1 ;; BEN-OTF.flt -- Font Layout Table for Bengali OpenType font
3 ;; National Institute of Advanced Industrial Science and Technology (AIST)
4 ;; Registration Number H15PRO112
6 ;; This file is part of the m17n database; a sub-part of the m17n
9 ;; The m17n library is free software; you can redistribute it and/or
10 ;; modify it under the terms of the GNU Lesser General Public License
11 ;; as published by the Free Software Foundation; either version 2.1 of
12 ;; the License, or (at your option) any later version.
14 ;; The m17n library is distributed in the hope that it will be useful,
15 ;; but WITHOUT ANY WARRANTY; without even the implied warranty of
16 ;; MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
17 ;; Lesser General Public License for more details.
19 ;; You should have received a copy of the GNU Lesser General Public
20 ;; License along with the m17n library; if not, write to the Free
21 ;; Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
26 ;;; For Bengali OpenType fonts to draw Bengali script. Tested with
27 ;;; MuktiNarrow.ttf <http://www.nongnu.org/freebangfont/index.html>
29 ;;; LikhanNormal.otf <http:http://www.stat.wisc.edu/~deepayan/Bengali/WebPage/Font/fonts.html>
32 ;; C: consonant (excluding B, Y and R)
33 ;; B: consonant BA (below)
34 ;; Y: consonant YA (post)
35 ;; R: consonant RA (reph, below)
41 ;; o: MATRA (two-part O)
42 ;; u: MATRA (two-part AU)
44 ;; A: vowel modifier (above)
45 ;; a: vowel modifier (post)
46 ;; V: independent vowel
47 ;; N: ZWNJ (ZERO WIDTH NON-JOINER)
48 ;; J: ZWJ (ZERO WIDTH JOINER)
53 (0x0964 0x0965 ?E) ; DANDA, DOUBLE DANDA
54 (0x0980 0x09FF ?E) ; ELSE
55 (0x0981 ?A) ; SIGN CANDRABINDU (above)
56 (0x0982 0x0983 ?a) ; SIGN ANUSWAR, VISARGA (post)
57 (0x0985 0x098C ?V) ; LETTER A .. VOCALIC L
58 (0x098F 0x0990 ?V) ; LETTER E .. AI
59 (0x0993 0x0994 ?V) ; LETTER O .. AU
60 (0x0995 0x09B9 ?C) ; LETTER KA .. HA
61 (0x09AC ?B) ; LETTER BA
62 (0x09AF ?Y) ; LETTER YA
63 (0x09B0 ?R) ; LETTER RA
64 (0x09BC ?n) ; SIGN NUKTA
65 (0x09BE ?p) ; VOWEL SIGN AA (post)
66 (0x09BF ?m) ; VOWEL SIGN I (pre)
67 (0x09C0 ?p) ; VOWEL SIGN II (post)
68 (0x09C1 0x09C4 ?b) ; VOWEL SIGN U, UU, R, RR (below)
69 (0x09C7 0x09C8 ?m) ; VOWEL SIGN E, AI (pre)
70 (0x09CB 0x09CC ?t) ; VOWEL SIGN O, AU (two-part)
71 (0x09CD ?H) ; SIGN VIRAMA (HASANT)
72 (0x09D7 ?p) ; AU LENGTH MARK
73 (0x09DC 0x09DD ?C) ; LETTER RRA, RHA
74 (0x09DF ?C) ; LETTER YYA
75 (0x09E0 0x09E1 ?V) ; LETTER VOCALIC RR, LL
76 (0x09E2 0x09E3 ?b) ; VOWEL SIGN L .. LL (below)
77 (0x09F1 0x09F2 ?C) ; LETTER RR', RR'' (assamese)
78 (0x09FE ?x) ; mark #1 (internal use)
79 (0x09FF ?y) ; mark #2 (internal use)
82 ;; Step 1 : Syllable identification. Recognised syllables are quoted
83 ;; by the virtual character, which is generated by the command "|" and
84 ;; has the category " " (space).
88 ;; Case A-C are for those syllables that end with an explicit vowel
89 ;; mark and/or a vowel modifier. They are divided into three cases
90 ;; for the readability of regular expression. The leading
91 ;; consonant-Hasant repetition is analysed for reordering in the
92 ;; next step. Two-part vowel, if any, is split for
95 ;; Case A : A syllable ending with a vowel modifier.
96 ("(([CRBY]n?H[NJ]?)*([CRBY]n?))([mbp]*)(t)?([Aa])"
104 ;; Case B : A syllable ending with a two-part vowel.
105 ("(([CRBY]n?H[NJ]?)*([CRBY]n?))(t)"
111 ;; Case C : A syllable ending with other vowel. Note that a
112 ;; two-part vowel may be expressed with two vowel marks for
113 ;; backward compatibility.
114 ("(([CRBY]n?H[NJ]?)*([CRBY]n?))([mbp]+)"
120 ;; Case D : Ya-phalaa. Reorder H and Y for the next step.
121 ;; The web page "Unicode FAQ for Indic Scripts and Languages"
122 ;; <http://www.unicode.org/faq/indic.html> says "it should be
123 ;; permissible for the Ya-phalla to be consistently formed by "ZWNJ
132 ;; Case E : No explicit vowel nor modifier. If the syllable ends
133 ;; with a consonant, analyse it for reordering in the next step.
134 ;; Otherwise, just identify the syllable without changing anything.
135 ("([CRBY]n?H[NJ]?)*[CRBY]n?(HN|HJ|H)?"
142 ;; Case F : Syllables that begin with an independent vowel. An
143 ;; optional HYp sequence appears when this syllable represents the
144 ;; sound "a" in English "bat" (see the FAQ above). If it appears,
145 ;; we reorder the H and Y for the next step.
147 < | (1 =) (2 ("HY(p)" 0x09AF 0x09CD (1 =))) (3 =) | >)
152 ;; Set mark #1 (x) at the position where below consonants begin, and
153 ;; mark #2 (y) at the position to which below and above signs will be
157 ;; At least one C and ends with Y.
158 ("(([CRBY]n?H[NJ]?)*Cn?)H(([BR]H)*)Y"
161 (3 = *) ; below consonants
163 0x09AF 0x09CD) ; YA + moved HASANT
164 ;; At least one C and ends with B or R.
165 ("(([CRBY]n?H[NJ]?)*Cn?)H(([BR]H)*[BR])"
168 (3 = *) ; below consonants
169 0x09CD ; moved HASANT
176 ;; Split two-part dependent vowel signs for canonicalisation.
179 ((0x09CB) 0x09C7 0x09BE)
180 ((0x09CC) 0x09C7 0x09D7)))
183 ;; Step 2 : Move Reph and Matra if necessary. From now on, we care
184 ;; for only those syllables that were identified in Step 1.
188 ;; This is the most generic pattern. It follows Case A-C and a
189 ;; part of Case E in Step 1. Now Mark #1 is used to indicate the
190 ;; critical part that require pre-base substitution in the
194 (" (RH)?([^ xy]+)x([^ y]*)y(YH)?(m)?(b)?(p)?(A)?(a)? "
198 ;; Actually, the nukt feature is not necessary for Bengali because
199 ;; all the necessary Nukta forms are precomposed in the Unicode
200 ;; standard. Even if a Nukta consonant is given in the form of
201 ;; the combination of the base consonant and a Nukta sign, we can
202 ;; safely perm the composition here because it does not affect
203 ;; surrounding letters in the syllable. The Akhand ligature
204 ;; operation is also applied here, before applying the half form
205 ;; operation because the Mukti font generates Akhand ligatures
206 ;; directly from the "C H C" sequence, not via the half form.
207 (2 otf:beng=nukt,akhn) ; {Cpre + H} + Cbase
209 (3 otf:beng=blwf) ; {Cbelow + H}
211 (1 otf:beng=rphf) ; [Reph]
213 (4 otf:beng=pstf) ; [Cpost + H]
218 ;; Syllables that begin with an independent vowel (following up
219 ;; Step 1, Case F). If a YH sequence exist, it is changed to the
220 ;; post base form. This type of syllables do not require further
229 ;; Ya-phalaa (following up Step 1, Case D). Remove N and change YH
230 ;; to the post base form. This type of syllables do not require
231 ;; further modification.
238 ;; Syllables that end with H and an optional N or J (following up a
239 ;; part of Step 1, Case E). This type of syllables also require
240 ;; pre-base substitution in the following steps.
244 ;; Only Nukt and Akhn are applied here. See the comment in the
245 ;; topmost sibling for explanation.
246 (1 otf:beng=nukt,akhn)
253 ;; Step 3 : Now only those syllables that contain the virtual
254 ;; character x require pre-base substition. This is the most
255 ;; complicated part in this FLT.
257 ;; If the sequence "C1 H C2" makes ligature L12, L12 replaces the
258 ;; original sequence.
260 ;; To test the availability of such a ligature, we try to generate it
261 ;; using the pre-base substitute feature, then see whether succeeded
262 ;; or not. In the case of failure, the pre-base feature does not
263 ;; change the original sequence.
265 ;; To create a ligature, the "C1 H" part must be first converted into
266 ;; the half form of C1. Creating the half form of a consonant always
269 ;; ligature(half(C1,H),C2)
270 ;; ==> ligature(C1half,C2)
274 ;; If the ligature is not available, the "C1 H" part must be converted
275 ;; into the _Halant_ (not half) form of C1. However, there is no way
276 ;; to reconvert C1half into C1halant nor to revert back to "C1 H".
277 ;; Thus we duplicate the critical part in two different forms so that
278 ;; we can select the appropriate one in the next step. The virtual
279 ;; character x is used to indicate the boundaries.
281 ;; ... C1 H C2 ... ==> ... x C1halant C2 x L12 x ...
283 ;; If the length of the L12 part is one, ligature generation was
284 ;; successful. In this case we wipe out the duplicated C1halant and
285 ;; C2. Otherwise we remove L12.
287 ;; In very few cases (I found only one in the Mukti font), the "C1 H"
288 ;; part need to be converted into C1halant to make a ligature with C2.
289 ;; So when try to generate a ligature form, we apply the GSUB features
290 ;; "half", "haln" and "pres" in this order.
293 ;; C: consonant (excluding B, Y and R)
295 ;; N: ZWNJ (ZERO WIDTH NON-JOINER)
296 ;; J: ZWJ (ZERO WIDTH JOINER)
301 (0x0964 0x0965 ?E) ; DANDA, DOUBLE DANDA
302 (0x0980 0x09FF ?E) ; ELSE
303 (0x09CD ?H) ; SIGN VIRAMA (HASANT)
304 (0x0995 ?K) ; LETTER KA
305 (0x09B7 ?S) ; LETTER SSA
306 (0x09A8 ?M) ; LETTER NA
307 (0x09AE ?M) ; LETTER MA
308 (0x09FE ?x) ; mark #1 (internal use)
315 ;; One pre-base and base.
317 (" ([^x ]*)x((.H)([^NJ]))(H)?x([^ ]*) "
321 (3 otf:beng=haln) ; C1halant
324 (2 otf:beng=half,haln,pres) ; ligature result
330 ;; One pre-base with ZWJ. According to the Unicode FAQ, the half
331 ;; form is forced in this case. So we fake as if ligature
332 ;; generation was failed.
333 (" ([^x ]*)x(.H)J(.)?x([^ ]*) "
337 (2 otf:beng=half) ; C1half
340 0x09FD ; pseudo result
341 0x09FD ; pseudo result
346 ;; One pre-base possibly with ZWNJ. Similar to above.
347 (" ([^x ]*)x(.H)N?(.)?x([^ ]*) "
351 (2 otf:beng=haln) ; C1halant
354 0x09FD ; pseudo result
355 0x09FD ; pseudo result
360 ;; Standalone base. There is nothing more to do.
361 (" ([^x ]*)x(.)x([^ ]*) "
368 ;; KA-SSA-NA and KA-SSA-MA are the only pre-base ligatures that
369 ;; consist of three consonants.
371 (" ([^x ]*)x((KH)(SH)(M))(H)?x([^ ]*) "
375 (3 otf:beng=haln) ; KAhalant
376 (4 otf:beng=haln) ; SSAhalant
379 (2 otf:beng=half,haln,pres) ; ligature result
385 ;; Two or more pre-bases plus base. Give up. Convert all
386 ;; pre-bases into halant form.
388 (" ([^x ]*)x(([^x]H[JN]?)+)([^x])?x([^ ]*) "
392 (2 force-haln) ; halant forms
395 0x09FD ; pseudo result
396 0x09FD ; pseudo result
404 ;; This is to remove ZWNJ and ZWJ. The half-form-force-effect of ZWJ
405 ;; is ignored. Sorry.
415 ;; Step 4 : Select the appropriate representation. Only those
416 ;; syllables that contain the virtual character x require
421 ;; Only one glyph in the ligature section (between the second and
422 ;; the third x). It means a ligature was successfully generated.
423 ;; C1halant and C2 (between the first and second x) are removed.
424 (" ([^x ]*)x[^x]+x(.)x([^ ]*) "
431 ;; Otherwise halant and base forms are used. The failed ligature
433 (" ([^x ]*)x([^x]+)x[^x]+x([^ ]*) "
440 ;; No need to care the other cases.
444 ;; Step 5 : Fine adjustments. Select appropriate glyph variants and
445 ;; apply GPOS features. Now the syllable boundary marks are removed
446 ;; so that the final step can find word boundaries.
451 (1 otf:beng=blws,abvs,psts,vatu))
457 ;; Step 6 : Word initial substitute. As the syllable boundaries has
458 ;; been eliminated in the previous step, this rule is applied to a run
459 ;; of Bengali glyphs, i.e. word by word. We finally apply the init
460 ;; feature to the word initial gylphs and everything is over.