1 ------------------------------------------------------------------------------
3 -- GNAT LIBRARY COMPONENTS --
5 -- G N A T . R E G P A T --
9 -- Copyright (C) 1986 by University of Toronto. --
10 -- Copyright (C) 1999-2006, AdaCore --
12 -- GNAT is free software; you can redistribute it and/or modify it under --
13 -- terms of the GNU General Public License as published by the Free Soft- --
14 -- ware Foundation; either version 2, or (at your option) any later ver- --
15 -- sion. GNAT is distributed in the hope that it will be useful, but WITH- --
16 -- OUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY --
17 -- or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License --
18 -- for more details. You should have received a copy of the GNU General --
19 -- Public License distributed with GNAT; see file COPYING. If not, write --
20 -- to the Free Software Foundation, 51 Franklin Street, Fifth Floor, --
21 -- Boston, MA 02110-1301, USA. --
23 -- As a special exception, if other files instantiate generics from this --
24 -- unit, or you link this unit with other files to produce an executable, --
25 -- this unit does not by itself cause the resulting executable to be --
26 -- covered by the GNU General Public License. This exception does not --
27 -- however invalidate any other reasons why the executable file might be --
28 -- covered by the GNU Public License. --
30 -- GNAT was originally developed by the GNAT team at New York University. --
31 -- Extensive contributions were provided by Ada Core Technologies Inc. --
33 ------------------------------------------------------------------------------
35 -- This is an altered Ada 95 version of the original V8 style regular
36 -- expression library written in C by Henry Spencer. Apart from the
37 -- translation to Ada, the interface has been considerably changed to
38 -- use the Ada String type instead of C-style nul-terminated strings.
40 -- Beware that some of this code is subtly aware of the way operator
41 -- precedence is structured in regular expressions. Serious changes in
42 -- regular-expression syntax might require a total rethink.
44 with System
.IO
; use System
.IO
;
45 with Ada
.Characters
.Handling
; use Ada
.Characters
.Handling
;
46 with Unchecked_Conversion
;
48 package body GNAT
.Regpat
is
50 MAGIC
: constant Character := Character'Val (10#
0234#
);
51 -- The first byte of the regexp internal "program" is actually
52 -- this magic number; the start node begins in the second byte.
54 -- This is used to make sure that a regular expression was correctly
57 ----------------------------
58 -- Implementation details --
59 ----------------------------
61 -- This is essentially a linear encoding of a nondeterministic
62 -- finite-state machine, also known as syntax charts or
63 -- "railroad normal form" in parsing technology.
65 -- Each node is an opcode plus a "next" pointer, possibly plus an
66 -- operand. "Next" pointers of all nodes except BRANCH implement
67 -- concatenation; a "next" pointer with a BRANCH on both ends of it
68 -- is connecting two alternatives.
70 -- The operand of some types of node is a literal string; for others,
71 -- it is a node leading into a sub-FSM. In particular, the operand of
72 -- a BRANCH node is the first node of the branch.
73 -- (NB this is *not* a tree structure: the tail of the branch connects
74 -- to the thing following the set of BRANCHes).
76 -- You can see the exact byte-compiled version by using the Dump
77 -- subprogram. However, here are a few examples:
80 -- 2 : BRANCH (next at 10)
81 -- 5 : EXACT (next at 18) operand=a
82 -- 10 : BRANCH (next at 18)
83 -- 13 : EXACT (next at 18) operand=b
84 -- 18 : EOP (next at 0)
87 -- 2 : CURLYX (next at 26) { 0, 32767}
88 -- 9 : OPEN 1 (next at 13)
89 -- 13 : EXACT (next at 19) operand=ab
90 -- 19 : CLOSE 1 (next at 23)
91 -- 23 : WHILEM (next at 0)
92 -- 26 : NOTHING (next at 29)
93 -- 29 : EOP (next at 0)
99 -- Name Operand? Meaning
101 (EOP
, -- no End of program
102 MINMOD
, -- no Next operator is not greedy
104 -- Classes of characters
106 ANY
, -- no Match any one character except newline
107 SANY
, -- no Match any character, including new line
108 ANYOF
, -- class Match any character in this class
109 EXACT
, -- str Match this string exactly
110 EXACTF
, -- str Match this string (case-folding is one)
111 NOTHING
, -- no Match empty string
112 SPACE
, -- no Match any whitespace character
113 NSPACE
, -- no Match any non-whitespace character
114 DIGIT
, -- no Match any numeric character
115 NDIGIT
, -- no Match any non-numeric character
116 ALNUM
, -- no Match any alphanumeric character
117 NALNUM
, -- no Match any non-alphanumeric character
121 BRANCH
, -- node Match this alternative, or the next
123 -- Simple loops (when the following node is one character in length)
125 STAR
, -- node Match this simple thing 0 or more times
126 PLUS
, -- node Match this simple thing 1 or more times
127 CURLY
, -- 2num node Match this simple thing between n and m times.
131 CURLYX
, -- 2num node Match this complex thing {n,m} times
132 -- The nums are coded on two characters each
134 WHILEM
, -- no Do curly processing and see if rest matches
136 -- Matches after or before a word
138 BOL
, -- no Match "" at beginning of line
139 MBOL
, -- no Same, assuming mutiline (match after \n)
140 SBOL
, -- no Same, assuming single line (don't match at \n)
141 EOL
, -- no Match "" at end of line
142 MEOL
, -- no Same, assuming mutiline (match before \n)
143 SEOL
, -- no Same, assuming single line (don't match at \n)
145 BOUND
, -- no Match "" at any word boundary
146 NBOUND
, -- no Match "" at any word non-boundary
148 -- Parenthesis groups handling
150 REFF
, -- num Match some already matched string, folded
151 OPEN
, -- num Mark this point in input as start of #n
152 CLOSE
); -- num Analogous to OPEN
154 for Opcode
'Size use 8;
159 -- The set of branches constituting a single choice are hooked
160 -- together with their "next" pointers, since precedence prevents
161 -- anything being concatenated to any individual branch. The
162 -- "next" pointer of the last BRANCH in a choice points to the
163 -- thing following the whole choice. This is also where the
164 -- final "next" pointer of each individual branch points; each
165 -- branch starts with the operand node of a BRANCH node.
168 -- '?', and complex '*' and '+', are implemented with CURLYX.
169 -- branches. Simple cases (one character per match) are implemented with
170 -- STAR and PLUS for speed and to minimize recursive plunges.
173 -- ...are numbered at compile time.
176 -- There are in fact two arguments, the first one is the length (minus
177 -- one of the string argument), coded on one character, the second
178 -- argument is the string itself, coded on length + 1 characters.
180 -- A node is one char of opcode followed by two chars of "next" pointer.
181 -- "Next" pointers are stored as two 8-bit pieces, high order first. The
182 -- value is a positive offset from the opcode of the node containing it.
183 -- An operand, if any, simply follows the node. (Note that much of the
184 -- code generation knows about this implicit relationship.)
186 -- Using two bytes for the "next" pointer is vast overkill for most
187 -- things, but allows patterns to get big without disasters.
189 -----------------------
190 -- Character classes --
191 -----------------------
192 -- This is the implementation for character classes ([...]) in the
193 -- syntax for regular expressions. Each character (0..256) has an
194 -- entry into the table. This makes for a very fast matching
197 type Class_Byte
is mod 256;
198 type Character_Class
is array (Class_Byte
range 0 .. 31) of Class_Byte
;
200 type Bit_Conversion_Array
is array (Class_Byte
range 0 .. 7) of Class_Byte
;
201 Bit_Conversion
: constant Bit_Conversion_Array
:=
202 (1, 2, 4, 8, 16, 32, 64, 128);
204 type Std_Class
is (ANYOF_NONE
,
205 ANYOF_ALNUM
, -- Alphanumeric class [a-zA-Z0-9]
207 ANYOF_SPACE
, -- Space class [ \t\n\r\f]
209 ANYOF_DIGIT
, -- Digit class [0-9]
211 ANYOF_ALNUMC
, -- Alphanumeric class [a-zA-Z0-9]
213 ANYOF_ALPHA
, -- Alpha class [a-zA-Z]
215 ANYOF_ASCII
, -- Ascii class (7 bits) 0..127
217 ANYOF_CNTRL
, -- Control class
219 ANYOF_GRAPH
, -- Graphic class
221 ANYOF_LOWER
, -- Lower case class [a-z]
223 ANYOF_PRINT
, -- printable class
227 ANYOF_UPPER
, -- Upper case class [A-Z]
229 ANYOF_XDIGIT
, -- Hexadecimal digit
233 procedure Set_In_Class
234 (Bitmap
: in out Character_Class
;
236 -- Set the entry to True for C in the class Bitmap
238 function Get_From_Class
239 (Bitmap
: Character_Class
;
240 C
: Character) return Boolean;
241 -- Return True if the entry is set for C in the class Bitmap
243 procedure Reset_Class
(Bitmap
: out Character_Class
);
244 -- Clear all the entries in the class Bitmap
246 pragma Inline
(Set_In_Class
);
247 pragma Inline
(Get_From_Class
);
248 pragma Inline
(Reset_Class
);
250 -----------------------
251 -- Local Subprograms --
252 -----------------------
254 function "=" (Left
: Character; Right
: Opcode
) return Boolean;
256 function Is_Alnum
(C
: Character) return Boolean;
257 -- Return True if C is an alphanum character or an underscore ('_')
259 function Is_White_Space
(C
: Character) return Boolean;
260 -- Return True if C is a whitespace character
262 function Is_Printable
(C
: Character) return Boolean;
263 -- Return True if C is a printable character
265 function Operand
(P
: Pointer
) return Pointer
;
266 -- Return a pointer to the first operand of the node at P
268 function String_Length
269 (Program
: Program_Data
;
270 P
: Pointer
) return Program_Size
;
271 -- Return the length of the string argument of the node at P
273 function String_Operand
(P
: Pointer
) return Pointer
;
274 -- Return a pointer to the string argument of the node at P
276 procedure Bitmap_Operand
277 (Program
: Program_Data
;
279 Op
: out Character_Class
);
280 -- Return a pointer to the string argument of the node at P
282 function Get_Next_Offset
283 (Program
: Program_Data
;
284 IP
: Pointer
) return Pointer
;
285 -- Get the offset field of a node. Used by Get_Next
288 (Program
: Program_Data
;
289 IP
: Pointer
) return Pointer
;
290 -- Dig the next instruction pointer out of a node
292 procedure Optimize
(Self
: in out Pattern_Matcher
);
293 -- Optimize a Pattern_Matcher by noting certain special cases
295 function Read_Natural
296 (Program
: Program_Data
;
297 IP
: Pointer
) return Natural;
298 -- Return the 2-byte natural coded at position IP
300 -- All of the subprograms above are tiny and should be inlined
303 pragma Inline
(Is_Alnum
);
304 pragma Inline
(Is_White_Space
);
305 pragma Inline
(Get_Next
);
306 pragma Inline
(Get_Next_Offset
);
307 pragma Inline
(Operand
);
308 pragma Inline
(Read_Natural
);
309 pragma Inline
(String_Length
);
310 pragma Inline
(String_Operand
);
312 type Expression_Flags
is record
313 Has_Width
, -- Known never to match null string
314 Simple
, -- Simple enough to be STAR/PLUS operand
315 SP_Start
: Boolean; -- Starts with * or +
318 Worst_Expression
: constant Expression_Flags
:= (others => False);
325 function "=" (Left
: Character; Right
: Opcode
) return Boolean is
327 return Character'Pos (Left
) = Opcode
'Pos (Right
);
334 procedure Bitmap_Operand
335 (Program
: Program_Data
;
337 Op
: out Character_Class
)
339 function Convert
is new Unchecked_Conversion
340 (Program_Data
, Character_Class
);
343 Op
(0 .. 31) := Convert
(Program
(P
+ 3 .. P
+ 34));
351 (Matcher
: out Pattern_Matcher
;
353 Final_Code_Size
: out Program_Size
;
354 Flags
: Regexp_Flags
:= No_Flags
)
356 -- We can't allocate space until we know how big the compiled form
357 -- will be, but we can't compile it (and thus know how big it is)
358 -- until we've got a place to put the code. So we cheat: we compile
359 -- it twice, once with code generation turned off and size counting
360 -- turned on, and once "for real".
362 -- This also means that we don't allocate space until we are sure
363 -- that the thing really will compile successfully, and we never
364 -- have to move the code and thus invalidate pointers into it.
366 -- Beware that the optimization-preparation code in here knows
367 -- about some of the structure of the compiled regexp.
369 PM
: Pattern_Matcher
renames Matcher
;
370 Program
: Program_Data
renames PM
.Program
;
372 Emit_Code
: constant Boolean := PM
.Size
> 0;
373 Emit_Ptr
: Pointer
:= Program_First
;
375 Parse_Pos
: Natural := Expression
'First; -- Input-scan pointer
376 Parse_End
: constant Natural := Expression
'Last;
378 ----------------------------
379 -- Subprograms for Create --
380 ----------------------------
382 procedure Emit
(B
: Character);
383 -- Output the Character B to the Program. If code-generation is
384 -- disabled, simply increments the program counter.
386 function Emit_Node
(Op
: Opcode
) return Pointer
;
387 -- If code-generation is enabled, Emit_Node outputs the
388 -- opcode Op and reserves space for a pointer to the next node.
389 -- Return value is the location of new opcode, ie old Emit_Ptr.
391 procedure Emit_Natural
(IP
: Pointer
; N
: Natural);
392 -- Split N on two characters at position IP
394 procedure Emit_Class
(Bitmap
: Character_Class
);
395 -- Emits a character class
397 procedure Case_Emit
(C
: Character);
398 -- Emit C, after converting is to lower-case if the regular
399 -- expression is case insensitive.
402 (Parenthesized
: Boolean;
403 Flags
: out Expression_Flags
;
405 -- Parse regular expression, i.e. main body or parenthesized thing
406 -- Caller must absorb opening parenthesis.
408 procedure Parse_Branch
409 (Flags
: out Expression_Flags
;
412 -- Implements the concatenation operator and handles '|'
413 -- First should be true if this is the first item of the alternative.
415 procedure Parse_Piece
416 (Expr_Flags
: out Expression_Flags
;
418 -- Parse something followed by possible [*+?]
421 (Expr_Flags
: out Expression_Flags
;
423 -- Parse_Atom is the lowest level parse procedure.
424 -- Optimization: gobbles an entire sequence of ordinary characters
425 -- so that it can turn them into a single node, which is smaller to
426 -- store and faster to run. Backslashed characters are exceptions,
427 -- each becoming a separate node; the code is simpler that way and
428 -- it's not worth fixing.
430 procedure Insert_Operator
433 Greedy
: Boolean := True);
434 -- Insert_Operator inserts an operator in front of an
435 -- already-emitted operand and relocates the operand.
436 -- This applies to PLUS and STAR.
437 -- If Minmod is True, then the operator is non-greedy.
439 procedure Insert_Curly_Operator
444 Greedy
: Boolean := True);
445 -- Insert an operator for CURLY ({Min}, {Min,} or {Min,Max}).
446 -- If Minmod is True, then the operator is non-greedy.
448 procedure Link_Tail
(P
, Val
: Pointer
);
449 -- Link_Tail sets the next-pointer at the end of a node chain
451 procedure Link_Operand_Tail
(P
, Val
: Pointer
);
452 -- Link_Tail on operand of first argument; nop if operandless
454 function Next_Instruction
(P
: Pointer
) return Pointer
;
455 -- Dig the "next" pointer out of a node
457 procedure Fail
(M
: String);
458 pragma No_Return
(Fail
);
459 -- Fail with a diagnostic message, if possible
461 function Is_Curly_Operator
(IP
: Natural) return Boolean;
462 -- Return True if IP is looking at a '{' that is the beginning
463 -- of a curly operator, ie it matches {\d+,?\d*}
465 function Is_Mult
(IP
: Natural) return Boolean;
466 -- Return True if C is a regexp multiplier: '+', '*' or '?'
468 procedure Get_Curly_Arguments
472 Greedy
: out Boolean);
473 -- Parse the argument list for a curly operator.
474 -- It is assumed that IP is indeed pointing at a valid operator.
475 -- So what is IP and how come IP is not referenced in the body ???
477 procedure Parse_Character_Class
(IP
: out Pointer
);
478 -- Parse a character class.
479 -- The calling subprogram should consume the opening '[' before.
481 procedure Parse_Literal
482 (Expr_Flags
: out Expression_Flags
;
484 -- Parse_Literal encodes a string of characters to be matched exactly
486 function Parse_Posix_Character_Class
return Std_Class
;
487 -- Parse a posic character class, like [:alpha:] or [:^alpha:].
488 -- The called is suppoed to absorbe the opening [.
490 pragma Inline
(Is_Mult
);
491 pragma Inline
(Emit_Natural
);
492 pragma Inline
(Parse_Character_Class
); -- since used only once
498 procedure Case_Emit
(C
: Character) is
500 if (Flags
and Case_Insensitive
) /= 0 then
504 -- Dump current character
514 procedure Emit
(B
: Character) is
517 Program
(Emit_Ptr
) := B
;
520 Emit_Ptr
:= Emit_Ptr
+ 1;
527 procedure Emit_Class
(Bitmap
: Character_Class
) is
528 subtype Program31
is Program_Data
(0 .. 31);
530 function Convert
is new Unchecked_Conversion
531 (Character_Class
, Program31
);
535 Program
(Emit_Ptr
.. Emit_Ptr
+ 31) := Convert
(Bitmap
);
538 Emit_Ptr
:= Emit_Ptr
+ 32;
545 procedure Emit_Natural
(IP
: Pointer
; N
: Natural) is
548 Program
(IP
+ 1) := Character'Val (N
/ 256);
549 Program
(IP
) := Character'Val (N
mod 256);
557 function Emit_Node
(Op
: Opcode
) return Pointer
is
558 Result
: constant Pointer
:= Emit_Ptr
;
562 Program
(Emit_Ptr
) := Character'Val (Opcode
'Pos (Op
));
563 Program
(Emit_Ptr
+ 1) := ASCII
.NUL
;
564 Program
(Emit_Ptr
+ 2) := ASCII
.NUL
;
567 Emit_Ptr
:= Emit_Ptr
+ 3;
575 procedure Fail
(M
: String) is
577 raise Expression_Error
with M
;
580 -------------------------
581 -- Get_Curly_Arguments --
582 -------------------------
584 procedure Get_Curly_Arguments
588 Greedy
: out Boolean)
590 pragma Unreferenced
(IP
);
592 Save_Pos
: Natural := Parse_Pos
+ 1;
596 Max
:= Max_Curly_Repeat
;
598 while Expression
(Parse_Pos
) /= '}'
599 and then Expression
(Parse_Pos
) /= ','
601 Parse_Pos
:= Parse_Pos
+ 1;
604 Min
:= Natural'Value (Expression
(Save_Pos
.. Parse_Pos
- 1));
606 if Expression
(Parse_Pos
) = ',' then
607 Save_Pos
:= Parse_Pos
+ 1;
608 while Expression
(Parse_Pos
) /= '}' loop
609 Parse_Pos
:= Parse_Pos
+ 1;
612 if Save_Pos
/= Parse_Pos
then
613 Max
:= Natural'Value (Expression
(Save_Pos
.. Parse_Pos
- 1));
620 if Parse_Pos
< Expression
'Last
621 and then Expression
(Parse_Pos
+ 1) = '?'
624 Parse_Pos
:= Parse_Pos
+ 1;
629 end Get_Curly_Arguments
;
631 ---------------------------
632 -- Insert_Curly_Operator --
633 ---------------------------
635 procedure Insert_Curly_Operator
640 Greedy
: Boolean := True)
642 Dest
: constant Pointer
:= Emit_Ptr
;
647 -- If the operand is not greedy, insert an extra operand before it
653 -- Move the operand in the byte-compilation, so that we can insert
654 -- the operator before it.
657 Program
(Operand
+ Size
.. Emit_Ptr
+ Size
) :=
658 Program
(Operand
.. Emit_Ptr
);
661 -- Insert the operator at the position previously occupied by the
667 Old
:= Emit_Node
(MINMOD
);
668 Link_Tail
(Old
, Old
+ 3);
671 Old
:= Emit_Node
(Op
);
672 Emit_Natural
(Old
+ 3, Min
);
673 Emit_Natural
(Old
+ 5, Max
);
675 Emit_Ptr
:= Dest
+ Size
;
676 end Insert_Curly_Operator
;
678 ---------------------
679 -- Insert_Operator --
680 ---------------------
682 procedure Insert_Operator
685 Greedy
: Boolean := True)
687 Dest
: constant Pointer
:= Emit_Ptr
;
692 pragma Warnings
(Off
, Discard
);
695 -- If not greedy, we have to emit another opcode first
701 -- Move the operand in the byte-compilation, so that we can insert
702 -- the operator before it.
705 Program
(Operand
+ Size
.. Emit_Ptr
+ Size
) :=
706 Program
(Operand
.. Emit_Ptr
);
709 -- Insert the operator at the position previously occupied by the
715 Old
:= Emit_Node
(MINMOD
);
716 Link_Tail
(Old
, Old
+ 3);
719 Discard
:= Emit_Node
(Op
);
720 Emit_Ptr
:= Dest
+ Size
;
723 -----------------------
724 -- Is_Curly_Operator --
725 -----------------------
727 function Is_Curly_Operator
(IP
: Natural) return Boolean is
728 Scan
: Natural := IP
;
731 if Expression
(Scan
) /= '{'
732 or else Scan
+ 2 > Expression
'Last
733 or else not Is_Digit
(Expression
(Scan
+ 1))
745 if Scan
> Expression
'Last then
749 exit when not Is_Digit
(Expression
(Scan
));
752 if Expression
(Scan
) = ',' then
756 if Scan
> Expression
'Last then
760 exit when not Is_Digit
(Expression
(Scan
));
764 return Expression
(Scan
) = '}';
765 end Is_Curly_Operator
;
771 function Is_Mult
(IP
: Natural) return Boolean is
772 C
: constant Character := Expression
(IP
);
778 or else (C
= '{' and then Is_Curly_Operator
(IP
));
781 -----------------------
782 -- Link_Operand_Tail --
783 -----------------------
785 procedure Link_Operand_Tail
(P
, Val
: Pointer
) is
787 if Emit_Code
and then Program
(P
) = BRANCH
then
788 Link_Tail
(Operand
(P
), Val
);
790 end Link_Operand_Tail
;
796 procedure Link_Tail
(P
, Val
: Pointer
) is
802 if not Emit_Code
then
810 Temp
:= Next_Instruction
(Scan
);
815 Offset
:= Val
- Scan
;
817 Emit_Natural
(Scan
+ 1, Natural (Offset
));
820 ----------------------
821 -- Next_Instruction --
822 ----------------------
824 function Next_Instruction
(P
: Pointer
) return Pointer
is
828 if not Emit_Code
then
832 Offset
:= Get_Next_Offset
(Program
, P
);
839 end Next_Instruction
;
845 -- Combining parenthesis handling with the base level
846 -- of regular expression is a trifle forced, but the
847 -- need to tie the tails of the branches to what follows
848 -- makes it hard to avoid.
851 (Parenthesized
: Boolean;
852 Flags
: out Expression_Flags
;
855 E
: String renames Expression
;
859 New_Flags
: Expression_Flags
;
860 Have_Branch
: Boolean := False;
863 Flags
:= (Has_Width
=> True, others => False); -- Tentatively
865 -- Make an OPEN node, if parenthesized
867 if Parenthesized
then
868 if Matcher
.Paren_Count
> Max_Paren_Count
then
869 Fail
("too many ()");
872 Par_No
:= Matcher
.Paren_Count
+ 1;
873 Matcher
.Paren_Count
:= Matcher
.Paren_Count
+ 1;
874 IP
:= Emit_Node
(OPEN
);
875 Emit
(Character'Val (Par_No
));
882 -- Pick up the branches, linking them together
884 Parse_Branch
(New_Flags
, True, Br
);
891 if Parse_Pos
<= Parse_End
892 and then E
(Parse_Pos
) = '|'
894 Insert_Operator
(BRANCH
, Br
);
899 Link_Tail
(IP
, Br
); -- OPEN -> first
904 if not New_Flags
.Has_Width
then
905 Flags
.Has_Width
:= False;
908 Flags
.SP_Start
:= Flags
.SP_Start
or New_Flags
.SP_Start
;
910 while Parse_Pos
<= Parse_End
911 and then (E
(Parse_Pos
) = '|')
913 Parse_Pos
:= Parse_Pos
+ 1;
914 Parse_Branch
(New_Flags
, False, Br
);
921 Link_Tail
(IP
, Br
); -- BRANCH -> BRANCH
923 if not New_Flags
.Has_Width
then
924 Flags
.Has_Width
:= False;
927 Flags
.SP_Start
:= Flags
.SP_Start
or New_Flags
.SP_Start
;
930 -- Make a closing node, and hook it on the end
932 if Parenthesized
then
933 Ender
:= Emit_Node
(CLOSE
);
934 Emit
(Character'Val (Par_No
));
936 Ender
:= Emit_Node
(EOP
);
939 Link_Tail
(IP
, Ender
);
943 -- Hook the tails of the branches to the closing node
948 Link_Operand_Tail
(Br
, Ender
);
949 Br
:= Next_Instruction
(Br
);
953 -- Check for proper termination
955 if Parenthesized
then
956 if Parse_Pos
> Parse_End
or else E
(Parse_Pos
) /= ')' then
957 Fail
("unmatched ()");
960 Parse_Pos
:= Parse_Pos
+ 1;
962 elsif Parse_Pos
<= Parse_End
then
963 if E
(Parse_Pos
) = ')' then
964 Fail
("unmatched ()");
966 Fail
("junk on end"); -- "Can't happen"
976 (Expr_Flags
: out Expression_Flags
;
982 -- Tentatively set worst expression case
984 Expr_Flags
:= Worst_Expression
;
986 C
:= Expression
(Parse_Pos
);
987 Parse_Pos
:= Parse_Pos
+ 1;
991 if (Flags
and Multiple_Lines
) /= 0 then
992 IP
:= Emit_Node
(MBOL
);
993 elsif (Flags
and Single_Line
) /= 0 then
994 IP
:= Emit_Node
(SBOL
);
996 IP
:= Emit_Node
(BOL
);
1000 if (Flags
and Multiple_Lines
) /= 0 then
1001 IP
:= Emit_Node
(MEOL
);
1002 elsif (Flags
and Single_Line
) /= 0 then
1003 IP
:= Emit_Node
(SEOL
);
1005 IP
:= Emit_Node
(EOL
);
1009 if (Flags
and Single_Line
) /= 0 then
1010 IP
:= Emit_Node
(SANY
);
1012 IP
:= Emit_Node
(ANY
);
1015 Expr_Flags
.Has_Width
:= True;
1016 Expr_Flags
.Simple
:= True;
1019 Parse_Character_Class
(IP
);
1020 Expr_Flags
.Has_Width
:= True;
1021 Expr_Flags
.Simple
:= True;
1025 New_Flags
: Expression_Flags
;
1028 Parse
(True, New_Flags
, IP
);
1034 Expr_Flags
.Has_Width
:=
1035 Expr_Flags
.Has_Width
or New_Flags
.Has_Width
;
1036 Expr_Flags
.SP_Start
:=
1037 Expr_Flags
.SP_Start
or New_Flags
.SP_Start
;
1040 when '|' | ASCII
.LF |
')' =>
1041 Fail
("internal urp"); -- Supposed to be caught earlier
1043 when '?' |
'+' |
'*' =>
1044 Fail
(C
& " follows nothing");
1047 if Is_Curly_Operator
(Parse_Pos
- 1) then
1048 Fail
(C
& " follows nothing");
1050 Parse_Literal
(Expr_Flags
, IP
);
1054 if Parse_Pos
> Parse_End
then
1055 Fail
("trailing \");
1058 Parse_Pos := Parse_Pos + 1;
1060 case Expression (Parse_Pos - 1) is
1062 IP := Emit_Node (BOUND);
1065 IP := Emit_Node (NBOUND);
1068 IP := Emit_Node (SPACE);
1069 Expr_Flags.Simple := True;
1070 Expr_Flags.Has_Width := True;
1073 IP := Emit_Node (NSPACE);
1074 Expr_Flags.Simple := True;
1075 Expr_Flags.Has_Width := True;
1078 IP := Emit_Node (DIGIT);
1079 Expr_Flags.Simple := True;
1080 Expr_Flags.Has_Width := True;
1083 IP := Emit_Node (NDIGIT);
1084 Expr_Flags.Simple := True;
1085 Expr_Flags.Has_Width := True;
1088 IP := Emit_Node (ALNUM);
1089 Expr_Flags.Simple := True;
1090 Expr_Flags.Has_Width := True;
1093 IP := Emit_Node (NALNUM);
1094 Expr_Flags.Simple := True;
1095 Expr_Flags.Has_Width := True;
1098 IP := Emit_Node (SBOL);
1101 IP := Emit_Node (SEOL);
1104 IP := Emit_Node (REFF);
1107 Save : constant Natural := Parse_Pos - 1;
1110 while Parse_Pos <= Expression'Last
1111 and then Is_Digit (Expression (Parse_Pos))
1113 Parse_Pos := Parse_Pos + 1;
1116 Emit (Character'Val (Natural'Value
1117 (Expression (Save .. Parse_Pos - 1))));
1121 Parse_Pos := Parse_Pos - 1;
1122 Parse_Literal (Expr_Flags, IP);
1126 Parse_Literal (Expr_Flags, IP);
1134 procedure Parse_Branch
1135 (Flags : out Expression_Flags;
1139 E : String renames Expression;
1142 New_Flags : Expression_Flags;
1145 pragma Warnings (Off, Discard);
1148 Flags := Worst_Expression; -- Tentatively
1153 IP := Emit_Node (BRANCH);
1158 while Parse_Pos <= Parse_End
1159 and then E (Parse_Pos) /= ')'
1160 and then E (Parse_Pos) /= ASCII.LF
1161 and then E (Parse_Pos) /= '|'
1163 Parse_Piece (New_Flags, Last);
1170 Flags.Has_Width := Flags.Has_Width or New_Flags.Has_Width;
1172 if Chain = 0 then -- First piece
1173 Flags.SP_Start := Flags.SP_Start or New_Flags.SP_Start;
1175 Link_Tail (Chain, Last);
1181 -- Case where loop ran zero CURLY
1184 Discard := Emit_Node (NOTHING);
1188 ---------------------------
1189 -- Parse_Character_Class --
1190 ---------------------------
1192 procedure Parse_Character_Class (IP : out Pointer) is
1193 Bitmap : Character_Class;
1194 Invert : Boolean := False;
1195 In_Range : Boolean := False;
1196 Named_Class : Std_Class := ANYOF_NONE;
1198 Last_Value : Character := ASCII.Nul;
1201 Reset_Class (Bitmap);
1203 -- Do we have an invert character class ?
1205 if Parse_Pos <= Parse_End
1206 and then Expression (Parse_Pos) = '^'
1209 Parse_Pos := Parse_Pos + 1;
1212 -- First character can be ] or - without closing the class
1214 if Parse_Pos <= Parse_End
1215 and then (Expression (Parse_Pos) = ']'
1216 or else Expression (Parse_Pos) = '-')
1218 Set_In_Class (Bitmap, Expression (Parse_Pos));
1219 Parse_Pos := Parse_Pos + 1;
1222 -- While we don't have the end of the class
1224 while Parse_Pos <= Parse_End
1225 and then Expression (Parse_Pos) /= ']'
1227 Named_Class := ANYOF_NONE;
1228 Value := Expression (Parse_Pos);
1229 Parse_Pos := Parse_Pos + 1;
1231 -- Do we have a Posix character class
1233 Named_Class := Parse_Posix_Character_Class;
1235 elsif Value = '\' then
1236 if Parse_Pos = Parse_End then
1237 Fail ("Trailing
\");
1239 Value
:= Expression
(Parse_Pos
);
1240 Parse_Pos
:= Parse_Pos
+ 1;
1243 when 'w' => Named_Class
:= ANYOF_ALNUM
;
1244 when 'W' => Named_Class
:= ANYOF_NALNUM
;
1245 when 's' => Named_Class
:= ANYOF_SPACE
;
1246 when 'S' => Named_Class
:= ANYOF_NSPACE
;
1247 when 'd' => Named_Class
:= ANYOF_DIGIT
;
1248 when 'D' => Named_Class
:= ANYOF_NDIGIT
;
1249 when 'n' => Value
:= ASCII
.LF
;
1250 when 'r' => Value
:= ASCII
.CR
;
1251 when 't' => Value
:= ASCII
.HT
;
1252 when 'f' => Value
:= ASCII
.FF
;
1253 when 'e' => Value
:= ASCII
.ESC
;
1254 when 'a' => Value
:= ASCII
.BEL
;
1256 -- when 'x' => ??? hexadecimal value
1257 -- when 'c' => ??? control character
1258 -- when '0'..'9' => ??? octal character
1260 when others => null;
1264 -- Do we have a character class?
1266 if Named_Class
/= ANYOF_NONE
then
1268 -- A range like 'a-\d' or 'a-[:digit:] is not a range
1271 Set_In_Class
(Bitmap
, Last_Value
);
1272 Set_In_Class
(Bitmap
, '-');
1279 when ANYOF_NONE
=> null;
1281 when ANYOF_ALNUM | ANYOF_ALNUMC
=>
1282 for Value
in Class_Byte
'Range loop
1283 if Is_Alnum
(Character'Val (Value
)) then
1284 Set_In_Class
(Bitmap
, Character'Val (Value
));
1288 when ANYOF_NALNUM | ANYOF_NALNUMC
=>
1289 for Value
in Class_Byte
'Range loop
1290 if not Is_Alnum
(Character'Val (Value
)) then
1291 Set_In_Class
(Bitmap
, Character'Val (Value
));
1296 for Value
in Class_Byte
'Range loop
1297 if Is_White_Space
(Character'Val (Value
)) then
1298 Set_In_Class
(Bitmap
, Character'Val (Value
));
1302 when ANYOF_NSPACE
=>
1303 for Value
in Class_Byte
'Range loop
1304 if not Is_White_Space
(Character'Val (Value
)) then
1305 Set_In_Class
(Bitmap
, Character'Val (Value
));
1310 for Value
in Class_Byte
'Range loop
1311 if Is_Digit
(Character'Val (Value
)) then
1312 Set_In_Class
(Bitmap
, Character'Val (Value
));
1316 when ANYOF_NDIGIT
=>
1317 for Value
in Class_Byte
'Range loop
1318 if not Is_Digit
(Character'Val (Value
)) then
1319 Set_In_Class
(Bitmap
, Character'Val (Value
));
1324 for Value
in Class_Byte
'Range loop
1325 if Is_Letter
(Character'Val (Value
)) then
1326 Set_In_Class
(Bitmap
, Character'Val (Value
));
1330 when ANYOF_NALPHA
=>
1331 for Value
in Class_Byte
'Range loop
1332 if not Is_Letter
(Character'Val (Value
)) then
1333 Set_In_Class
(Bitmap
, Character'Val (Value
));
1338 for Value
in 0 .. 127 loop
1339 Set_In_Class
(Bitmap
, Character'Val (Value
));
1342 when ANYOF_NASCII
=>
1343 for Value
in 128 .. 255 loop
1344 Set_In_Class
(Bitmap
, Character'Val (Value
));
1348 for Value
in Class_Byte
'Range loop
1349 if Is_Control
(Character'Val (Value
)) then
1350 Set_In_Class
(Bitmap
, Character'Val (Value
));
1354 when ANYOF_NCNTRL
=>
1355 for Value
in Class_Byte
'Range loop
1356 if not Is_Control
(Character'Val (Value
)) then
1357 Set_In_Class
(Bitmap
, Character'Val (Value
));
1362 for Value
in Class_Byte
'Range loop
1363 if Is_Graphic
(Character'Val (Value
)) then
1364 Set_In_Class
(Bitmap
, Character'Val (Value
));
1368 when ANYOF_NGRAPH
=>
1369 for Value
in Class_Byte
'Range loop
1370 if not Is_Graphic
(Character'Val (Value
)) then
1371 Set_In_Class
(Bitmap
, Character'Val (Value
));
1376 for Value
in Class_Byte
'Range loop
1377 if Is_Lower
(Character'Val (Value
)) then
1378 Set_In_Class
(Bitmap
, Character'Val (Value
));
1382 when ANYOF_NLOWER
=>
1383 for Value
in Class_Byte
'Range loop
1384 if not Is_Lower
(Character'Val (Value
)) then
1385 Set_In_Class
(Bitmap
, Character'Val (Value
));
1390 for Value
in Class_Byte
'Range loop
1391 if Is_Printable
(Character'Val (Value
)) then
1392 Set_In_Class
(Bitmap
, Character'Val (Value
));
1396 when ANYOF_NPRINT
=>
1397 for Value
in Class_Byte
'Range loop
1398 if not Is_Printable
(Character'Val (Value
)) then
1399 Set_In_Class
(Bitmap
, Character'Val (Value
));
1404 for Value
in Class_Byte
'Range loop
1405 if Is_Printable
(Character'Val (Value
))
1406 and then not Is_White_Space
(Character'Val (Value
))
1407 and then not Is_Alnum
(Character'Val (Value
))
1409 Set_In_Class
(Bitmap
, Character'Val (Value
));
1413 when ANYOF_NPUNCT
=>
1414 for Value
in Class_Byte
'Range loop
1415 if not Is_Printable
(Character'Val (Value
))
1416 or else Is_White_Space
(Character'Val (Value
))
1417 or else Is_Alnum
(Character'Val (Value
))
1419 Set_In_Class
(Bitmap
, Character'Val (Value
));
1424 for Value
in Class_Byte
'Range loop
1425 if Is_Upper
(Character'Val (Value
)) then
1426 Set_In_Class
(Bitmap
, Character'Val (Value
));
1430 when ANYOF_NUPPER
=>
1431 for Value
in Class_Byte
'Range loop
1432 if not Is_Upper
(Character'Val (Value
)) then
1433 Set_In_Class
(Bitmap
, Character'Val (Value
));
1437 when ANYOF_XDIGIT
=>
1438 for Value
in Class_Byte
'Range loop
1439 if Is_Hexadecimal_Digit
(Character'Val (Value
)) then
1440 Set_In_Class
(Bitmap
, Character'Val (Value
));
1444 when ANYOF_NXDIGIT
=>
1445 for Value
in Class_Byte
'Range loop
1446 if not Is_Hexadecimal_Digit
1447 (Character'Val (Value
))
1449 Set_In_Class
(Bitmap
, Character'Val (Value
));
1455 -- Not a character range
1457 elsif not In_Range
then
1458 Last_Value
:= Value
;
1460 if Expression
(Parse_Pos
) = '-'
1461 and then Parse_Pos
< Parse_End
1462 and then Expression
(Parse_Pos
+ 1) /= ']'
1464 Parse_Pos
:= Parse_Pos
+ 1;
1466 -- Do we have a range like '\d-a' and '[:space:]-a'
1467 -- which is not a real range
1469 if Named_Class
/= ANYOF_NONE
then
1470 Set_In_Class
(Bitmap
, '-');
1476 Set_In_Class
(Bitmap
, Value
);
1480 -- Else in a character range
1483 if Last_Value
> Value
then
1484 Fail
("Invalid Range [" & Last_Value
'Img
1485 & "-" & Value
'Img & "]");
1488 while Last_Value
<= Value
loop
1489 Set_In_Class
(Bitmap
, Last_Value
);
1490 Last_Value
:= Character'Succ (Last_Value
);
1499 -- Optimize case-insensitive ranges (put the upper case or lower
1500 -- case character into the bitmap)
1502 if (Flags
and Case_Insensitive
) /= 0 then
1503 for C
in Character'Range loop
1504 if Get_From_Class
(Bitmap
, C
) then
1505 Set_In_Class
(Bitmap
, To_Lower
(C
));
1506 Set_In_Class
(Bitmap
, To_Upper
(C
));
1511 -- Optimize inverted classes
1514 for J
in Bitmap
'Range loop
1515 Bitmap
(J
) := not Bitmap
(J
);
1519 Parse_Pos
:= Parse_Pos
+ 1;
1523 IP
:= Emit_Node
(ANYOF
);
1524 Emit_Class
(Bitmap
);
1525 end Parse_Character_Class
;
1531 -- This is a bit tricky due to quoted chars and due to
1532 -- the multiplier characters '*', '+', and '?' that
1533 -- take the SINGLE char previous as their operand.
1535 -- On entry, the character at Parse_Pos - 1 is going to go
1536 -- into the string, no matter what it is. It could be
1537 -- following a \ if Parse_Atom was entered from the '\' case.
1539 -- Basic idea is to pick up a good char in C and examine
1540 -- the next char. If Is_Mult (C) then twiddle, if it's a \
1541 -- then frozzle and if it's another magic char then push C and
1542 -- terminate the string. If none of the above, push C on the
1543 -- string and go around again.
1545 -- Start_Pos is used to remember where "the current character"
1546 -- starts in the string, if due to an Is_Mult we need to back
1547 -- up and put the current char in a separate 1-character string.
1548 -- When Start_Pos is 0, C is the only char in the string;
1549 -- this is used in Is_Mult handling, and in setting the SIMPLE
1552 procedure Parse_Literal
1553 (Expr_Flags
: out Expression_Flags
;
1556 Start_Pos
: Natural := 0;
1558 Length_Ptr
: Pointer
;
1560 Has_Special_Operator
: Boolean := False;
1563 Parse_Pos
:= Parse_Pos
- 1; -- Look at current character
1565 if (Flags
and Case_Insensitive
) /= 0 then
1566 IP
:= Emit_Node
(EXACTF
);
1568 IP
:= Emit_Node
(EXACT
);
1571 Length_Ptr
:= Emit_Ptr
;
1572 Emit_Ptr
:= String_Operand
(IP
);
1576 C
:= Expression
(Parse_Pos
); -- Get current character
1579 when '.' |
'[' |
'(' |
')' |
'|' | ASCII
.LF |
'$' |
'^' =>
1581 if Start_Pos
= 0 then
1582 Start_Pos
:= Parse_Pos
;
1583 Emit
(C
); -- First character is always emitted
1585 exit Parse_Loop
; -- Else we are done
1588 when '?' |
'+' |
'*' |
'{' =>
1590 if Start_Pos
= 0 then
1591 Start_Pos
:= Parse_Pos
;
1592 Emit
(C
); -- First character is always emitted
1594 -- Are we looking at an operator, or is this
1595 -- simply a normal character ?
1597 elsif not Is_Mult
(Parse_Pos
) then
1598 Start_Pos
:= Parse_Pos
;
1602 -- We've got something like "abc?d". Mark this as a
1603 -- special case. What we want to emit is a first
1604 -- constant string for "ab", then one for "c" that will
1605 -- ultimately be transformed with a CURLY operator, A
1606 -- special case has to be handled for "a?", since there
1607 -- is no initial string to emit.
1609 Has_Special_Operator
:= True;
1614 Start_Pos
:= Parse_Pos
;
1616 if Parse_Pos
= Parse_End
then
1617 Fail
("Trailing \");
1620 case Expression (Parse_Pos + 1) is
1621 when 'b' | 'B' | 's' | 'S' | 'd' | 'D'
1622 | 'w' | 'W' | '0' .. '9' | 'G' | 'A'
1624 when 'n' => Emit (ASCII.LF);
1625 when 't' => Emit (ASCII.HT);
1626 when 'r' => Emit (ASCII.CR);
1627 when 'f' => Emit (ASCII.FF);
1628 when 'e' => Emit (ASCII.ESC);
1629 when 'a' => Emit (ASCII.BEL);
1630 when others => Emit (Expression (Parse_Pos + 1));
1633 Parse_Pos := Parse_Pos + 1;
1637 Start_Pos := Parse_Pos;
1641 exit Parse_Loop when Emit_Ptr - Length_Ptr = 254;
1643 Parse_Pos := Parse_Pos + 1;
1645 exit Parse_Loop when Parse_Pos > Parse_End;
1646 end loop Parse_Loop;
1648 -- Is the string followed by a '*+?{' operator ? If yes, and if there
1649 -- is an initial string to emit, do it now.
1651 if Has_Special_Operator
1652 and then Emit_Ptr >= Length_Ptr + 3
1654 Emit_Ptr := Emit_Ptr - 1;
1655 Parse_Pos := Start_Pos;
1659 Program (Length_Ptr) := Character'Val (Emit_Ptr - Length_Ptr - 2);
1662 Expr_Flags.Has_Width := True;
1664 -- Slight optimization when there is a single character
1666 if Emit_Ptr = Length_Ptr + 2 then
1667 Expr_Flags.Simple := True;
1675 -- Note that the branching code sequences used for '?' and the
1676 -- general cases of '*' and + are somewhat optimized: they use
1677 -- the same NOTHING node as both the endmarker for their branch
1678 -- list and the body of the last branch. It might seem that
1679 -- this node could be dispensed with entirely, but the endmarker
1680 -- role is not redundant.
1682 procedure Parse_Piece
1683 (Expr_Flags : out Expression_Flags;
1687 New_Flags : Expression_Flags;
1688 Greedy : Boolean := True;
1691 Parse_Atom (New_Flags, IP);
1697 if Parse_Pos > Parse_End
1698 or else not Is_Mult (Parse_Pos)
1700 Expr_Flags := New_Flags;
1704 Op := Expression (Parse_Pos);
1707 Expr_Flags := (SP_Start => True, others => False);
1709 Expr_Flags := (Has_Width => True, others => False);
1712 -- Detect non greedy operators in the easy cases
1715 and then Parse_Pos + 1 <= Parse_End
1716 and then Expression (Parse_Pos + 1) = '?'
1719 Parse_Pos := Parse_Pos + 1;
1722 -- Generate the byte code
1727 if New_Flags.Simple then
1728 Insert_Operator (STAR, IP, Greedy);
1730 Link_Tail (IP, Emit_Node (WHILEM));
1731 Insert_Curly_Operator
1732 (CURLYX, 0, Max_Curly_Repeat, IP, Greedy);
1733 Link_Tail (IP, Emit_Node (NOTHING));
1738 if New_Flags.Simple then
1739 Insert_Operator (PLUS, IP, Greedy);
1741 Link_Tail (IP, Emit_Node (WHILEM));
1742 Insert_Curly_Operator
1743 (CURLYX, 1, Max_Curly_Repeat, IP, Greedy);
1744 Link_Tail (IP, Emit_Node (NOTHING));
1748 if New_Flags.Simple then
1749 Insert_Curly_Operator (CURLY, 0, 1, IP, Greedy);
1751 Link_Tail (IP, Emit_Node (WHILEM));
1752 Insert_Curly_Operator (CURLYX, 0, 1, IP, Greedy);
1753 Link_Tail (IP, Emit_Node (NOTHING));
1761 Get_Curly_Arguments (Parse_Pos, Min, Max, Greedy);
1763 if New_Flags.Simple then
1764 Insert_Curly_Operator (CURLY, Min, Max, IP, Greedy);
1766 Link_Tail (IP, Emit_Node (WHILEM));
1767 Insert_Curly_Operator (CURLYX, Min, Max, IP, Greedy);
1768 Link_Tail (IP, Emit_Node (NOTHING));
1776 Parse_Pos := Parse_Pos + 1;
1778 if Parse_Pos <= Parse_End
1779 and then Is_Mult (Parse_Pos)
1781 Fail ("nested
*+{");
1785 ---------------------------------
1786 -- Parse_Posix_Character_Class --
1787 ---------------------------------
1789 function Parse_Posix_Character_Class return Std_Class is
1790 Invert : Boolean := False;
1791 Class : Std_Class := ANYOF_NONE;
1792 E : String renames Expression;
1794 -- Class names. Note that code assumes that the length of all
1795 -- classes starting with the same letter have the same length.
1797 Alnum : constant String := "alnum
:]";
1798 Alpha : constant String := "alpha
:]";
1799 Ascii_C : constant String := "ascii
:]";
1800 Cntrl : constant String := "cntrl
:]";
1801 Digit : constant String := "digit
:]";
1802 Graph : constant String := "graph
:]";
1803 Lower : constant String := "lower
:]";
1804 Print : constant String := "print
:]";
1805 Punct : constant String := "punct
:]";
1806 Space : constant String := "space
:]";
1807 Upper : constant String := "upper
:]";
1808 Word : constant String := "word
:]";
1809 Xdigit : constant String := "xdigit
:]";
1812 -- Case of character class specified
1814 if Parse_Pos <= Parse_End
1815 and then Expression (Parse_Pos) = ':'
1817 Parse_Pos := Parse_Pos + 1;
1819 -- Do we have something like: [[:^alpha:]]
1821 if Parse_Pos <= Parse_End
1822 and then Expression (Parse_Pos) = '^'
1825 Parse_Pos := Parse_Pos + 1;
1828 -- Check for class names based on first letter
1830 case Expression (Parse_Pos) is
1834 -- All 'a' classes have the same length (Alnum'Length)
1836 if Parse_Pos + Alnum'Length - 1 <= Parse_End then
1838 if E (Parse_Pos .. Parse_Pos + Alnum'Length - 1) =
1842 Class := ANYOF_NALNUMC;
1844 Class := ANYOF_ALNUMC;
1847 Parse_Pos := Parse_Pos + Alnum'Length;
1849 elsif E (Parse_Pos .. Parse_Pos + Alpha'Length - 1) =
1853 Class := ANYOF_NALPHA;
1855 Class := ANYOF_ALPHA;
1858 Parse_Pos := Parse_Pos + Alpha'Length;
1860 elsif E (Parse_Pos .. Parse_Pos + Ascii_C'Length - 1) =
1864 Class := ANYOF_NASCII;
1866 Class := ANYOF_ASCII;
1869 Parse_Pos := Parse_Pos + Ascii_C'Length;
1874 if Parse_Pos + Cntrl'Length - 1 <= Parse_End
1875 and then E (Parse_Pos .. Parse_Pos + Cntrl'Length - 1) =
1879 Class := ANYOF_NCNTRL;
1881 Class := ANYOF_CNTRL;
1884 Parse_Pos := Parse_Pos + Cntrl'Length;
1888 if Parse_Pos + Digit'Length - 1 <= Parse_End
1889 and then E (Parse_Pos .. Parse_Pos + Digit'Length - 1) =
1893 Class := ANYOF_NDIGIT;
1895 Class := ANYOF_DIGIT;
1898 Parse_Pos := Parse_Pos + Digit'Length;
1902 if Parse_Pos + Graph'Length - 1 <= Parse_End
1903 and then E (Parse_Pos .. Parse_Pos + Graph'Length - 1) =
1907 Class := ANYOF_NGRAPH;
1909 Class := ANYOF_GRAPH;
1911 Parse_Pos := Parse_Pos + Graph'Length;
1915 if Parse_Pos + Lower'Length - 1 <= Parse_End
1916 and then E (Parse_Pos .. Parse_Pos + Lower'Length - 1) =
1920 Class := ANYOF_NLOWER;
1922 Class := ANYOF_LOWER;
1924 Parse_Pos := Parse_Pos + Lower'Length;
1929 -- All 'p' classes have the same length
1931 if Parse_Pos + Print'Length - 1 <= Parse_End then
1932 if E (Parse_Pos .. Parse_Pos + Print'Length - 1) =
1936 Class := ANYOF_NPRINT;
1938 Class := ANYOF_PRINT;
1941 Parse_Pos := Parse_Pos + Print'Length;
1943 elsif E (Parse_Pos .. Parse_Pos + Punct'Length - 1) =
1947 Class := ANYOF_NPUNCT;
1949 Class := ANYOF_PUNCT;
1952 Parse_Pos := Parse_Pos + Punct'Length;
1957 if Parse_Pos + Space'Length - 1 <= Parse_End
1958 and then E (Parse_Pos .. Parse_Pos + Space'Length - 1) =
1962 Class := ANYOF_NSPACE;
1964 Class := ANYOF_SPACE;
1967 Parse_Pos := Parse_Pos + Space'Length;
1972 if Parse_Pos + Upper'Length - 1 <= Parse_End
1973 and then E (Parse_Pos .. Parse_Pos + Upper'Length - 1) =
1977 Class := ANYOF_NUPPER;
1979 Class := ANYOF_UPPER;
1981 Parse_Pos := Parse_Pos + Upper'Length;
1986 if Parse_Pos + Word'Length - 1 <= Parse_End
1987 and then E (Parse_Pos .. Parse_Pos + Word'Length - 1) =
1991 Class := ANYOF_NALNUM;
1993 Class := ANYOF_ALNUM;
1995 Parse_Pos := Parse_Pos + Word'Length;
2000 if Parse_Pos + Xdigit'Length - 1 <= Parse_End
2001 and then E (Parse_Pos .. Parse_Pos + Xdigit'Length - 1)
2005 Class := ANYOF_NXDIGIT;
2007 Class := ANYOF_XDIGIT;
2010 Parse_Pos := Parse_Pos + Xdigit'Length;
2014 Fail ("Invalid
character class
");
2017 -- Character class not specified
2024 end Parse_Posix_Character_Class;
2026 Expr_Flags : Expression_Flags;
2029 -- Start of processing for Compile
2033 Parse (False, Expr_Flags, Result);
2036 Fail ("Couldn
't compile expression
");
2039 Final_Code_Size := Emit_Ptr - 1;
2041 -- Do we want to actually compile the expression, or simply get the
2052 (Expression : String;
2053 Flags : Regexp_Flags := No_Flags) return Pattern_Matcher
2055 Size : Program_Size;
2056 Dummy : Pattern_Matcher (0);
2059 Compile (Dummy, Expression, Size, Flags);
2062 Result : Pattern_Matcher (Size);
2064 Compile (Result, Expression, Size, Flags);
2070 (Matcher : out Pattern_Matcher;
2071 Expression : String;
2072 Flags : Regexp_Flags := No_Flags)
2074 Size : Program_Size;
2077 Compile (Matcher, Expression, Size, Flags);
2084 procedure Dump (Self : Pattern_Matcher) is
2086 -- Index : Pointer := Program_First + 1;
2087 -- What is the above line for ???
2090 Program : Program_Data renames Self.Program;
2092 procedure Dump_Until
2095 Indent : Natural := 0);
2096 -- Dump the program until the node Till (not included) is met.
2097 -- Every line is indented with Index spaces at the beginning
2098 -- Dumps till the end if Till is 0.
2104 procedure Dump_Until
2107 Indent : Natural := 0)
2110 Index : Pointer := Start;
2111 Local_Indent : Natural := Indent;
2115 while Index < Till loop
2117 Op := Opcode'Val (Character'Pos ((Self.Program (Index))));
2120 Local_Indent := Local_Indent - 3;
2124 Point : constant String := Pointer'Image (Index);
2127 for J in 1 .. 6 - Point'Length loop
2133 & (1 .. Local_Indent => ' ')
2134 & Opcode'Image (Op));
2137 -- Print the parenthesis number
2139 if Op = OPEN or else Op = CLOSE or else Op = REFF then
2140 Put (Natural'Image (Character'Pos (Program (Index + 3))));
2143 Next := Index + Get_Next_Offset (Program, Index);
2145 if Next = Index then
2146 Put (" (next
at 0)");
2148 Put (" (next
at " & Pointer'Image (Next) & ")");
2153 -- Character class operand
2157 Bitmap : Character_Class;
2158 Last : Character := ASCII.Nul;
2159 Current : Natural := 0;
2161 Current_Char : Character;
2164 Bitmap_Operand (Program, Index, Bitmap);
2167 while Current <= 255 loop
2168 Current_Char := Character'Val (Current);
2170 -- First item in a range
2172 if Get_From_Class (Bitmap, Current_Char) then
2173 Last := Current_Char;
2175 -- Search for the last item in the range
2178 Current := Current + 1;
2179 exit when Current > 255;
2180 Current_Char := Character'Val (Current);
2182 not Get_From_Class (Bitmap, Current_Char);
2192 if Character'Succ (Last) /= Current_Char then
2193 Put ("-" & Character'Pred (Current_Char));
2197 Current := Current + 1;
2202 Index := Index + 3 + Bitmap'Length;
2207 when EXACT | EXACTF =>
2208 Length := String_Length (Program, Index);
2209 Put (" operand
(length
:" & Program_Size'Image (Length + 1)
2211 & String (Program (String_Operand (Index)
2212 .. String_Operand (Index)
2214 Index := String_Operand (Index) + Length + 1;
2221 Dump_Until (Index + 3, Next, Local_Indent + 3);
2227 -- Only one instruction
2229 Dump_Until (Index + 3, Index + 4, Local_Indent + 3);
2232 when CURLY | CURLYX =>
2234 & Natural'Image (Read_Natural (Program, Index + 3))
2236 & Natural'Image (Read_Natural (Program, Index + 5))
2239 Dump_Until (Index + 7, Next, Local_Indent + 3);
2245 Local_Indent := Local_Indent + 3;
2247 when CLOSE | REFF =>
2265 -- Start of processing for Dump
2268 pragma Assert (Self.Program (Program_First) = MAGIC,
2269 "Corrupted Pattern_Matcher
");
2271 Put_Line ("Must start
with (Self
.First
) = "
2272 & Character'Image (Self.First));
2274 if (Self.Flags and Case_Insensitive) /= 0 then
2275 Put_Line (" Case_Insensitive mode
");
2278 if (Self.Flags and Single_Line) /= 0 then
2279 Put_Line (" Single_Line mode
");
2282 if (Self.Flags and Multiple_Lines) /= 0 then
2283 Put_Line (" Multiple_Lines mode
");
2286 Put_Line (" 1 : MAGIC
");
2287 Dump_Until (Program_First + 1, Self.Program'Last + 1);
2290 --------------------
2291 -- Get_From_Class --
2292 --------------------
2294 function Get_From_Class
2295 (Bitmap : Character_Class;
2296 C : Character) return Boolean
2298 Value : constant Class_Byte := Character'Pos (C);
2302 (Bitmap (Value / 8) and Bit_Conversion (Value mod 8)) /= 0;
2309 function Get_Next (Program : Program_Data; IP : Pointer) return Pointer is
2310 Offset : constant Pointer := Get_Next_Offset (Program, IP);
2320 ---------------------
2321 -- Get_Next_Offset --
2322 ---------------------
2324 function Get_Next_Offset
2325 (Program : Program_Data;
2326 IP : Pointer) return Pointer
2329 return Pointer (Read_Natural (Program, IP + 1));
2330 end Get_Next_Offset;
2336 function Is_Alnum (C : Character) return Boolean is
2338 return Is_Alphanumeric (C) or else C = '_';
2345 function Is_Printable (C : Character) return Boolean is
2347 -- Printable if space or graphic character or other whitespace
2348 -- Other white space includes (HT/LF/VT/FF/CR = codes 9-13)
2350 return C in Character'Val (32) .. Character'Val (126)
2351 or else C in ASCII.HT .. ASCII.CR;
2354 --------------------
2355 -- Is_White_Space --
2356 --------------------
2358 function Is_White_Space (C : Character) return Boolean is
2360 -- Note: HT = 9, LF = 10, VT = 11, FF = 12, CR = 13
2362 return C = ' ' or else C in ASCII.HT .. ASCII.CR;
2370 (Self : Pattern_Matcher;
2372 Matches : out Match_Array;
2373 Data_First : Integer := -1;
2374 Data_Last : Positive := Positive'Last)
2376 pragma Assert (Matches'First = 0);
2378 Program : Program_Data renames Self.Program; -- Shorter notation
2380 First_In_Data : constant Integer := Integer'Max (Data_First, Data'First);
2381 Last_In_Data : constant Integer := Integer'Min (Data_Last, Data'Last);
2383 -- Global work variables
2385 Input_Pos : Natural; -- String-input pointer
2386 BOL_Pos : Natural; -- Beginning of input, for ^ check
2387 Matched : Boolean := False; -- Until proven True
2389 Matches_Full : Match_Array (0 .. Natural'Max (Self.Paren_Count,
2391 -- Stores the value of all the parenthesis pairs.
2392 -- We do not use directly Matches, so that we can also use back
2393 -- references (REFF) even if Matches is too small.
2395 type Natural_Array is array (Match_Count range <>) of Natural;
2396 Matches_Tmp : Natural_Array (Matches_Full'Range);
2397 -- Save the opening position of parenthesis
2399 Last_Paren : Natural := 0;
2400 -- Last parenthesis seen
2402 Greedy : Boolean := True;
2403 -- True if the next operator should be greedy
2405 type Current_Curly_Record;
2406 type Current_Curly_Access is access all Current_Curly_Record;
2407 type Current_Curly_Record is record
2408 Paren_Floor : Natural; -- How far back to strip parenthesis data
2409 Cur : Integer; -- How many instances of scan we've matched
2410 Min : Natural; -- Minimal number of scans to match
2411 Max : Natural; -- Maximal number of scans to match
2412 Greedy : Boolean; -- Whether to work our way up or down
2413 Scan : Pointer; -- The thing to match
2414 Next : Pointer; -- What has to match after it
2415 Lastloc : Natural; -- Where we started matching this scan
2416 Old_Cc : Current_Curly_Access; -- Before we started this one
2418 -- Data used to handle the curly operator and the plus and star
2419 -- operators for complex expressions.
2421 Current_Curly : Current_Curly_Access := null;
2422 -- The curly currently being processed
2424 -----------------------
2425 -- Local Subprograms --
2426 -----------------------
2428 function Index (Start : Positive; C : Character) return Natural;
2429 -- Find character C in Data starting at Start and return position
2433 Max : Natural := Natural'Last) return Natural;
2434 -- Repeatedly match something simple, report how many
2435 -- It only matches on things of length 1.
2436 -- Starting from Input_Pos, it matches at most Max CURLY.
2438 function Try (Pos : Positive) return Boolean;
2439 -- Try to match at specific point
2441 function Match (IP : Pointer) return Boolean;
2442 -- This is the main matching routine. Conceptually the strategy
2443 -- is simple: check to see whether the current node matches,
2444 -- call self recursively to see whether the rest matches,
2445 -- and then act accordingly.
2447 -- In practice Match makes some effort to avoid recursion, in
2448 -- particular by going through "ordinary
" nodes (that don't
2449 -- need to know whether the rest of the match failed) by
2450 -- using a loop instead of recursion.
2451 -- Why is the above comment part of the spec rather than body ???
2453 function Match_Whilem (IP : Pointer) return Boolean;
2454 -- Return True if a WHILEM matches
2455 -- How come IP is unreferenced in the body ???
2457 function Recurse_Match (IP : Pointer; From : Natural) return Boolean;
2458 pragma Inline (Recurse_Match);
2459 -- Calls Match recursively. It saves and restores the parenthesis
2460 -- status and location in the input stream correctly, so that
2461 -- backtracking is possible
2463 function Match_Simple_Operator
2467 Greedy : Boolean) return Boolean;
2468 -- Return True it the simple operator (possibly non-greedy) matches
2470 pragma Inline (Index);
2471 pragma Inline (Repeat);
2473 -- These are two complex functions, but used only once
2475 pragma Inline (Match_Whilem);
2476 pragma Inline (Match_Simple_Operator);
2482 function Index (Start : Positive; C : Character) return Natural is
2484 for J in Start .. Last_In_Data loop
2485 if Data (J) = C then
2497 function Recurse_Match (IP : Pointer; From : Natural) return Boolean is
2498 L : constant Natural := Last_Paren;
2500 Tmp_F : constant Match_Array :=
2501 Matches_Full (From + 1 .. Matches_Full'Last);
2503 Start : constant Natural_Array :=
2504 Matches_Tmp (From + 1 .. Matches_Tmp'Last);
2505 Input : constant Natural := Input_Pos;
2513 Matches_Full (Tmp_F'Range) := Tmp_F;
2514 Matches_Tmp (Start'Range) := Start;
2523 function Match (IP : Pointer) return Boolean is
2524 Scan : Pointer := IP;
2531 pragma Assert (Scan /= 0);
2533 -- Determine current opcode and count its usage in debug mode
2535 Op := Opcode'Val (Character'Pos (Program (Scan)));
2537 -- Calculate offset of next instruction.
2538 -- Second character is most significant in Program_Data.
2540 Next := Get_Next (Program, Scan);
2544 return True; -- Success !
2547 if Program (Next) /= BRANCH then
2548 Next := Operand (Scan); -- No choice, avoid recursion
2552 if Recurse_Match (Operand (Scan), 0) then
2556 Scan := Get_Next (Program, Scan);
2557 exit when Scan = 0 or else Program (Scan) /= BRANCH;
2567 exit State_Machine when Input_Pos /= BOL_Pos
2568 and then ((Self.Flags and Multiple_Lines) = 0
2569 or else Data (Input_Pos - 1) /= ASCII.LF);
2572 exit State_Machine when Input_Pos /= BOL_Pos
2573 and then Data (Input_Pos - 1) /= ASCII.LF;
2576 exit State_Machine when Input_Pos /= BOL_Pos;
2579 exit State_Machine when Input_Pos <= Data'Last
2580 and then ((Self.Flags and Multiple_Lines) = 0
2581 or else Data (Input_Pos) /= ASCII.LF);
2584 exit State_Machine when Input_Pos <= Data'Last
2585 and then Data (Input_Pos) /= ASCII.LF;
2588 exit State_Machine when Input_Pos <= Data'Last;
2590 when BOUND | NBOUND =>
2592 -- Was last char in word ?
2595 N : Boolean := False;
2596 Ln : Boolean := False;
2599 if Input_Pos /= First_In_Data then
2600 N := Is_Alnum (Data (Input_Pos - 1));
2603 if Input_Pos > Last_In_Data then
2606 Ln := Is_Alnum (Data (Input_Pos));
2621 exit State_Machine when Input_Pos > Last_In_Data
2622 or else not Is_White_Space (Data (Input_Pos));
2623 Input_Pos := Input_Pos + 1;
2626 exit State_Machine when Input_Pos > Last_In_Data
2627 or else Is_White_Space (Data (Input_Pos));
2628 Input_Pos := Input_Pos + 1;
2631 exit State_Machine when Input_Pos > Last_In_Data
2632 or else not Is_Digit (Data (Input_Pos));
2633 Input_Pos := Input_Pos + 1;
2636 exit State_Machine when Input_Pos > Last_In_Data
2637 or else Is_Digit (Data (Input_Pos));
2638 Input_Pos := Input_Pos + 1;
2641 exit State_Machine when Input_Pos > Last_In_Data
2642 or else not Is_Alnum (Data (Input_Pos));
2643 Input_Pos := Input_Pos + 1;
2646 exit State_Machine when Input_Pos > Last_In_Data
2647 or else Is_Alnum (Data (Input_Pos));
2648 Input_Pos := Input_Pos + 1;
2651 exit State_Machine when Input_Pos > Last_In_Data
2652 or else Data (Input_Pos) = ASCII.LF;
2653 Input_Pos := Input_Pos + 1;
2656 exit State_Machine when Input_Pos > Last_In_Data;
2657 Input_Pos := Input_Pos + 1;
2661 Opnd : Pointer := String_Operand (Scan);
2662 Current : Positive := Input_Pos;
2664 Last : constant Pointer :=
2665 Opnd + String_Length (Program, Scan);
2668 while Opnd <= Last loop
2669 exit State_Machine when Current > Last_In_Data
2670 or else Program (Opnd) /= Data (Current);
2671 Current := Current + 1;
2675 Input_Pos := Current;
2680 Opnd : Pointer := String_Operand (Scan);
2681 Current : Positive := Input_Pos;
2683 Last : constant Pointer :=
2684 Opnd + String_Length (Program, Scan);
2687 while Opnd <= Last loop
2688 exit State_Machine when Current > Last_In_Data
2689 or else Program (Opnd) /= To_Lower (Data (Current));
2690 Current := Current + 1;
2694 Input_Pos := Current;
2699 Bitmap : Character_Class;
2702 Bitmap_Operand (Program, Scan, Bitmap);
2703 exit State_Machine when Input_Pos > Last_In_Data
2704 or else not Get_From_Class (Bitmap, Data (Input_Pos));
2705 Input_Pos := Input_Pos + 1;
2710 No : constant Natural :=
2711 Character'Pos (Program (Operand (Scan)));
2714 Matches_Tmp (No) := Input_Pos;
2719 No : constant Natural :=
2720 Character'Pos (Program (Operand (Scan)));
2723 Matches_Full (No) := (Matches_Tmp (No), Input_Pos - 1);
2725 if Last_Paren < No then
2732 No : constant Natural :=
2733 Character'Pos (Program (Operand (Scan)));
2738 -- If we haven't seen that parenthesis yet
2740 if Last_Paren < No then
2744 Data_Pos := Matches_Full (No).First;
2746 while Data_Pos <= Matches_Full (No).Last loop
2747 if Input_Pos > Last_In_Data
2748 or else Data (Input_Pos) /= Data (Data_Pos)
2753 Input_Pos := Input_Pos + 1;
2754 Data_Pos := Data_Pos + 1;
2761 when STAR | PLUS | CURLY =>
2763 Greed : constant Boolean := Greedy;
2767 return Match_Simple_Operator (Op, Scan, Next, Greed);
2772 -- Looking at something like:
2774 -- 1: CURLYX {n,m} (->4)
2775 -- 2: code for complex thing (->3)
2780 Min : constant Natural :=
2781 Read_Natural (Program, Scan + 3);
2782 Max : constant Natural :=
2783 Read_Natural (Program, Scan + 5);
2784 Cc : aliased Current_Curly_Record;
2786 Has_Match : Boolean;
2789 Cc := (Paren_Floor => Last_Paren,
2797 Old_Cc => Current_Curly);
2798 Current_Curly := Cc'Unchecked_Access;
2800 Has_Match := Match (Next - 3);
2802 -- Start on the WHILEM
2804 Current_Curly := Cc.Old_Cc;
2809 return Match_Whilem (IP);
2813 end loop State_Machine;
2815 -- If we get here, there is no match.
2816 -- For successful matches when EOP is the terminating point.
2821 ---------------------------
2822 -- Match_Simple_Operator --
2823 ---------------------------
2825 function Match_Simple_Operator
2829 Greedy : Boolean) return Boolean
2831 Next_Char : Character := ASCII.Nul;
2832 Next_Char_Known : Boolean := False;
2833 No : Integer; -- Can be negative
2835 Max : Natural := Natural'Last;
2836 Operand_Code : Pointer;
2839 Save : constant Natural := Input_Pos;
2842 -- Lookahead to avoid useless match attempts
2843 -- when we know what character comes next.
2845 if Program (Next) = EXACT then
2846 Next_Char := Program (String_Operand (Next));
2847 Next_Char_Known := True;
2850 -- Find the minimal and maximal values for the operator
2855 Operand_Code := Operand (Scan);
2859 Operand_Code := Operand (Scan);
2862 Min := Read_Natural (Program, Scan + 3);
2863 Max := Read_Natural (Program, Scan + 5);
2864 Operand_Code := Scan + 7;
2867 -- Non greedy operators
2871 -- Test the minimal repetitions
2874 and then Repeat (Operand_Code, Min) < Min
2881 -- Find the place where 'next' could work
2883 if Next_Char_Known then
2884 -- Last position to check
2886 if Max = Natural'Last then
2887 Last_Pos := Last_In_Data;
2889 Last_Pos := Input_Pos + Max;
2891 if Last_Pos > Last_In_Data then
2892 Last_Pos := Last_In_Data;
2896 -- Look for the first possible opportunity
2899 -- Find the next possible position
2901 while Input_Pos <= Last_Pos
2902 and then Data (Input_Pos) /= Next_Char
2904 Input_Pos := Input_Pos + 1;
2907 if Input_Pos > Last_Pos then
2911 -- Check that we still match if we stop
2912 -- at the position we just found.
2915 Num : constant Natural := Input_Pos - Old;
2920 if Repeat (Operand_Code, Num) < Num then
2925 -- Input_Pos now points to the new position
2927 if Match (Get_Next (Program, Scan)) then
2932 Input_Pos := Input_Pos + 1;
2935 -- We know what the next character is
2938 while Max >= Min loop
2940 -- If the next character matches
2942 if Match (Next) then
2946 Input_Pos := Save + Min;
2948 -- Could not or did not match -- move forward
2950 if Repeat (Operand_Code, 1) /= 0 then
2963 No := Repeat (Operand_Code, Max);
2965 -- ??? Perl has some special code here in case the
2966 -- next instruction is of type EOL, since $ and \Z
2967 -- can match before *and* after newline at the end.
2969 -- ??? Perl has some special code here in case (paren)
2972 -- Else, if we don't have any parenthesis
2974 while No >= Min loop
2975 if not Next_Char_Known
2976 or else (Input_Pos <= Last_In_Data
2977 and then Data (Input_Pos) = Next_Char)
2979 if Match (Next) then
2984 -- Could not or did not work, we back up
2987 Input_Pos := Save + No;
2992 end Match_Simple_Operator;
2998 -- This is really hard to understand, because after we match what we
2999 -- are trying to match, we must make sure the rest of the REx is going
3000 -- to match for sure, and to do that we have to go back UP the parse
3001 -- tree by recursing ever deeper. And if it fails, we have to reset
3002 -- our parent's current state that we can try again after backing off.
3004 function Match_Whilem (IP : Pointer) return Boolean is
3005 pragma Unreferenced (IP);
3007 Cc : constant Current_Curly_Access := Current_Curly;
3008 N : constant Natural := Cc.Cur + 1;
3011 Lastloc : constant Natural := Cc.Lastloc;
3012 -- Detection of 0-len
3015 -- If degenerate scan matches "", assume scan done
3017 if Input_Pos = Cc.Lastloc
3018 and then N >= Cc.Min
3020 -- Temporarily restore the old context, and check that we
3021 -- match was comes after CURLYX.
3023 Current_Curly := Cc.Old_Cc;
3025 if Current_Curly /= null then
3026 Ln := Current_Curly.Cur;
3029 if Match (Cc.Next) then
3033 if Current_Curly /= null then
3034 Current_Curly.Cur := Ln;
3037 Current_Curly := Cc;
3041 -- First, just match a string of min scans
3045 Cc.Lastloc := Input_Pos;
3047 if Match (Cc.Scan) then
3052 Cc.Lastloc := Lastloc;
3056 -- Prefer next over scan for minimal matching
3058 if not Cc.Greedy then
3059 Current_Curly := Cc.Old_Cc;
3061 if Current_Curly /= null then
3062 Ln := Current_Curly.Cur;
3065 if Recurse_Match (Cc.Next, Cc.Paren_Floor) then
3069 if Current_Curly /= null then
3070 Current_Curly.Cur := Ln;
3073 Current_Curly := Cc;
3075 -- Maximum greed exceeded ?
3081 -- Try scanning more and see if it helps
3083 Cc.Lastloc := Input_Pos;
3085 if Recurse_Match (Cc.Scan, Cc.Paren_Floor) then
3090 Cc.Lastloc := Lastloc;
3094 -- Prefer scan over next for maximal matching
3096 if N < Cc.Max then -- more greed allowed ?
3098 Cc.Lastloc := Input_Pos;
3100 if Recurse_Match (Cc.Scan, Cc.Paren_Floor) then
3105 -- Failed deeper matches of scan, so see if this one works
3107 Current_Curly := Cc.Old_Cc;
3109 if Current_Curly /= null then
3110 Ln := Current_Curly.Cur;
3113 if Match (Cc.Next) then
3117 if Current_Curly /= null then
3118 Current_Curly.Cur := Ln;
3121 Current_Curly := Cc;
3123 Cc.Lastloc := Lastloc;
3133 Max : Natural := Natural'Last) return Natural
3135 Scan : Natural := Input_Pos;
3137 Op : constant Opcode := Opcode'Val (Character'Pos (Program (IP)));
3140 Is_First : Boolean := True;
3141 Bitmap : Character_Class;
3144 if Max = Natural'Last or else Scan + Max - 1 > Last_In_Data then
3145 Last := Last_In_Data;
3147 Last := Scan + Max - 1;
3153 and then Data (Scan) /= ASCII.LF
3163 -- The string has only one character if Repeat was called
3165 C := Program (String_Operand (IP));
3167 and then C = Data (Scan)
3174 -- The string has only one character if Repeat was called
3176 C := Program (String_Operand (IP));
3178 and then To_Lower (C) = Data (Scan)
3185 Bitmap_Operand (Program, IP, Bitmap);
3190 and then Get_From_Class (Bitmap, Data (Scan))
3197 and then Is_Alnum (Data (Scan))
3204 and then not Is_Alnum (Data (Scan))
3211 and then Is_White_Space (Data (Scan))
3218 and then not Is_White_Space (Data (Scan))
3225 and then Is_Digit (Data (Scan))
3232 and then not Is_Digit (Data (Scan))
3238 raise Program_Error;
3241 Count := Scan - Input_Pos;
3250 function Try (Pos : Positive) return Boolean is
3254 Matches_Full := (others => No_Match);
3256 if Match (Program_First + 1) then
3257 Matches_Full (0) := (Pos, Input_Pos - 1);
3264 -- Start of processing for Match
3267 -- Do we have the regexp Never_Match?
3269 if Self.Size = 0 then
3270 Matches (0) := No_Match;
3274 -- Check validity of program
3277 (Program (Program_First) = MAGIC,
3278 "Corrupted Pattern_Matcher
");
3280 -- If there is a "must appear
" string, look for it
3282 if Self.Must_Have_Length > 0 then
3284 First : constant Character := Program (Self.Must_Have);
3285 Must_First : constant Pointer := Self.Must_Have;
3286 Must_Last : constant Pointer :=
3287 Must_First + Pointer (Self.Must_Have_Length - 1);
3288 Next_Try : Natural := Index (First_In_Data, First);
3292 and then Data (Next_Try .. Next_Try + Self.Must_Have_Length - 1)
3293 = String (Program (Must_First .. Must_Last))
3295 Next_Try := Index (Next_Try + 1, First);
3298 if Next_Try = 0 then
3299 Matches_Full := (others => No_Match);
3300 return; -- Not present
3305 -- Mark beginning of line for ^
3307 BOL_Pos := Data'First;
3309 -- Simplest case first: an anchored match need be tried only once
3311 if Self.Anchored and then (Self.Flags and Multiple_Lines) = 0 then
3312 Matched := Try (First_In_Data);
3314 elsif Self.Anchored then
3316 Next_Try : Natural := First_In_Data;
3318 -- Test the first position in the buffer
3319 Matched := Try (Next_Try);
3321 -- Else only test after newlines
3324 while Next_Try <= Last_In_Data loop
3325 while Next_Try <= Last_In_Data
3326 and then Data (Next_Try) /= ASCII.LF
3328 Next_Try := Next_Try + 1;
3331 Next_Try := Next_Try + 1;
3333 if Next_Try <= Last_In_Data then
3334 Matched := Try (Next_Try);
3341 elsif Self.First /= ASCII.NUL then
3342 -- We know what char it must start with
3345 Next_Try : Natural := Index (First_In_Data, Self.First);
3348 while Next_Try /= 0 loop
3349 Matched := Try (Next_Try);
3351 Next_Try := Index (Next_Try + 1, Self.First);
3356 -- Messy cases: try all locations (including for the empty string)
3358 Matched := Try (First_In_Data);
3361 for S in First_In_Data + 1 .. Last_In_Data loop
3368 -- Matched has its value
3370 for J in Last_Paren + 1 .. Matches'Last loop
3371 Matches_Full (J) := No_Match;
3374 Matches := Matches_Full (Matches'Range);
3383 (Self : Pattern_Matcher;
3385 Data_First : Integer := -1;
3386 Data_Last : Positive := Positive'Last) return Natural
3388 Matches : Match_Array (0 .. 0);
3391 Match (Self, Data, Matches, Data_First, Data_Last);
3392 if Matches (0) = No_Match then
3393 return Data'First - 1;
3395 return Matches (0).First;
3400 (Self : Pattern_Matcher;
3402 Data_First : Integer := -1;
3403 Data_Last : Positive := Positive'Last) return Boolean
3405 Matches : Match_Array (0 .. 0);
3408 Match (Self, Data, Matches, Data_First, Data_Last);
3409 return Matches (0).First >= Data'First;
3413 (Expression : String;
3415 Matches : out Match_Array;
3416 Size : Program_Size := Auto_Size;
3417 Data_First : Integer := -1;
3418 Data_Last : Positive := Positive'Last)
3420 PM : Pattern_Matcher (Size);
3421 Finalize_Size : Program_Size;
3425 Match (Compile (Expression), Data, Matches, Data_First, Data_Last);
3427 Compile (PM, Expression, Finalize_Size);
3428 Match (PM, Data, Matches, Data_First, Data_Last);
3437 (Expression : String;
3439 Size : Program_Size := Auto_Size;
3440 Data_First : Integer := -1;
3441 Data_Last : Positive := Positive'Last) return Natural
3443 PM : Pattern_Matcher (Size);
3444 Final_Size : Program_Size; -- unused
3448 return Match (Compile (Expression), Data, Data_First, Data_Last);
3450 Compile (PM, Expression, Final_Size);
3451 return Match (PM, Data, Data_First, Data_Last);
3460 (Expression : String;
3462 Size : Program_Size := Auto_Size;
3463 Data_First : Integer := -1;
3464 Data_Last : Positive := Positive'Last) return Boolean
3466 Matches : Match_Array (0 .. 0);
3467 PM : Pattern_Matcher (Size);
3468 Final_Size : Program_Size; -- unused
3472 Match (Compile (Expression), Data, Matches, Data_First, Data_Last);
3474 Compile (PM, Expression, Final_Size);
3475 Match (PM, Data, Matches, Data_First, Data_Last);
3478 return Matches (0).First >= Data'First;
3485 function Operand (P : Pointer) return Pointer is
3494 procedure Optimize (Self : in out Pattern_Matcher) is
3495 Max_Length : Program_Size;
3496 This_Length : Program_Size;
3499 Program : Program_Data renames Self.Program;
3502 -- Start with safe defaults (no optimization):
3503 -- * No known first character of match
3504 -- * Does not necessarily start at beginning of line
3505 -- * No string known that has to appear in data
3507 Self.First := ASCII.NUL;
3508 Self.Anchored := False;
3509 Self.Must_Have := Program'Last + 1;
3510 Self.Must_Have_Length := 0;
3512 Scan := Program_First + 1; -- First instruction (can be anything)
3514 if Program (Scan) = EXACT then
3515 Self.First := Program (String_Operand (Scan));
3517 elsif Program (Scan) = BOL
3518 or else Program (Scan) = SBOL
3519 or else Program (Scan) = MBOL
3521 Self.Anchored := True;
3524 -- If there's something expensive in the regexp, find the
3525 -- longest literal string that must appear and make it the
3526 -- regmust. Resolve ties in favor of later strings, since
3527 -- the regstart check works with the beginning of the regexp.
3528 -- and avoiding duplication strengthens checking. Not a
3529 -- strong reason, but sufficient in the absence of others.
3531 if False then -- if Flags.SP_Start then ???
3534 while Scan /= 0 loop
3535 if Program (Scan) = EXACT or else Program (Scan) = EXACTF then
3536 This_Length := String_Length (Program, Scan);
3538 if This_Length >= Max_Length then
3539 Longest := String_Operand (Scan);
3540 Max_Length := This_Length;
3544 Scan := Get_Next (Program, Scan);
3547 Self.Must_Have := Longest;
3548 Self.Must_Have_Length := Natural (Max_Length) + 1;
3556 function Paren_Count (Regexp : Pattern_Matcher) return Match_Count is
3558 return Regexp.Paren_Count;
3565 function Quote (Str : String) return String is
3566 S : String (1 .. Str'Length * 2);
3567 Last : Natural := 0;
3570 for J in Str'Range loop
3572 when '^' | '$' | '|' | '*' | '+' | '?' | '{' |
3573 '}' | '[' | ']' | '(' | ')' | '\' =>
3575 S (Last + 1) := '\';
3576 S (Last + 2) := Str (J);
3580 S (Last + 1) := Str (J);
3585 return S (1 .. Last);
3592 function Read_Natural
3593 (Program : Program_Data;
3594 IP : Pointer) return Natural
3597 return Character'Pos (Program (IP)) +
3598 256 * Character'Pos (Program (IP + 1));
3605 procedure Reset_Class (Bitmap : out Character_Class) is
3607 Bitmap := (others => 0);
3614 procedure Set_In_Class
3615 (Bitmap : in out Character_Class;
3618 Value : constant Class_Byte := Character'Pos (C);
3620 Bitmap (Value / 8) := Bitmap (Value / 8)
3621 or Bit_Conversion (Value mod 8);
3628 function String_Length
3629 (Program : Program_Data;
3630 P : Pointer) return Program_Size
3633 pragma Assert (Program (P) = EXACT or else Program (P) = EXACTF);
3634 return Character'Pos (Program (P + 3));
3637 --------------------
3638 -- String_Operand --
3639 --------------------
3641 function String_Operand (P : Pointer) return Pointer is