:- module(regex). :- comment(summary, "Interface to POSIX regular expression handling"). :- comment(author, "J Chamois"). :- comment(date, "$Date: 2004/11/12 18:55:21 $"). :- comment(copyright, "Public domain"). :- export compile_pattern/3, match/2, match/3, match/4, matchsub/4, matchall/4, split/4. :- get_flag(hostarch, Arch), get_flag(object_suffix, SO), concat_string([Arch,"/eregex.",SO], SOFile), ( exists(SOFile) -> load(SOFile), external(compile_pattern/3, ec_regcomp),% (Pattern,Options,Compiled) external(match/3, ec_regmatch), % (Pattern,String,Options) external(match/4, ec_regmatch4), % (Pattern,String,Options,Match) external(matchsub/4, ec_regmatchsub), % (Pattern,String,Options,Subs) external(matchall/4, ec_regmatchall), % (Pattern,String,Options,All) external(split/4, ec_regsplit) % (Pattern,String,Options,All) ; % Because of libraries test, don't make error messages at load time compile_term([ compile_pattern(P,O,C) :- error(141, compile_pattern(P,O,C)), match(P,S,O) :- error(141, match(P,S,O)), match(P,S,O,M) :- error(141, match(P,S,O,M)), matchsub(P,S,O,M) :- error(141, matchsub(P,S,O,M)), matchall(P,S,O,M) :- error(141, matchall(P,S,O,M)), split(P,S,O,M) :- error(141, split(P,S,O,M)) ]) % printf(error, % "Sorry, library(regex) currently not available on %s architecture%n", % [Arch]) ). match(Pattern, String) :- match(Pattern, String, []). /* bench(N) :- P = "ab+a", S = "shghcgbfhscggfcgabahdfnavlkb", cputime(T0), bench1(N, P, S), cputime(T1), R1 is T1-T0, writeln(R1), bench2(N, P, S), cputime(T2), R2 is T2-T1, writeln(R2), bench3(N, P, S), cputime(T3), R3 is T3-T2, writeln(R3), true. bench1(N, P, S) :- between(1,N,1,_), match(P, S, []), fail. bench1(_, _, _). bench2(N, P, S) :- between(1,N,1,_), match(P, S, [], _), fail. bench2(_, _, _). bench3(N, P, S) :- compile_pattern(P, [], C), between(1,N,1,_), match(C, S, [], _), fail. bench3(_, _, _). bench1(N, P, S) :- ( for(_,1,N),param(P,S) do match(P, S, []) ). bench2(N, P, S) :- ( for(_,1,N),param(P,S) do match(P, S, [], _) ). bench3(N, P, S) :- compile_pattern(P, [], C), ( for(_,1,N),param(C,S) do match(C, S, [], _) ). */ %---------------------------------------------------------------------- % DOCUMENTATION %---------------------------------------------------------------------- :- comment(desc, html(" This library implements an ECLiPSe API for POSIX 1003.2 regular expressions (on Unix systems it calls the regular expression functions from the standard library, on Windows it uses Henry Spencer's regex library version 3.8).

Regular Expressions

This is just a very brief summary of the essentials. For details of regular expressions see any POSIX regex(7) man page. Two types of regular expressions are supported:
Extended Regular Expressions (the default)
These are described below and correspond essentially to those understood by the UNIX egrep command.
Basic Regular Expressions
These correspond essentially to those in the UNIX ed editor or the grep command, and are mostly obsolete.
Note that our choice of default differs from the POSIX 1003.2 C API.

Characters

Every character stands for itself, except for the characters ^.[$()|*+?{\\ which must be escaped with a \\ to prevent them from having special meaning.
.
Matches any character
[aeiou]
Matches any of the characters between the bracktes
[^aeiou]
Matches any character except those listed
[a-z0-9]
Matches any character in the given ranges

Anchors

^
Matches at the beginning of the string (or line)
$
Matches at the end of the string (or line)

Repetition

?
Matches the preceding element 0 or 1 times
*
Matches the preceding element 0 or more times
+
Matches the preceding element 1 or more times
{3}
Matches the preceding element 3 times
{1,3}
Matches the preceding element 1 to 3 times

Grouping

(subexpr)
Matches the parenthesized expression. This grouping is used in connection with the repetition operators, or for indicating subexpressions whose matches are to be captured and returned
(one|two|three)
Matches any of the alternative expressions

Options

Most of the predicates in this library accept a list of options. The accepted options are:
basic
Interpret the pattern as a Basic Regular Expression, rather than the default Extended Regular Expression.
extended
Interpret the pattern as an Extended Regular Expression (this flag is redundant since this is the default).
icase
Ignore case when matching.
newline
Treat newlines specially, i.e. don't treat them as normal characters and make ^ match after a newline and $ before a newline. By default, newlines are treated as ordinary characters.
notbol
Don't interpret the beginning of the string as the beginning of a line, i.e. don't let ^ match there.
noteol
Don't interpret the end of the string as the end of a line, i.e. don't let $ match there.

Shortcomings

  1. Due to limitations of the underlying implementation, the predicates in this library do not handle embedded NUL characters in strings correctly (they are interpreted as the end of the string).
  2. POSIX regular expressions don't seem to have a notion of \"noncapturing parentheses\", i.e. parentheses that are only used for grouping, not for indicating that one wants to capture the matching substring.
  3. In an environment like ECLiPSe, one would like to be able to do things like 
    	?- ideal_match(\"(/[^/]*)+\", \"/usr/local/eclipse\", L).
	L = [\"/usr\", \"/local\", \"/eclipse\"]
	Yes
    i.e. capture every instance of a matching subexpression. There seems to be no way to do that with a POSIX regexp implementation.
")). :- comment(match/2, [ summary:"A substring of String matches the regular expression Pattern", args:["Pattern":"A string (or a compiled pattern handle)", "String":"A string"], amode:match(+,+), fail_if:"String does not match Pattern", desc:html(" Succeeds if all or a substring of String matches the regular expression Pattern. For the description of regular expressions see the library(regex) page.

Equivalent to match(Pattern, String, []). "), see_also:[library(regex),match/3,match/4,matchsub/4,matchall/4,split/4,compile_pattern/3], eg:" ?- match(\"aca\", \"abracadabra\"). Yes ?- match(\"^a[cd]a$\", \"abracadabra\"). No ?- match(\"^a[cd]a$\", \"ada\"). Yes " ]). :- comment(match/3, [ summary:"A substring of String matches the regular expression Pattern", args:["Pattern":"A string (or a compiled pattern handle)", "String":"A string", "Options":"List of atoms"], amode:match(+,+,+), fail_if:"String does not match Pattern", see_also:[library(regex),match/2,match/4,matchall/4,matchsub/4,split/4,compile_pattern/3], desc:html(" Succeeds if all or a substring of String matches the regular expression Pattern. For the description of regular expressions see the library(regex) page.

Options is a (possibly empty) list of atomic option names, as described in the library(regex) page. "), eg:" ?- match(\"april\", \"April\", []). No ?- match(\"april\", \"April\", [icase]). Yes ?- match(\"(^(a[cd]a)$)+\", \"aca\\nada\", []). No ?- match(\"(^(a[cd]a)$)+\", \"aca\\nada\", [newline]). Yes ?- match(\"\\\\\", \"a word only\", [basic]). Yes ?- match(\"\\\\\", \"not words though\", [basic]). No " ]). :- comment(match/4, [ summary:"Match is the first substring of String that matches the regular expression Pattern", args:["Pattern":"A string (or a compiled pattern handle)", "String":"A string", "Options":"List of atoms", "Match":"Output: a string"], amode:match(+,+,+,-), fail_if:"String does not match Pattern", see_also:[library(regex),match/2,match/3,matchsub/4,matchall/4,compile_pattern/3], desc:html(" Succeeds if all or a substring of String matches the regular expression Pattern. For the description of regular expressions see the library(regex) page.

Options is a (possibly empty) list of atomic option names, as described in the library(regex) page.

Match is bound to the first substring of String which matches Pattern.

Note that this predicate does not return any information about matching (parenthesised) sub-expressions! "), eg:" ?- match(\"<[a-z]+>\", \"Text with tags\", [icase], M). M = \"\" Yes " ]). :- comment(matchsub/4, [ summary:"A substring of String matches the regular expression Pattern and SubMatches are matching sub-expressions", args:["Pattern":"A string (or a compiled pattern handle)", "String":"A string", "Options":"List of atoms", "SubMatches":"Output: List of strings"], amode:matchsub(+,+,+,-), fail_if:"String does not match Pattern", see_also:[library(regex),match/2,match/3,match/4,compile_pattern/3], desc:html(" Succeeds if all or a substring of String matches the regular expression Pattern. For the description of regular expressions see the library(regex) page.

Options is a (possibly empty) list of atomic option names, as described in the library(regex) page.

SubMatches is bound to a list of strings, each corresponding to a parenthesized subexpression in Pattern. These subexpressions are ordered according to the position of their opening parenthesis within the pattern. The matching string appears on the corresponding position in the SubMatches list. Note that, if a subexpression matches several times, only the last match is returned.

"), eg:" ?- matchsub(\"Name:([^,]+), Age:([0-9]+),\", \"Name:Fred, Age:34,\", [], L). L = [\"Fred\", \"34\"] Yes " ]). :- comment(matchall/4, [ summary:"AllMatches is a list of substrings of String which match the regular expression Pattern", args:["Pattern":"A string (or a compiled pattern handle)", "String":"A string", "Options":"List of atoms", "AllMatches":"Output: List of strings"], amode:matchall(+,+,+,-), fail_if:"None", see_also:[library(regex),match/2,match/3,match/4,split/4,compile_pattern/3], desc:html(" This predicates always succeeds.

Options is a (possibly empty) list of atomic option names, as described in the library(regex) page.

AllMatches is bound to a list of strings. If the input string String does not match the pattern, the list is empty. Otherwise the list contains substrings of String which match the entire pattern, ordered according to their occurrence within String. No overlapping matches are returned, i.e. the next match is found by examining the remainder of String after the previous match.

Note that this predicate does not return any information about matching (parenthesised) sub-expressions! "), eg:" ?- matchall(\"[0-9]+\", \" blue 27 red123 green99\", [], L). L = [\"27\", \"123\", \"99\"] Yes ?- matchall(\"([0-9]+|[^0-9]+)\", \" blue 27 red123 green99\", [], L). L = [\" blue \", \"27\", \" red\", \"123\", \" green\", \"99\"] Yes " ]). :- comment(split/4, [ summary:"Parts is a list of substrings, partitioning String according to Pattern", args:["Pattern":"A string (or a compiled pattern handle)", "String":"A string", "Options":"List of atoms", "Parts":"Output: List of strings"], amode:split(+,+,+,-), fail_if:"None", see_also:[library(regex),matchall/4,compile_pattern/3,concat_string/2,split_string/4], desc:html(" This predicates always succeeds.

Options is a (possibly empty) list of atomic option names, as described in the library(regex) page.

Parts is bound to a list of strings which are consecutive substrings of the input string String (i.e. concatenating this list using concat_strings/2 will yield the original String). The list is constructed such that it has an odd number of elements, where the even numbered elements match the pattern, and the odd numbered elements contains those portions of String that did not match the Pattern. Some of these substrings may be empty.

This partitioning of the String can be used to construct a new string with the matches replaced by something else. Use the following code pattern: 

	    split(Pattern, String, Options, Parts),
	    (
		fromto(Parts, [NoMatch,Match|NMs], NMs, [Last]),
		fromto(Repl,  [NoMatch,Subst|NSs], NSs, [Last])
	    do
		%%% compute Subst from Match here %%%
	    ),
	    concat_string(Repl, NewString),

Note that the split/4 predicate does not return any information about matching (parenthesised) sub-expressions! "), eg:" ?- split(\"cad\", \"abracadabra\", [], Parts). Parts = [\"abra\", \"cad\", \"abra\"] Yes (0.00s cpu) ?- split(\"bra\", \"abracadabra\", [], Parts). Parts = [\"a\", \"bra\", \"cada\", \"bra\", \"\"] Yes (0.00s cpu) ?- split(\"bla\", \"abracadabra\", [], Parts). Parts = [\"abracadabra\"] Yes (0.00s cpu) ?- split(\"a\", \"aaa\", [], Parts). Parts = [\"\", \"a\", \"\", \"a\", \"\", \"a\", \"\"] Yes (0.00s cpu) ?- split(\"%[a-z]\", \"format %s %f=%d.\", [], Parts). Parts = [\"format \", \"%s\", \" \", \"%f\", \"=\", \"%d\", \".\"] Yes (0.00s cpu) % With the following definition replace(Pattern, New, String, NewString) :- split(Pattern, String, [], Parts), ( fromto(Parts, [NoMatch,_Match|NMs], NMs, [Last]), fromto(Repl, [NoMatch, Subst|NSs], NSs, [Last]), param(New) do Subst = New ), concat_string(Repl, NewString). ?- replace(\"2\", \"to\", \"2 be or not 2 be\", S). S = \"to be or not to be\" Yes (0.00s cpu) " ]). :- comment(compile_pattern/3, [ summary:"Precompile a pattern for repeated use", args:["Pattern":"A string", "Options":"List of atoms", "CompiledPattern":"Output: a compiled pattern handle"], amode:compile_pattern(+,+,-), fail_if:"None", see_also:[library(regex),match/2,match/3,match/4,matchsub/4,matchall/4], desc:html(" All matching predicates in this library accept either a regular expression in string form, or a precompiled regular expression. The matching predicates will execute faster if invoked with a precompiled pattern instead of the string. Whenever a pattern needs to be matched more than once, it will typically be more efficient to work with a precompiled pattern.

Options is a (possibly empty) list of atomic option names, as described in the library(regex) page. The options give here should be the same as the ones given to the matching-predicates later.

The space consumed by the compiled pattern will be automatically reclaimed on failure, or on garbage collection when no longer needed. "), eg:" ?- compile_pattern(\"ab+a\", [], C), match(C, \"bbabbbaab\", [], M1), match(C, \"abacus\", [], M2). C = 'HANDLE'(16'00025c60) M1 = \"abbba\" M2 = \"aba\" Yes " ]).