -- xcaddr u' ) |
: -u8/string ( c-addr1 u1 -- c-addr2 u2 ) |
over + u8<< over - ; |
over dup u8<< swap - /string ; |
|
|
|
: u8@ ( c-addr -- u ) |
: u8@ ( c-addr -- u ) |
u8@+ nip ; |
u8@+ nip ; |
Line 106 $80 Value max-single-byte
|
Line 110 $80 Value max-single-byte
|
>r u8!+ r> r> swap - true |
>r u8!+ r> r> swap - true |
then ; |
then ; |
|
|
: u8addrlen ( u8-addr -- u ) |
: u8addrlen ( u8-addr u -- u ) drop |
\ length of UTF-8 char starting at u8-addr (accesses only u8-addr) |
\ length of UTF-8 char starting at u8-addr (accesses only u8-addr) |
c@ |
c@ |
dup $80 u< if drop 1 exit endif |
dup $80 u< if drop 1 exit endif |
Line 120 $80 Value max-single-byte
|
Line 124 $80 Value max-single-byte
|
|
|
: -u8trailing-garbage ( addr u1 -- addr u2 ) |
: -u8trailing-garbage ( addr u1 -- addr u2 ) |
2dup + dup u8<< ( addr u1 end1 end2 ) |
2dup + dup u8<< ( addr u1 end1 end2 ) |
2dup dup u8addrlen + = if \ last character ok |
2dup dup over over - u8addrlen + = if \ last character ok |
2drop |
2drop |
else |
else |
nip nip over - |
nip nip over - |
then ; |
then ; |
|
|
|
[IFUNDEF] wcwidth |
|
: wc,3 ( n low high -- ) 1+ , , , ; |
|
|
|
Create wc-table \ derived from wcwidth source code, for UCS32 |
|
0 0300 0357 wc,3 |
|
0 035D 036F wc,3 |
|
0 0483 0486 wc,3 |
|
0 0488 0489 wc,3 |
|
0 0591 05A1 wc,3 |
|
0 05A3 05B9 wc,3 |
|
0 05BB 05BD wc,3 |
|
0 05BF 05BF wc,3 |
|
0 05C1 05C2 wc,3 |
|
0 05C4 05C4 wc,3 |
|
0 0600 0603 wc,3 |
|
0 0610 0615 wc,3 |
|
0 064B 0658 wc,3 |
|
0 0670 0670 wc,3 |
|
0 06D6 06E4 wc,3 |
|
0 06E7 06E8 wc,3 |
|
0 06EA 06ED wc,3 |
|
0 070F 070F wc,3 |
|
0 0711 0711 wc,3 |
|
0 0730 074A wc,3 |
|
0 07A6 07B0 wc,3 |
|
0 0901 0902 wc,3 |
|
0 093C 093C wc,3 |
|
0 0941 0948 wc,3 |
|
0 094D 094D wc,3 |
|
0 0951 0954 wc,3 |
|
0 0962 0963 wc,3 |
|
0 0981 0981 wc,3 |
|
0 09BC 09BC wc,3 |
|
0 09C1 09C4 wc,3 |
|
0 09CD 09CD wc,3 |
|
0 09E2 09E3 wc,3 |
|
0 0A01 0A02 wc,3 |
|
0 0A3C 0A3C wc,3 |
|
0 0A41 0A42 wc,3 |
|
0 0A47 0A48 wc,3 |
|
0 0A4B 0A4D wc,3 |
|
0 0A70 0A71 wc,3 |
|
0 0A81 0A82 wc,3 |
|
0 0ABC 0ABC wc,3 |
|
0 0AC1 0AC5 wc,3 |
|
0 0AC7 0AC8 wc,3 |
|
0 0ACD 0ACD wc,3 |
|
0 0AE2 0AE3 wc,3 |
|
0 0B01 0B01 wc,3 |
|
0 0B3C 0B3C wc,3 |
|
0 0B3F 0B3F wc,3 |
|
0 0B41 0B43 wc,3 |
|
0 0B4D 0B4D wc,3 |
|
0 0B56 0B56 wc,3 |
|
0 0B82 0B82 wc,3 |
|
0 0BC0 0BC0 wc,3 |
|
0 0BCD 0BCD wc,3 |
|
0 0C3E 0C40 wc,3 |
|
0 0C46 0C48 wc,3 |
|
0 0C4A 0C4D wc,3 |
|
0 0C55 0C56 wc,3 |
|
0 0CBC 0CBC wc,3 |
|
0 0CBF 0CBF wc,3 |
|
0 0CC6 0CC6 wc,3 |
|
0 0CCC 0CCD wc,3 |
|
0 0D41 0D43 wc,3 |
|
0 0D4D 0D4D wc,3 |
|
0 0DCA 0DCA wc,3 |
|
0 0DD2 0DD4 wc,3 |
|
0 0DD6 0DD6 wc,3 |
|
0 0E31 0E31 wc,3 |
|
0 0E34 0E3A wc,3 |
|
0 0E47 0E4E wc,3 |
|
0 0EB1 0EB1 wc,3 |
|
0 0EB4 0EB9 wc,3 |
|
0 0EBB 0EBC wc,3 |
|
0 0EC8 0ECD wc,3 |
|
0 0F18 0F19 wc,3 |
|
0 0F35 0F35 wc,3 |
|
0 0F37 0F37 wc,3 |
|
0 0F39 0F39 wc,3 |
|
0 0F71 0F7E wc,3 |
|
0 0F80 0F84 wc,3 |
|
0 0F86 0F87 wc,3 |
|
0 0F90 0F97 wc,3 |
|
0 0F99 0FBC wc,3 |
|
0 0FC6 0FC6 wc,3 |
|
0 102D 1030 wc,3 |
|
0 1032 1032 wc,3 |
|
0 1036 1037 wc,3 |
|
0 1039 1039 wc,3 |
|
0 1058 1059 wc,3 |
|
1 0000 1100 wc,3 |
|
2 1100 115f wc,3 |
|
0 1160 11FF wc,3 |
|
0 1712 1714 wc,3 |
|
0 1732 1734 wc,3 |
|
0 1752 1753 wc,3 |
|
0 1772 1773 wc,3 |
|
0 17B4 17B5 wc,3 |
|
0 17B7 17BD wc,3 |
|
0 17C6 17C6 wc,3 |
|
0 17C9 17D3 wc,3 |
|
0 17DD 17DD wc,3 |
|
0 180B 180D wc,3 |
|
0 18A9 18A9 wc,3 |
|
0 1920 1922 wc,3 |
|
0 1927 1928 wc,3 |
|
0 1932 1932 wc,3 |
|
0 1939 193B wc,3 |
|
0 200B 200F wc,3 |
|
0 202A 202E wc,3 |
|
0 2060 2063 wc,3 |
|
0 206A 206F wc,3 |
|
0 20D0 20EA wc,3 |
|
2 2329 232A wc,3 |
|
0 302A 302F wc,3 |
|
2 2E80 303E wc,3 |
|
0 3099 309A wc,3 |
|
2 3040 A4CF wc,3 |
|
2 AC00 D7A3 wc,3 |
|
2 F900 FAFF wc,3 |
|
0 FB1E FB1E wc,3 |
|
0 FE00 FE0F wc,3 |
|
0 FE20 FE23 wc,3 |
|
2 FE30 FE6F wc,3 |
|
0 FEFF FEFF wc,3 |
|
2 FF00 FF60 wc,3 |
|
2 FFE0 FFE6 wc,3 |
|
0 FFF9 FFFB wc,3 |
|
0 1D167 1D169 wc,3 |
|
0 1D173 1D182 wc,3 |
|
0 1D185 1D18B wc,3 |
|
0 1D1AA 1D1AD wc,3 |
|
2 20000 2FFFD wc,3 |
|
2 30000 3FFFD wc,3 |
|
0 E0001 E0001 wc,3 |
|
0 E0020 E007F wc,3 |
|
0 E0100 E01EF wc,3 |
|
here wc-table - Constant #wc-table |
|
|
|
\ inefficient table walk: |
|
|
|
: wcwidth ( xc -- n ) |
|
wc-table #wc-table over + swap ?DO |
|
dup I 2@ within IF I 2 cells + @ UNLOOP EXIT THEN |
|
3 cells +LOOP 1 ; |
|
[THEN] |
|
|
|
: u8width ( xcaddr u -- n ) |
|
0 rot rot over + swap ?DO |
|
I xc@+ swap >r wcwidth + |
|
r> I - +LOOP ; |
|
|
: set-encoding-utf-8 ( -- ) |
: set-encoding-utf-8 ( -- ) |
['] u8emit is xemit |
['] u8emit is xemit |
['] u8key is xkey |
['] u8key is xkey |
['] u8>> is xchar+ |
['] u8>> is xchar+ |
['] u8<< is xchar- |
['] u8<< is xchar- |
|
[ [IFDEF] xstring+ ] |
|
['] u8\string- is xstring- |
|
['] +u8/string is +xstring |
|
[ [THEN] ] |
|
[ [IFDEF] +x/string ] |
|
['] u8\string- is x\string- |
['] +u8/string is +x/string |
['] +u8/string is +x/string |
['] -u8/string is -x/string |
[ [THEN] ] |
['] u8@ is xc@ |
['] u8@ is xc@ |
['] u8!+? is xc!+? |
['] u8!+? is xc!+? |
['] u8@+ is xc@+ |
['] u8@+ is xc@+ |
['] u8len is xc-size |
['] u8len is xc-size |
|
[ [IFDEF] x-width ] |
|
['] u8width is x-width |
|
[ [THEN] ] |
|
[ [IFDEF] x-size ] |
|
['] u8addrlen is x-size |
|
[ [THEN] ] |
['] -u8trailing-garbage is -trailing-garbage |
['] -u8trailing-garbage is -trailing-garbage |
; |
; |
|
|
\ input editor |
: utf-8-cold ( -- ) |
|
s" LC_ALL" getenv 2dup d0= IF 2drop |
[IFUNDEF] #esc 27 Constant #esc [THEN] |
s" LC_CTYPE" getenv 2dup d0= IF 2drop |
|
s" LANG" getenv 2dup d0= IF 2drop |
: save-cursor ( -- ) #esc emit '7 emit ; |
s" C" THEN THEN THEN |
: restore-cursor ( -- ) #esc emit '8 emit ; |
s" UTF-8" search nip nip |
: .rest ( addr pos1 -- addr pos1 ) |
IF set-encoding-utf-8 ELSE set-encoding-fixed-width THEN ; |
restore-cursor 2dup type ; |
|
: .all ( span addr pos1 -- span addr pos1 ) |
environment-wordlist set-current |
restore-cursor >r 2dup swap type r> ; |
: xchar-encoding ( -- addr u ) \ xchar-ext |
|
\G Returns a printable ASCII string that reperesents the encoding, |
: <u8ins> ( max span addr pos1 u8char -- max span addr pos2 ) |
\G and use the preferred MIME name (if any) or the name in |
>r 2over r@ u8len + u< IF rdrop bell EXIT THEN |
\G @url{http://www.iana.org/assignments/character-sets} like |
>string over r@ u8len + swap move 2dup chars + r@ swap u8!+ drop |
\G ``ISO-LATIN-1'' or ``UTF-8'', with the exception of ``ASCII'', where |
r> u8len >r rot r@ chars + -rot r> chars + ; |
\G we prefer the alias ``ASCII''. |
: (u8ins) ( max span addr pos1 u8char -- max span addr pos2 ) |
max-single-byte $80 = IF s" UTF-8" ELSE s" ISO-LATIN-1" THEN ; |
<u8ins> .all .rest ; |
forth definitions |
: u8back ( max span addr pos1 -- max span addr pos2 f ) |
|
dup IF over + u8<< over - 0 max .all .rest |
:noname ( -- ) |
ELSE bell THEN 0 ; |
defers 'cold |
: u8forw ( max span addr pos1 -- max span addr pos2 f ) |
utf-8-cold |
2 pick over <> IF over + u8@+ u8emit over - ELSE bell THEN 0 ; |
; is 'cold |
: (u8del) ( max span addr pos1 -- max span addr pos2 ) |
|
over + dup u8<< tuck - >r over - |
|
>string over r@ + -rot move |
|
rot r> - -rot ; |
|
: ?u8del ( max span addr pos1 -- max span addr pos2 0 ) |
|
dup IF (u8del) .all 2 spaces .rest THEN 0 ; |
|
: <u8del> ( max span addr pos1 -- max span addr pos2 0 ) |
|
2 pick over <> |
|
IF u8forw drop (u8del) .all 2 spaces .rest |
|
ELSE bell THEN 0 ; |
|
: u8eof 2 pick over or 0= IF bye ELSE <u8del> THEN ; |
|
|
|
: u8first-pos ( max span addr pos1 -- max span addr 0 0 ) |
|
drop 0 .all .rest 0 ; |
|
: u8end-pos ( max span addr pos1 -- max span addr span 0 ) |
|
drop over .all 0 ; |
|
|
|
|
|
: u8clear-line ( max span addr pos1 -- max addr ) |
|
drop restore-cursor swap spaces restore-cursor ; |
|
: u8clear-tib ( max span addr pos -- max 0 addr 0 false ) |
|
u8clear-line 0 tuck dup ; |
|
|
|
: (u8enter) ( max span addr pos1 -- max span addr pos2 true ) |
|
>r end^ 2@ hist-setpos |
|
2dup swap history write-line drop ( throw ) \ don't worry about errors |
|
hist-pos 2dup backward^ 2! end^ 2! |
|
r> .all space true ; |
|
|
|
: u8kill-expand ( max span addr pos1 -- max span addr pos2 ) |
|
prefix-found cell+ @ ?dup IF >r |
|
r@ - >string over r@ + -rot move |
|
rot r@ - -rot .all r> spaces .rest THEN ; |
|
|
|
: insert ( string length buffer size -- ) |
|
rot over min >r r@ - ( left over ) |
|
over dup r@ + rot move r> move ; |
|
|
|
: u8tab-expand ( max span addr pos1 -- max span addr pos2 0 ) |
|
key? IF #tab (u8ins) 0 EXIT THEN |
|
u8kill-expand 2dup extract-word dup 0= IF nip EXIT THEN |
|
search-prefix tib-full? |
|
IF 7 emit 2drop prefix-off |
|
ELSE dup >r |
|
2>r >string r@ + 2r> 2swap insert |
|
r@ + rot r> + -rot |
|
THEN |
|
prefix-found @ IF bl (u8ins) ELSE .all .rest THEN 0 ; |
|
|
|
: utf-8-io ( -- ) |
|
['] u8forw ctrl F bindkey |
|
['] u8back ctrl B bindkey |
|
['] ?u8del ctrl H bindkey |
|
['] u8eof ctrl D bindkey |
|
['] <u8del> ctrl X bindkey |
|
['] u8clear-tib ctrl K bindkey |
|
['] u8first-pos ctrl A bindkey |
|
['] u8end-pos ctrl E bindkey |
|
['] (u8enter) #lf bindkey |
|
['] (u8enter) #cr bindkey |
|
['] u8tab-expand #tab bindkey |
|
['] (u8ins) IS insert-char |
|
['] kill-prefix IS everychar |
|
['] save-cursor IS everyline |
|
['] u8key IS key |
|
['] u8emit IS emit ; |
|
|
|
|
utf-8-cold |