| 1 |
############################################################################ |
| 2 |
# |
| 3 |
# Name: ibreader.icn |
| 4 |
# |
| 5 |
# Title: reader for Ibpag2 source files |
| 6 |
# |
| 7 |
# Author: Richard L. Goerwitz |
| 8 |
# |
| 9 |
# Version: 1.29 |
| 10 |
# |
| 11 |
############################################################################ |
| 12 |
# |
| 13 |
# This file contains a collection of procedures that 1) read in an |
| 14 |
# Ibpag2 source file, 2) output token defines, 3) emit action code, |
| 15 |
# and finally 4) pass a start symbol, list of productions, and token |
| 16 |
# table back to the calling procedure. Described formally: |
| 17 |
# |
| 18 |
# ibreader: file x file x string -> ib_grammar record |
| 19 |
# (in, out, module) -> grammar |
| 20 |
# |
| 21 |
# In is the input stream; out is the output stream; module is an |
| 22 |
# optional string that distinguishes this grammar from others that |
| 23 |
# might also be running simultaneously. Grammar is an ib_grammar |
| 24 |
# record containing the start symbol in its first field and the |
| 25 |
# production list in its second. Its third field contains a table |
| 26 |
# used to map integers to actual token names or character literals, |
| 27 |
# i.e. its keys are things like -1, 0, etc. and its values are things |
| 28 |
# like "error," "EOF," etc. |
| 29 |
# |
| 30 |
# Note that if a module argument is supplied to ibreader(), one must |
| 31 |
# also be supplied to ibwriter(). See ibwriter.icn. |
| 32 |
# |
| 33 |
# The format of the input file is highly reminiscent of YACC. It |
| 34 |
# consists of three basic sections, the first two of which are |
| 35 |
# followed by %%. See the main documentation to Ibpag2 for |
| 36 |
# specifics. Major differences between Ibpag2 and YACC input format |
| 37 |
# include: |
| 38 |
# |
| 39 |
# 1) "$$ = x" constructs are replaced by "return x" (e.g. "$$ = |
| 40 |
# $1 + $3" -> "return $1 + $3") |
| 41 |
# |
| 42 |
# 2) all variables within a given action are, by default, local |
| 43 |
# to that action; i.e. they cannot be accessed by other |
| 44 |
# actions unless you declare them global elsewhere (e.g. in |
| 45 |
# the pass-through part of the declarations section %{ ... %}) |
| 46 |
# |
| 47 |
# 3) the %union declaration is not needed by Ibpag |
| 48 |
# |
| 49 |
# 4) tokens and symbols are separated from each other by a comma |
| 50 |
# (e.g. %token '+', '-' and S : NP, VP) |
| 51 |
# |
| 52 |
# 5) epsilon is indicated by the keyword "epsilon" (e.g. REL : |
| 53 |
# epsilon) |
| 54 |
# |
| 55 |
# 6) both epsilon and error *may* be declared as %tokens for |
| 56 |
# reasons of precedence, although they retain hard-coded |
| 57 |
# internal values (-2 and -1, respectively) |
| 58 |
# |
| 59 |
# 7) all actions must follow the last RHS symbol of the rule they |
| 60 |
# apply to (preceded by an optional %prec directive); to |
| 61 |
# achieve S : NP { action1 }, VP { action2 }, insert a dummy |
| 62 |
# rule: S : NP, dummy, VP { action2 }; dummy : epsilon { |
| 63 |
# action1 } ; |
| 64 |
# |
| 65 |
# 8) YYERROR, YYACCEPT, yyclearin, and yyerrok are the same, |
| 66 |
# except they are written IIERROR, IIACCEPT, iiclearin, and |
| 67 |
# iierrok (i.e. "ii" replaces "yy") |
| 68 |
# |
| 69 |
# 9) Ibpag2's input files are tokenized like modified Icon files, |
| 70 |
# and, as a consequence, Icon's reserved words must not be |
| 71 |
# used as symbols (e.g. "if : if, then" is no go) |
| 72 |
# |
| 73 |
############################################################################ |
| 74 |
# |
| 75 |
# Links: itokens, escape |
| 76 |
# |
| 77 |
# See also: ibwriter |
| 78 |
# |
| 79 |
############################################################################ |
| 80 |
|
| 81 |
#link itokens, escape |
| 82 |
link escape |
| 83 |
|
| 84 |
record ib_grammar(start, rules, tbl) |
| 85 |
record tokstats(str, no, prec, assoc) |
| 86 |
|
| 87 |
# Declared in itokens.icn: |
| 88 |
# global line_number |
| 89 |
|
| 90 |
# |
| 91 |
# ibreader: file x file x string x string -> ib_grammar record |
| 92 |
# (in, out, module, source_fname) -> grammar |
| 93 |
# |
| 94 |
# Where in is an input stream, out is an output stream, module is |
| 95 |
# some string uniquely identifying this module (optional), and |
| 96 |
# where grammar is an ib_grammar record containing the start |
| 97 |
# symbol in its first field and a list of production records in |
| 98 |
# its second. Source_fname is the string name of Ibpag2's input |
| 99 |
# grammar file. Defaults to "source file." |
| 100 |
# |
| 101 |
procedure ibreader(in, out, module, source_fname) |
| 102 |
|
| 103 |
local tmp, grammar, toktbl, next_token, next_token_no_nl, |
| 104 |
token, LHS, t |
| 105 |
|
| 106 |
/source_fname := "source file" |
| 107 |
grammar := ib_grammar(&null, list(), table()) |
| 108 |
toktbl := table() |
| 109 |
next_token := create itokens(in, 1) |
| 110 |
next_token_no_nl := create 1(tmp := |@next_token, \tmp.sym) |
| 111 |
token := @next_token_no_nl | iohno(4) |
| 112 |
|
| 113 |
# Do the %{ $} and %token stuff, i.e. everything up to %% |
| 114 |
# (NEWSECT). |
| 115 |
# |
| 116 |
until token.sym == "NEWSECT" do { |
| 117 |
case token.sym of { |
| 118 |
default : { |
| 119 |
iohno(48, "token "||image(token.str) ||"; line "|| line_number) |
| 120 |
} |
| 121 |
"SEMICOL" : { |
| 122 |
# Skip semicolon. Get another token while we're at it. |
| 123 |
token := @next_token_no_nl | iohno(47, "line "||line_number) |
| 124 |
} |
| 125 |
"BEGGLOB" : { |
| 126 |
write(out, "\n$line ", line_number, " ", image(source_fname)) |
| 127 |
# Copy token values to out until we reach "%}" (ENDGLOB). |
| 128 |
(token := copy_icon_stuff(next_token, out)).sym == "ENDGLOB" |
| 129 |
token := @next_token_no_nl |
| 130 |
} |
| 131 |
"MOD" : { |
| 132 |
(token := @next_token_no_nl).sym == "IDENT" | |
| 133 |
iohno(30, "line " || line_number) |
| 134 |
# |
| 135 |
# Read in token declarations, set associativity and |
| 136 |
# precedences, and enter the tokens into toktbl. |
| 137 |
# |
| 138 |
token := { |
| 139 |
case token.str of { |
| 140 |
default : iohno(30, "line " || line_number) |
| 141 |
"token" : read_decl(next_token_no_nl, toktbl, &null) |
| 142 |
"right" : read_decl(next_token_no_nl, toktbl, "r") |
| 143 |
"left" : read_decl(next_token_no_nl, toktbl, "l") |
| 144 |
"nonassoc": read_decl(next_token_no_nl, toktbl, "n") |
| 145 |
"union" : iohno(45, "line "|| line_number) |
| 146 |
"start" : { |
| 147 |
(token := @next_token_no_nl).sym == "IDENT" | |
| 148 |
iohno(31, "line " || line_number) |
| 149 |
/grammar.start := token.str | |
| 150 |
iohno(32, "line " || line_number) |
| 151 |
@next_token_no_nl | iohno(4) |
| 152 |
} |
| 153 |
} |
| 154 |
} |
| 155 |
} |
| 156 |
} |
| 157 |
} |
| 158 |
# Skip past %% (NEWSECT) and semicolon (if present). |
| 159 |
token := @next_token_no_nl | iohno(47, "line "|| line_number) |
| 160 |
(token := token | @next_token_no_nl | iohno(4)).sym ~== "SEMICOL" |
| 161 |
token.sym == "NEWSECT" & iohno(47, "line "|| line_number) |
| 162 |
|
| 163 |
# |
| 164 |
# Fetch start symbol if it wasn't defined above via %start; by |
| 165 |
# default the start symbol is the LHS of rule 1. |
| 166 |
# |
| 167 |
/grammar.start := token.str |
| 168 |
|
| 169 |
# Having reached the end of the declarations section, we can now |
| 170 |
# copy out a define for each token number, not counting character |
| 171 |
# literals (which are stored as integers). While we're at it, |
| 172 |
# create a table that maps token numbers back to character |
| 173 |
# literals and strings (for use in later verbose and debugging |
| 174 |
# displays). |
| 175 |
# |
| 176 |
write(out, "\n") |
| 177 |
every t := !toktbl do { |
| 178 |
if type(t.str) == "integer" then |
| 179 |
insert(grammar.tbl, t.no, image(char(t.str))) |
| 180 |
else { |
| 181 |
insert(grammar.tbl, t.no, t.str) |
| 182 |
write(out, "$define ", t.str, "\t", t.no) |
| 183 |
} |
| 184 |
} |
| 185 |
|
| 186 |
# Now, finally, read in rules up until we reach EOF or %% (i.e. |
| 187 |
# NEWSECT). EOF is signaled below by failure of read_RHS(). |
| 188 |
# |
| 189 |
until token.sym == "NEWSECT" do { |
| 190 |
token.sym == "IDENT" | iohno(33, token.str ||" line "|| line_number) |
| 191 |
LHS := token.str |
| 192 |
token := @next_token_no_nl | iohno(4) |
| 193 |
token.sym == "COLON" | iohno(34, token.str ||" line "|| line_number) |
| 194 |
# |
| 195 |
# Read in RHS, then the action (if any) then the prec (if |
| 196 |
# any). If we see a BAR, then repeat, re-using the same |
| 197 |
# left-hand side symbol. |
| 198 |
# |
| 199 |
while token := |
| 200 |
read_RHS(next_token, next_token_no_nl, out, toktbl, LHS, |
| 201 |
grammar, module, source_fname) | |
| 202 |
# if read_RHS fails, we're at EOF |
| 203 |
break break |
| 204 |
do token.sym == "BAR" | break |
| 205 |
} |
| 206 |
|
| 207 |
# Copy the remainder of the file to out as Icon code. |
| 208 |
write(out, "\n$line ", line_number, " ", image(source_fname)) |
| 209 |
every copy_icon_stuff(next_token, out, "EOFX") |
| 210 |
|
| 211 |
# Do final setup on the reverse token table. This table will be |
| 212 |
# used later to map integers to their original names in verbose or |
| 213 |
# debugging displays. |
| 214 |
# |
| 215 |
insert(grammar.tbl, 0, "$") |
| 216 |
|
| 217 |
return grammar |
| 218 |
|
| 219 |
end |
| 220 |
|
| 221 |
|
| 222 |
# |
| 223 |
# copy_icon_stuff: coexpression x file x string -> ib_TOK records |
| 224 |
# (next_token, out, except) -> token records |
| 225 |
# |
| 226 |
# Copy Icon code to output stream, also suspending as we go. |
| 227 |
# Insert separators between tokens where needed. Do not output |
| 228 |
# any token whose sym field matches except. The point in |
| 229 |
# suspending tokens as we go is to enable the calling procedure to |
| 230 |
# look for signal tokens that indicate insertion or termination |
| 231 |
# points. |
| 232 |
# |
| 233 |
procedure copy_icon_stuff(next_token, out, except) |
| 234 |
|
| 235 |
local separator, T |
| 236 |
|
| 237 |
separator := "" |
| 238 |
while T := @next_token do { |
| 239 |
if \T.sym then suspend T |
| 240 |
if \T.sym == \except then next |
| 241 |
if any(&digits ++ &letters ++ '_.', \T.str, 1, 2) & \T.sym ~== "DOT" |
| 242 |
then writes(out, separator) |
| 243 |
writes(out, T.str) |
| 244 |
if any(&digits ++ &letters ++ '_.', \T.str, -1, 0) & \T.sym ~== "DOT" |
| 245 |
then separator := " " else separator := "" |
| 246 |
} |
| 247 |
|
| 248 |
# unexpected EOF error |
| 249 |
(except === "EOFX") | iohno(4) |
| 250 |
|
| 251 |
end |
| 252 |
|
| 253 |
|
| 254 |
# |
| 255 |
# read_decl: coexpression x table x string -> ib_TOK |
| 256 |
# (next_token_no_nl, toktbl, assoc) -> token |
| 257 |
# |
| 258 |
# Read in token declarations, assigning them the correct |
| 259 |
# precedence and associativity. Number the tokens for later |
| 260 |
# $define preprocessor directives. When done, return the last |
| 261 |
# token processed. Toktbl is the table that holds the stats for |
| 262 |
# each declared token. |
| 263 |
# |
| 264 |
procedure read_decl(next_token_no_nl, toktbl, assoc) |
| 265 |
|
| 266 |
local token, c |
| 267 |
static token_no, prec |
| 268 |
initial { |
| 269 |
token_no := 256 |
| 270 |
prec := 0 |
| 271 |
} |
| 272 |
|
| 273 |
# All tokens in this list have the same prec and assoc. |
| 274 |
# Precedence is determined by order. Associativity is determined |
| 275 |
# by keyword in the calling procedure, and is passed as arg 3. |
| 276 |
# |
| 277 |
prec +:= 1 |
| 278 |
assoc === ("n"|"r"|"l"|&null) | iohno(5, image(assoc)) |
| 279 |
|
| 280 |
# As long as we find commas and token names, keep on adding tokens |
| 281 |
# to the token table. Return the unused token when done. If we |
| 282 |
# reach EOF, there's been an error. |
| 283 |
# |
| 284 |
repeat { |
| 285 |
token := @next_token_no_nl | iohno(4) |
| 286 |
case token.sym of { |
| 287 |
default : iohno(31, token.str ||" line "|| line_number) |
| 288 |
"CSETLIT" | "STRING": { |
| 289 |
# Enter character literals as integers. |
| 290 |
*escape(token.str[2:-1]) = 1 | iohno(49, token.str) |
| 291 |
c := ord(escape(token.str[2:-1])) |
| 292 |
toktbl[c] := tokstats(c, c, prec, assoc) |
| 293 |
} |
| 294 |
"IDENT" : { |
| 295 |
case token.str of { |
| 296 |
"error" : |
| 297 |
toktbl[token.str] := tokstats("error", -1, prec, assoc) |
| 298 |
"epsilon": |
| 299 |
toktbl[token.str] := tokstats("epsilon",-2,prec, assoc) |
| 300 |
default : { |
| 301 |
# Enter TOKENs as string-keyed records in toktbl. |
| 302 |
token_no +:= 1 |
| 303 |
toktbl[token.str] := |
| 304 |
tokstats(token.str, token_no, prec, assoc) |
| 305 |
} |
| 306 |
} |
| 307 |
} |
| 308 |
} |
| 309 |
# As long as we're seeing commas, go back for more tokens. |
| 310 |
token := @next_token_no_nl | iohno(4) |
| 311 |
token.sym == "COMMA" | break |
| 312 |
} |
| 313 |
|
| 314 |
# Skip past semicolon, if present (as set up now, it shouldn't be). |
| 315 |
(token := token | @next_token_no_nl | iohno(4)).sym ~== "SEMICOL" |
| 316 |
return token |
| 317 |
|
| 318 |
end |
| 319 |
|
| 320 |
|
| 321 |
# |
| 322 |
# read_RHS: coexpression x coexpression x file x table x |
| 323 |
# string x ib_grammar record x string x string -> token |
| 324 |
# |
| 325 |
# Read_RHS goes through the RHS of rule definitions, inserting the |
| 326 |
# resulting productions into a master rule list. At the same |
| 327 |
# time, it outputs the actions corresponding to those productions |
| 328 |
# as procedures that are given names corresponding to the numbers |
| 329 |
# of the productions. I.e. production 1, if endowed with an { |
| 330 |
# action }, will correspond to procedure _1_. Prec and assoc are |
| 331 |
# automatically set to that of the last RHS nonterminal, but this |
| 332 |
# may be changed explicitly by the %prec keyword, as in YACC. |
| 333 |
# Source_fname is the name of the source grammar file we're pro- |
| 334 |
# cessing (caller will give us some reasonable default if we're |
| 335 |
# reading &input). |
| 336 |
# |
| 337 |
# Fails on EOF. |
| 338 |
# |
| 339 |
procedure read_RHS(next_token, next_token_no_nl, out, toktbl, LHS, |
| 340 |
grammar, module, source_fname) |
| 341 |
|
| 342 |
local token, rule, c |
| 343 |
static rule_no |
| 344 |
initial rule_no := 0 |
| 345 |
|
| 346 |
rule_no +:= 1 |
| 347 |
# LHS RHS POS LOOK no prec assoc |
| 348 |
rule := production(LHS, list(), &null, &null, rule_no, &null, &null) |
| 349 |
put(grammar.rules, rule) |
| 350 |
|
| 351 |
# Read in RHS symbols. |
| 352 |
# |
| 353 |
repeat { |
| 354 |
token := @next_token_no_nl | iohno(4) |
| 355 |
case token.sym of { |
| 356 |
default : |
| 357 |
iohno(35, "token "|| image(token.str)||"; line "|| line_number) |
| 358 |
"CSETLIT" | "STRING": { |
| 359 |
*escape(token.str[2:-1]) = 1 | iohno(49, token.str) |
| 360 |
c := ord(escape(token.str[2:-1])) |
| 361 |
if \toktbl[c] then { |
| 362 |
rule.prec := toktbl[c].prec |
| 363 |
rule.assoc := toktbl[c].assoc |
| 364 |
} |
| 365 |
# literals not declared earlier will get caught here |
| 366 |
else insert(grammar.tbl, c, image(char(c))) |
| 367 |
put(rule.RHS, c) |
| 368 |
} |
| 369 |
"IDENT" : { |
| 370 |
# If it's a terminal (i.e. a declared token), assign |
| 371 |
# this rule its precedence and associativity. If it's |
| 372 |
# not in toktbl, then it's not a declared token.... |
| 373 |
if \toktbl[token.str] then { |
| 374 |
rule.prec := toktbl[token.str].prec |
| 375 |
rule.assoc := toktbl[token.str].assoc |
| 376 |
put(rule.RHS, toktbl[token.str].no) |
| 377 |
if toktbl[token.str].no = -2 then { |
| 378 |
*rule.RHS > 1 & iohno(44, "line ", line_number) |
| 379 |
rule.POS := 2 |
| 380 |
} |
| 381 |
} |
| 382 |
# ...undeclared stuff. Could be a nonterminal. If |
| 383 |
# error and/or epsilon weren't declared as tokens, |
| 384 |
# they will get caught here, too. |
| 385 |
else { |
| 386 |
case token.str of { |
| 387 |
&null : stop("What is going on here?") |
| 388 |
default : put(rule.RHS, token.str) |
| 389 |
"error" : { |
| 390 |
put(rule.RHS, -1) |
| 391 |
insert(grammar.tbl, -1, "error") |
| 392 |
} |
| 393 |
"epsilon" : { |
| 394 |
if *put(rule.RHS, -2) > 1 |
| 395 |
then iohno(44, "line ", line_number) |
| 396 |
else rule.POS := 2 |
| 397 |
insert(grammar.tbl, -2, "epsilon") |
| 398 |
} |
| 399 |
} |
| 400 |
} |
| 401 |
} |
| 402 |
} |
| 403 |
# Comma means: Go back for another RHS symbol. |
| 404 |
token := @next_token_no_nl | fail |
| 405 |
token.sym == "COMMA" | break |
| 406 |
} |
| 407 |
|
| 408 |
# Skip semicolon token, if present. |
| 409 |
(token := token | @next_token_no_nl | fail).sym ~== "SEMICOL" |
| 410 |
|
| 411 |
# Read and set (optional) precedence. |
| 412 |
# |
| 413 |
if token.sym == "MOD" then { |
| 414 |
token := @next_token_no_nl | iohno(4) |
| 415 |
(token.sym == "IDENT" & token.str == "prec") | |
| 416 |
iohno(43, token.str || " line " || line_number) |
| 417 |
token := @next_token_no_nl | iohno(4) |
| 418 |
case token.sym of { |
| 419 |
"CSETLIT" | "STRING" : { |
| 420 |
*escape(token.str[2:-1]) = 1 | iohno(49, token.str) |
| 421 |
c := ord(escape(token.str[2:-1])) & |
| 422 |
rule.prec := toktbl[c].prec & |
| 423 |
rule.assoc := toktbl[c].assoc |
| 424 |
} |
| 425 |
"IDENT" : { |
| 426 |
\toktbl[token.str] | |
| 427 |
iohno(43, token.str || " line " || line_number) |
| 428 |
rule.prec := toktbl[token.str].prec & |
| 429 |
rule.assoc := toktbl[token.str].assoc |
| 430 |
} |
| 431 |
default : 1 = 4 # deliberate failure |
| 432 |
} | iohno(43, "line ", line_number) |
| 433 |
token := @next_token_no_nl | fail |
| 434 |
} |
| 435 |
|
| 436 |
# Skip semicolon token, if present. |
| 437 |
(token := token | @next_token_no_nl | fail).sym ~== "SEMICOL" |
| 438 |
|
| 439 |
# Read in (optional) action. |
| 440 |
# |
| 441 |
if token.sym == "LBRACE" then { |
| 442 |
write_action_as_procedure(next_token, out, rule, |
| 443 |
module, source_fname) |
| 444 |
token := @next_token_no_nl | fail |
| 445 |
} |
| 446 |
|
| 447 |
# Skip semicolon token, if present. |
| 448 |
(token := token | @next_token_no_nl | fail).sym ~== "SEMICOL" |
| 449 |
return token |
| 450 |
|
| 451 |
end |
| 452 |
|
| 453 |
|
| 454 |
# |
| 455 |
# write_action_as_procedure |
| 456 |
# |
| 457 |
procedure write_action_as_procedure(next_token, out, rule, |
| 458 |
module, source_fname) |
| 459 |
|
| 460 |
local argstr, bracelevel, token, i, neg |
| 461 |
|
| 462 |
/module := "" |
| 463 |
argstr := "" |
| 464 |
# |
| 465 |
# Decide the number of arguments based on the length of the RHS of |
| 466 |
# rule. Exception: Epsilon productions are empty, and pop nothing |
| 467 |
# off the stack, so take zero args. |
| 468 |
# |
| 469 |
if rule.RHS[1] ~=== -2 then { |
| 470 |
every argstr ||:= "arg" || (1 to *rule.RHS) || "," |
| 471 |
argstr := trim(argstr, ',') |
| 472 |
} |
| 473 |
write(out, "procedure _", rule.no, "_", module, "(", argstr, ")") |
| 474 |
write(out, "\n$line ", line_number, " ", image(source_fname)) |
| 475 |
|
| 476 |
bracelevel := 1 |
| 477 |
until bracelevel = 0 do { |
| 478 |
every token := copy_icon_stuff(next_token, out, "RHSARG") do { |
| 479 |
case token.sym of { |
| 480 |
default : next |
| 481 |
"LBRACE" : bracelevel +:= 1 |
| 482 |
"RBRACE" : bracelevel -:= 1 |
| 483 |
"RHSARG" : { |
| 484 |
until \ (token := @next_token).sym do |
| 485 |
writes(out, token.str) |
| 486 |
if neg := (token.sym == "MINUS") then |
| 487 |
until \ (token := @next_token).sym do |
| 488 |
writes(out, token.str) |
| 489 |
else neg := &null |
| 490 |
token.sym == "INTLIT" | iohno(37, "$"||token.str) |
| 491 |
if /neg & token.str ~== "0" then { |
| 492 |
token.str <= *rule.RHS | iohno(38, "$"||token.str) |
| 493 |
writes(out, " arg", token.str, " ") |
| 494 |
} else { |
| 495 |
# Code for $0, $-1, etc. |
| 496 |
# |
| 497 |
# Warning! If the name of the stack is changed |
| 498 |
# in iiparse.lib, it has to be changed here, too. |
| 499 |
# |
| 500 |
i := abs(token.str)+1 |
| 501 |
writes(out, " value_stack", module, "[", i, "] ") |
| 502 |
} |
| 503 |
} |
| 504 |
} |
| 505 |
if bracelevel = 0 then { |
| 506 |
write(out, "\nend\n") |
| 507 |
return token |
| 508 |
} |
| 509 |
} |
| 510 |
} |
| 511 |
|
| 512 |
iohno(39, "line "|| line_number) |
| 513 |
|
| 514 |
end |
| 515 |
|