1 2#***************************************************************************** 3# 4# Copyright (C) 2002-2003, International Business Machines Corporation and others. 5# All Rights Reserved. 6# 7#***************************************************************************** 8# 9# file: rbbirpt.txt 10# ICU Break Iterator Rule Parser State Table 11# 12# This state table is used when reading and parsing a set of RBBI rules 13# The rule parser uses a state machine; the data in this file define the 14# state transitions that occur for each input character. 15# 16# *** This file defines the RBBI rule grammar. This is it. 17# *** The determination of what is accepted is here. 18# 19# This file is processed by a perl script "rbbicst.pl" to produce initialized C arrays 20# that are then built with the rule parser. 21# 22 23# 24# Here is the syntax of the state definitions in this file: 25# 26# 27#StateName: 28# input-char n next-state ^push-state action 29# input-char n next-state ^push-state action 30# | | | | | 31# | | | | |--- action to be performed by state machine 32# | | | | See function RBBIRuleScanner::doParseActions() 33# | | | | 34# | | | |--- Push this named state onto the state stack. 35# | | | Later, when next state is specified as "pop", 36# | | | the pushed state will become the current state. 37# | | | 38# | | |--- Transition to this state if the current input character matches the input 39# | | character or char class in the left hand column. "pop" causes the next 40# | | state to be popped from the state stack. 41# | | 42# | |--- When making the state transition specified on this line, advance to the next 43# | character from the input only if 'n' appears here. 44# | 45# |--- Character or named character classes to test for. If the current character being scanned 46# matches, peform the actions and go to the state specified on this line. 47# The input character is tested sequentally, in the order written. The characters and 48# character classes tested for do not need to be mutually exclusive. The first match wins. 49# 50 51 52 53 54# 55# start state, scan position is at the beginning of the rules file, or in between two rules. 56# 57start: 58 escaped term ^break-rule-end doExprStart 59 white_space n start 60 '$' scan-var-name ^assign-or-rule doExprStart 61 '!' n rev-option 62 ';' n start # ignore empty rules. 63 eof exit 64 default term ^break-rule-end doExprStart 65 66# 67# break-rule-end: Returned from doing a break-rule expression. 68# 69break-rule-end: 70 ';' n start doEndOfRule 71 white_space n break-rule-end 72 default errorDeath doRuleError 73 74 75# 76# ! We've just scanned a '!', indicating either a !!key word flag or a 77# !Reverse rule. 78# 79rev-option: 80 '!' n option-scan1 81 default reverse-rule ^break-rule-end doReverseDir 82 83option-scan1: 84 name_start_char n option-scan2 doOptionStart 85 default errorDeath doRuleError 86 87option-scan2: 88 name_char n option-scan2 89 default option-scan3 doOptionEnd 90 91option-scan3: 92 ';' n start 93 white_space n option-scan3 94 default errorDeath doRuleError 95 96 97reverse-rule: 98 default term ^break-rule-end doExprStart 99 100 101# 102# term. Eat through a single rule character, or a composite thing, which 103# could be a parenthesized expression, a variable name, or a Unicode Set. 104# 105term: 106 escaped n expr-mod doRuleChar 107 white_space n term 108 rule_char n expr-mod doRuleChar 109 '[' scan-unicode-set ^expr-mod 110 '(' n term ^expr-mod doLParen 111 '$' scan-var-name ^term-var-ref 112 '.' n expr-mod doDotAny 113 default errorDeath doRuleError 114 115 116 117# 118# term-var-ref We've just finished scanning a reference to a $variable. 119# Check that the variable was defined. 120# The variable name scanning is in common with assignment statements, 121# so the check can't be done there. 122term-var-ref: 123 default expr-mod doCheckVarDef 124 125 126# 127# expr-mod We've just finished scanning a term, now look for the optional 128# trailing '*', '?', '+' 129# 130expr-mod: 131 white_space n expr-mod 132 '*' n expr-cont doUnaryOpStar 133 '+' n expr-cont doUnaryOpPlus 134 '?' n expr-cont doUnaryOpQuestion 135 default expr-cont 136 137 138# 139# expr-cont Expression, continuation. At a point where additional terms are 140# allowed, but not required. 141# 142expr-cont: 143 escaped term doExprCatOperator 144 white_space n expr-cont 145 rule_char term doExprCatOperator 146 '[' term doExprCatOperator 147 '(' term doExprCatOperator 148 '$' term doExprCatOperator 149 '.' term doExprCatOperator 150 '/' look-ahead doExprCatOperator 151 '{' n tag-open doExprCatOperator 152 '|' n term doExprOrOperator 153 ')' n pop doExprRParen 154 default pop doExprFinished 155 156 157# 158# look-ahead Scanning a '/', which identifies a break point, assuming that the 159# remainder of the expression matches. 160# 161# Generate a parse tree as if this was a special kind of input symbol 162# appearing in an otherwise normal concatenation expression. 163# 164look-ahead: 165 '/' n expr-cont-no-slash doSlash 166 default errorDeath 167 168 169# 170# expr-cont-no-slash Expression, continuation. At a point where additional terms are 171# allowed, but not required. Just like 172# expr-cont, above, except that no '/' 173# look-ahead symbol is permitted. 174# 175expr-cont-no-slash: 176 escaped term doExprCatOperator 177 white_space n expr-cont 178 rule_char term doExprCatOperator 179 '[' term doExprCatOperator 180 '(' term doExprCatOperator 181 '$' term doExprCatOperator 182 '.' term doExprCatOperator 183 '|' n term doExprOrOperator 184 ')' n pop doExprRParen 185 default pop doExprFinished 186 187 188# 189# tags scanning a '{', the opening delimiter for a tag that identifies 190# the kind of match. Scan the whole {dddd} tag, where d=digit 191# 192tag-open: 193 white_space n tag-open 194 digit_char tag-value doStartTagValue 195 default errorDeath doTagExpectedError 196 197tag-value: 198 white_space n tag-close 199 '}' tag-close 200 digit_char n tag-value doTagDigit 201 default errorDeath doTagExpectedError 202 203tag-close: 204 white_space n tag-close 205 '}' n expr-cont-no-tag doTagValue 206 default errorDeath doTagExpectedError 207 208 209 210# 211# expr-cont-no-tag Expression, continuation. At a point where additional terms are 212# allowed, but not required. Just like 213# expr-cont, above, except that no "{ddd}" 214# tagging is permitted. 215# 216expr-cont-no-tag: 217 escaped term doExprCatOperator 218 white_space n expr-cont-no-tag 219 rule_char term doExprCatOperator 220 '[' term doExprCatOperator 221 '(' term doExprCatOperator 222 '$' term doExprCatOperator 223 '.' term doExprCatOperator 224 '/' look-ahead doExprCatOperator 225 '|' n term doExprOrOperator 226 ')' n pop doExprRParen 227 default pop doExprFinished 228 229 230 231 232# 233# Variable Name Scanning. 234# 235# The state that branched to here must have pushed a return state 236# to go to after completion of the variable name scanning. 237# 238# The current input character must be the $ that introduces the name. 239# The $ is consummed here rather than in the state that first detected it 240# so that the doStartVariableName action only needs to happen in one 241# place (here), and the other states don't need to worry about it. 242# 243scan-var-name: 244 '$' n scan-var-start doStartVariableName 245 default errorDeath 246 247 248scan-var-start: 249 name_start_char n scan-var-body 250 default errorDeath doVariableNameExpectedErr 251 252scan-var-body: 253 name_char n scan-var-body 254 default pop doEndVariableName 255 256 257 258# 259# scan-unicode-set Unicode Sets are parsed by the the UnicodeSet class. 260# Within the RBBI parser, after finding the first character 261# of a Unicode Set, we just hand the rule input at that 262# point of to the Unicode Set constructor, then pick 263# up parsing after the close of the set. 264# 265# The action for this state invokes the UnicodeSet parser. 266# 267scan-unicode-set: 268 '[' n pop doScanUnicodeSet 269 'p' n pop doScanUnicodeSet 270 'P' n pop doScanUnicodeSet 271 default errorDeath 272 273 274 275 276 277 278 279# 280# assign-or-rule. A $variable was encountered at the start of something, could be 281# either an assignment statement or a rule, depending on whether an '=' 282# follows the variable name. We get to this state when the variable name 283# scanning does a return. 284# 285assign-or-rule: 286 white_space n assign-or-rule 287 '=' n term ^assign-end doStartAssign # variable was target of assignment 288 default term-var-ref ^break-rule-end # variable was a term in a rule 289 290 291 292# 293# assign-end This state is entered when the end of the expression on the 294# right hand side of an assignment is found. We get here via 295# a pop; this state is pushed when the '=' in an assignment is found. 296# 297# The only thing allowed at this point is a ';'. The RHS of an 298# assignment must look like a rule expression, and we come here 299# when what is being scanned no longer looks like an expression. 300# 301assign-end: 302 ';' n start doEndAssign 303 default errorDeath doRuleErrorAssignExpr 304 305 306 307# 308# errorDeath. This state is specified as the next state whenever a syntax error 309# in the source rules is detected. Barring bugs, the state machine will never 310# actually get here, but will stop because of the action associated with the error. 311# But, just in case, this state asks the state machine to exit. 312errorDeath: 313 default n errorDeath doExit 314 315 316