Mercurial > projects > ldc
view tango/tango/core/Array.d @ 132:1700239cab2e trunk
[svn r136] MAJOR UNSTABLE UPDATE!!!
Initial commit after moving to Tango instead of Phobos.
Lots of bugfixes...
This build is not suitable for most things.
author | lindquist |
---|---|
date | Fri, 11 Jan 2008 17:57:40 +0100 |
parents | |
children |
line wrap: on
line source
/** * The array module provides array manipulation routines in a manner that * balances performance and flexibility. Operations are provided for sorting, * and for processing both sorted and unsorted arrays. * * Copyright: Copyright (C) 2005-2006 Sean Kelly. All rights reserved. * License: BSD style: $(LICENSE) * Authors: Sean Kelly */ module tango.core.Array; private import tango.core.Traits; private import tango.stdc.stdlib : alloca; version( DDoc ) { typedef int Num; typedef int Elem; typedef bool function( Elem ) Pred1E; typedef bool function( Elem, Elem ) Pred2E; } private { struct IsEqual( T ) { static bool opCall( T p1, T p2 ) { // TODO: Fix this if/when opEquals is changed to return a bool. static if( is( T == class ) || is( T == struct ) ) return (p1 == p2) != 0; else return p1 == p2; } } struct IsLess( T ) { static bool opCall( T p1, T p2 ) { return p1 < p2; } } template ElemTypeOf( T ) { alias typeof(T[0]) ElemTypeOf; } } //////////////////////////////////////////////////////////////////////////////// // Find //////////////////////////////////////////////////////////////////////////////// version( DDoc ) { /** * Performs a linear scan of buf from $(LB)0 .. buf.length$(RP), returning * the index of the first element matching pat, or buf.length if no match * was found. Comparisons will be performed using the supplied predicate * or '==' if none is supplied. * * Params: * buf = The array to search. * pat = The pattern to search for. * pred = The evaluation predicate, which should return true if e1 is * equal to e2 and false if not. This predicate may be any * callable type. * * Returns: * The index of the first match or buf.length if no match was found. */ size_t find( Elem[] buf, Elem pat, Pred2E pred = Pred2E.init ); /** * Performs a linear scan of buf from $(LB)0 .. buf.length$(RP), returning * the index of the first element matching pat, or buf.length if no match * was found. Comparisons will be performed using the supplied predicate * or '==' if none is supplied. * * Params: * buf = The array to search. * pat = The pattern to search for. * pred = The evaluation predicate, which should return true if e1 is * equal to e2 and false if not. This predicate may be any * callable type. * * Returns: * The index of the first match or buf.length if no match was found. */ size_t find( Elem[] buf, Elem[] pat, Pred2E pred = Pred2E.init ); } else { template find_( Elem, Pred = IsEqual!(Elem) ) { static assert( isCallableType!(Pred) ); size_t fn( Elem[] buf, Elem pat, Pred pred = Pred.init ) { foreach( size_t pos, Elem cur; buf ) { if( pred( cur, pat ) ) return pos; } return buf.length; } size_t fn( Elem[] buf, Elem[] pat, Pred pred = Pred.init ) { if( buf.length == 0 || pat.length == 0 || buf.length < pat.length ) { return buf.length; } size_t end = buf.length - pat.length + 1; for( size_t pos = 0; pos < end; ++pos ) { if( pred( buf[pos], pat[0] ) ) { size_t mat = 0; do { if( ++mat >= pat.length ) return pos - pat.length + 1; if( ++pos >= buf.length ) return buf.length; } while( pred( buf[pos], pat[mat] ) ); pos -= mat; } } return buf.length; } } template find( Buf, Pat ) { size_t find( Buf buf, Pat pat ) { return find_!(ElemTypeOf!(Buf)).fn( buf, pat ); } } template find( Buf, Pat, Pred ) { size_t find( Buf buf, Pat pat, Pred pred ) { return find_!(ElemTypeOf!(Buf), Pred).fn( buf, pat, pred ); } } debug( UnitTest ) { unittest { // find element assert( find( "", 'a' ) == 0 ); assert( find( "abc", 'a' ) == 0 ); assert( find( "abc", 'b' ) == 1 ); assert( find( "abc", 'c' ) == 2 ); assert( find( "abc", 'd' ) == 3 ); // null parameters assert( find( "", "" ) == 0 ); assert( find( "a", "" ) == 1 ); assert( find( "", "a" ) == 0 ); // exact match assert( find( "abc", "abc" ) == 0 ); // simple substring match assert( find( "abc", "a" ) == 0 ); assert( find( "abca", "a" ) == 0 ); assert( find( "abc", "b" ) == 1 ); assert( find( "abc", "c" ) == 2 ); assert( find( "abc", "d" ) == 3 ); // multi-char substring match assert( find( "abc", "ab" ) == 0 ); assert( find( "abcab", "ab" ) == 0 ); assert( find( "abc", "bc" ) == 1 ); assert( find( "abc", "ac" ) == 3 ); assert( find( "abrabracadabra", "abracadabra" ) == 3 ); } } } //////////////////////////////////////////////////////////////////////////////// // Reverse Find //////////////////////////////////////////////////////////////////////////////// version( DDoc ) { /** * Performs a linear scan of buf from $(LP)buf.length .. 0$(RB), returning * the index of the first element matching pat, or buf.length if no match * was found. Comparisons will be performed using the supplied predicate * or '==' if none is supplied. * * Params: * buf = The array to search. * pat = The pattern to search for. * pred = The evaluation predicate, which should return true if e1 is * equal to e2 and false if not. This predicate may be any * callable type. * * Returns: * The index of the first match or buf.length if no match was found. */ size_t rfind( Elem[] buf, Elem pat, Pred2E pred = Pred2E.init ); /** * Performs a linear scan of buf from $(LP)buf.length .. 0$(RB), returning * the index of the first element matching pat, or buf.length if no match * was found. Comparisons will be performed using the supplied predicate * or '==' if none is supplied. * * Params: * buf = The array to search. * pat = The pattern to search for. * pred = The evaluation predicate, which should return true if e1 is * equal to e2 and false if not. This predicate may be any * callable type. * * Returns: * The index of the first match or buf.length if no match was found. */ size_t rfind( Elem[] buf, Elem[] pat, Pred2E pred = Pred2E.init ); } else { template rfind_( Elem, Pred = IsEqual!(Elem) ) { static assert( isCallableType!(Pred) ); size_t fn( Elem[] buf, Elem pat, Pred pred = Pred.init ) { if( buf.length == 0 ) return buf.length; size_t pos = buf.length; do { if( pred( buf[--pos], pat ) ) return pos; } while( pos > 0 ); return buf.length; } size_t fn( Elem[] buf, Elem[] pat, Pred pred = Pred.init ) { if( buf.length == 0 || pat.length == 0 || buf.length < pat.length ) { return buf.length; } size_t pos = buf.length - pat.length + 1; do { if( pred( buf[--pos], pat[0] ) ) { size_t mat = 0; do { if( ++mat >= pat.length ) return pos - pat.length + 1; if( ++pos >= buf.length ) return buf.length; } while( pred( buf[pos], pat[mat] ) ); pos -= mat; } } while( pos > 0 ); return buf.length; } } template rfind( Buf, Pat ) { size_t rfind( Buf buf, Pat pat ) { return rfind_!(ElemTypeOf!(Buf)).fn( buf, pat ); } } template rfind( Buf, Pat, Pred ) { size_t rfind( Buf buf, Pat pat, Pred pred ) { return rfind_!(ElemTypeOf!(Buf), Pred).fn( buf, pat, pred ); } } debug( UnitTest ) { unittest { // rfind element assert( rfind( "", 'a' ) == 0 ); assert( rfind( "abc", 'a' ) == 0 ); assert( rfind( "abc", 'b' ) == 1 ); assert( rfind( "abc", 'c' ) == 2 ); assert( rfind( "abc", 'd' ) == 3 ); // null parameters assert( rfind( "", "" ) == 0 ); assert( rfind( "a", "" ) == 1 ); assert( rfind( "", "a" ) == 0 ); // exact match assert( rfind( "abc", "abc" ) == 0 ); // simple substring match assert( rfind( "abc", "a" ) == 0 ); assert( rfind( "abca", "a" ) == 3 ); assert( rfind( "abc", "b" ) == 1 ); assert( rfind( "abc", "c" ) == 2 ); assert( rfind( "abc", "d" ) == 3 ); // multi-char substring match assert( rfind( "abc", "ab" ) == 0 ); assert( rfind( "abcab", "ab" ) == 3 ); assert( rfind( "abc", "bc" ) == 1 ); assert( rfind( "abc", "ac" ) == 3 ); assert( rfind( "abracadabrabra", "abracadabra" ) == 0 ); } } } //////////////////////////////////////////////////////////////////////////////// // KMP Find //////////////////////////////////////////////////////////////////////////////// version( DDoc ) { /** * Performs a linear scan of buf from $(LB)0 .. buf.length$(RP), returning * the index of the first element matching pat, or buf.length if no match * was found. Comparisons will be performed using the supplied predicate * or '==' if none is supplied. * * This function uses the KMP algorithm and offers O(M+N) performance but * must allocate a temporary buffer of size pat.sizeof to do so. If it is * available on the target system, alloca will be used for the allocation, * otherwise a standard dynamic memory allocation will occur. * * Params: * buf = The array to search. * pat = The pattern to search for. * pred = The evaluation predicate, which should return true if e1 is * equal to e2 and false if not. This predicate may be any * callable type. * * Returns: * The index of the first match or buf.length if no match was found. */ size_t kfind( Elem[] buf, Elem pat, Pred2E pred = Pred2E.init ); /** * Performs a linear scan of buf from $(LB)0 .. buf.length$(RP), returning * the index of the first element matching pat, or buf.length if no match * was found. Comparisons will be performed using the supplied predicate * or '==' if none is supplied. * * This function uses the KMP algorithm and offers O(M+N) performance but * must allocate a temporary buffer of size pat.sizeof to do so. If it is * available on the target system, alloca will be used for the allocation, * otherwise a standard dynamic memory allocation will occur. * * Params: * buf = The array to search. * pat = The pattern to search for. * pred = The evaluation predicate, which should return true if e1 is * equal to e2 and false if not. This predicate may be any * callable type. * * Returns: * The index of the first match or buf.length if no match was found. */ size_t kfind( Elem[] buf, Elem[] pat, Pred2E pred = Pred2E.init ); } else { template kfind_( Elem, Pred = IsEqual!(Elem) ) { static assert( isCallableType!(Pred) ); size_t fn( Elem[] buf, Elem pat, Pred pred = Pred.init ) { foreach( size_t pos, Elem cur; buf ) { if( pred( cur, pat ) ) return pos; } return buf.length; } size_t fn( Elem[] buf, Elem[] pat, Pred pred = Pred.init ) { if( buf.length == 0 || pat.length == 0 || buf.length < pat.length ) { return buf.length; } static if( is( alloca ) ) { size_t[] func = (cast(size_t*) alloca( (pat.length + 1) * size_t.sizeof ))[0 .. pat.length + 1]; } else { size_t[] func = new size_t[pat.length + 1]; scope( exit ) delete func; // force cleanup } func[0] = 0; // // building prefix-function // for( size_t m = 0, i = 1 ; i < pat.length ; ++i ) { while( ( m > 0 ) && !pred( pat[m], pat[i] ) ) m = func[m - 1]; if( pred( pat[m], pat[i] ) ) ++m; func[i] = m; } // // searching // for( size_t m = 0, i = 0; i < buf.length; ++i ) { while( ( m > 0 ) && !pred( pat[m], buf[i] ) ) m = func[m - 1]; if( pred( pat[m], buf[i] ) ) { ++m; if( m == pat.length ) { return i - pat.length + 1; } } } return buf.length; } } template kfind( Buf, Pat ) { size_t kfind( Buf buf, Pat pat ) { return kfind_!(ElemTypeOf!(Buf)).fn( buf, pat ); } } template kfind( Buf, Pat, Pred ) { size_t kfind( Buf buf, Pat pat, Pred pred ) { return kfind_!(ElemTypeOf!(Buf), Pred).fn( buf, pat, pred ); } } debug( UnitTest ) { unittest { // find element assert( kfind( "", 'a' ) == 0 ); assert( kfind( "abc", 'a' ) == 0 ); assert( kfind( "abc", 'b' ) == 1 ); assert( kfind( "abc", 'c' ) == 2 ); assert( kfind( "abc", 'd' ) == 3 ); // null parameters assert( kfind( "", "" ) == 0 ); assert( kfind( "a", "" ) == 1 ); assert( kfind( "", "a" ) == 0 ); // exact match assert( kfind( "abc", "abc" ) == 0 ); // simple substring match assert( kfind( "abc", "a" ) == 0 ); assert( kfind( "abca", "a" ) == 0 ); assert( kfind( "abc", "b" ) == 1 ); assert( kfind( "abc", "c" ) == 2 ); assert( kfind( "abc", "d" ) == 3 ); // multi-char substring match assert( kfind( "abc", "ab" ) == 0 ); assert( kfind( "abcab", "ab" ) == 0 ); assert( kfind( "abc", "bc" ) == 1 ); assert( kfind( "abc", "ac" ) == 3 ); assert( kfind( "abrabracadabra", "abracadabra" ) == 3 ); } } } //////////////////////////////////////////////////////////////////////////////// // KMP Reverse Find //////////////////////////////////////////////////////////////////////////////// version( DDoc ) { /** * Performs a linear scan of buf from $(LP)buf.length .. 0$(RB), returning * the index of the first element matching pat, or buf.length if no match * was found. Comparisons will be performed using the supplied predicate * or '==' if none is supplied. * * This function uses the KMP algorithm and offers O(M+N) performance but * must allocate a temporary buffer of size pat.sizeof to do so. If it is * available on the target system, alloca will be used for the allocation, * otherwise a standard dynamic memory allocation will occur. * * Params: * buf = The array to search. * pat = The pattern to search for. * pred = The evaluation predicate, which should return true if e1 is * equal to e2 and false if not. This predicate may be any * callable type. * * Returns: * The index of the first match or buf.length if no match was found. */ size_t krfind( Elem[] buf, Elem pat, Pred2E pred = Pred2E.init ); /** * Performs a linear scan of buf from $(LP)buf.length .. 0$(RB), returning * the index of the first element matching pat, or buf.length if no match * was found. Comparisons will be performed using the supplied predicate * or '==' if none is supplied. * * This function uses the KMP algorithm and offers O(M+N) performance but * must allocate a temporary buffer of size pat.sizeof to do so. If it is * available on the target system, alloca will be used for the allocation, * otherwise a standard dynamic memory allocation will occur. * * Params: * buf = The array to search. * pat = The pattern to search for. * pred = The evaluation predicate, which should return true if e1 is * equal to e2 and false if not. This predicate may be any * callable type. * * Returns: * The index of the first match or buf.length if no match was found. */ size_t krfind( Elem[] buf, Elem[] pat, Pred2E pred = Pred2E.init ); } else { template krfind_( Elem, Pred = IsEqual!(Elem) ) { static assert( isCallableType!(Pred) ); size_t fn( Elem[] buf, Elem pat, Pred pred = Pred.init ) { if( buf.length == 0 ) return buf.length; size_t pos = buf.length; do { if( pred( buf[--pos], pat ) ) return pos; } while( pos > 0 ); return buf.length; } size_t fn( Elem[] buf, Elem[] pat, Pred pred = Pred.init ) { if( buf.length == 0 || pat.length == 0 || buf.length < pat.length ) { return buf.length; } static if( is( alloca ) ) { size_t[] func = (cast(size_t*) alloca( (pat.length + 1) * size_t.sizeof ))[0 .. pat.length + 1]; } else { size_t[] func = new size_t[pat.length + 1]; scope( exit ) delete func; // force cleanup } func[$ - 1] = 0; // // building prefix-function // for( size_t m = 0, i = pat.length - 1; i > 0; --i ) { while( ( m > 0 ) && !pred( pat[length - m - 1], pat[i - 1] ) ) m = func[length - m]; if( pred( pat[length - m - 1], pat[i - 1] ) ) ++m; func[i - 1] = m; } // // searching // size_t m = 0; size_t i = buf.length; do { --i; while( ( m > 0 ) && !pred( pat[length - m - 1], buf[i] ) ) m = func[length - m - 1]; if( pred( pat[length - m - 1], buf[i] ) ) { ++m; if ( m == pat.length ) { return i; } } } while( i > 0 ); return buf.length; } } template krfind( Buf, Pat ) { size_t krfind( Buf buf, Pat pat ) { return krfind_!(ElemTypeOf!(Buf)).fn( buf, pat ); } } template krfind( Buf, Pat, Pred ) { size_t krfind( Buf buf, Pat pat, Pred pred ) { return krfind_!(ElemTypeOf!(Buf), Pred).fn( buf, pat, pred ); } } debug( UnitTest ) { unittest { // rfind element assert( krfind( "", 'a' ) == 0 ); assert( krfind( "abc", 'a' ) == 0 ); assert( krfind( "abc", 'b' ) == 1 ); assert( krfind( "abc", 'c' ) == 2 ); assert( krfind( "abc", 'd' ) == 3 ); // null parameters assert( krfind( "", "" ) == 0 ); assert( krfind( "a", "" ) == 1 ); assert( krfind( "", "a" ) == 0 ); // exact match assert( krfind( "abc", "abc" ) == 0 ); // simple substring match assert( krfind( "abc", "a" ) == 0 ); assert( krfind( "abca", "a" ) == 3 ); assert( krfind( "abc", "b" ) == 1 ); assert( krfind( "abc", "c" ) == 2 ); assert( krfind( "abc", "d" ) == 3 ); // multi-char substring match assert( krfind( "abc", "ab" ) == 0 ); assert( krfind( "abcab", "ab" ) == 3 ); assert( krfind( "abc", "bc" ) == 1 ); assert( krfind( "abc", "ac" ) == 3 ); assert( krfind( "abracadabrabra", "abracadabra" ) == 0 ); } } } //////////////////////////////////////////////////////////////////////////////// // Find-If //////////////////////////////////////////////////////////////////////////////// version( DDoc ) { /** * Performs a linear scan of buf from $(LB)0 .. buf.length$(RP), returning * the index of the first element where pred returns true. * * Params: * buf = The array to search. * pred = The evaluation predicate, which should return true if the * element is a valid match and false if not. This predicate * may be any callable type. * * Returns: * The index of the first match or buf.length if no match was found. */ size_t findIf( Elem[] buf, Pred1E pred ); } else { template findIf_( Elem, Pred ) { static assert( isCallableType!(Pred) ); size_t fn( Elem[] buf, Pred pred ) { foreach( size_t pos, Elem cur; buf ) { if( pred( cur ) ) return pos; } return buf.length; } } template findIf( Buf, Pred ) { size_t findIf( Buf buf, Pred pred ) { return findIf_!(ElemTypeOf!(Buf), Pred).fn( buf, pred ); } } debug( UnitTest ) { unittest { assert( findIf( "bcecg", ( char c ) { return c == 'a'; } ) == 5 ); assert( findIf( "bcecg", ( char c ) { return c == 'b'; } ) == 0 ); assert( findIf( "bcecg", ( char c ) { return c == 'c'; } ) == 1 ); assert( findIf( "bcecg", ( char c ) { return c == 'd'; } ) == 5 ); assert( findIf( "bcecg", ( char c ) { return c == 'g'; } ) == 4 ); assert( findIf( "bcecg", ( char c ) { return c == 'h'; } ) == 5 ); } } } //////////////////////////////////////////////////////////////////////////////// // Reverse Find-If //////////////////////////////////////////////////////////////////////////////// version( DDoc ) { /** * Performs a linear scan of buf from $(LP)buf.length .. 0$(RB), returning * the index of the first element where pred returns true. * * Params: * buf = The array to search. * pred = The evaluation predicate, which should return true if the * element is a valid match and false if not. This predicate * may be any callable type. * * Returns: * The index of the first match or buf.length if no match was found. */ size_t rfindIf( Elem[] buf, Pred1E pred ); } else { template rfindIf_( Elem, Pred ) { static assert( isCallableType!(Pred) ); size_t fn( Elem[] buf, Pred pred ) { if( buf.length == 0 ) return buf.length; size_t pos = buf.length; do { if( pred( buf[--pos] ) ) return pos; } while( pos > 0 ); return buf.length; } } template rfindIf( Buf, Pred ) { size_t rfindIf( Buf buf, Pred pred ) { return rfindIf_!(ElemTypeOf!(Buf), Pred).fn( buf, pred ); } } debug( UnitTest ) { unittest { assert( rfindIf( "bcecg", ( char c ) { return c == 'a'; } ) == 5 ); assert( rfindIf( "bcecg", ( char c ) { return c == 'b'; } ) == 0 ); assert( rfindIf( "bcecg", ( char c ) { return c == 'c'; } ) == 3 ); assert( rfindIf( "bcecg", ( char c ) { return c == 'd'; } ) == 5 ); assert( rfindIf( "bcecg", ( char c ) { return c == 'g'; } ) == 4 ); assert( rfindIf( "bcecg", ( char c ) { return c == 'h'; } ) == 5 ); } } } //////////////////////////////////////////////////////////////////////////////// // Find Adjacent //////////////////////////////////////////////////////////////////////////////// version( DDoc ) { /** * Performs a linear scan of buf from $(LB)0 .. buf.length$(RP), returning * the index of the first element that compares equal to the next element * in the sequence. Comparisons will be performed using the supplied * predicate or '==' if none is supplied. * * Params: * buf = The array to scan. * pred = The evaluation predicate, which should return true if e1 is * equal to e2 and false if not. This predicate may be any * callable type. * * Returns: * The index of the first match or buf.length if no match was found. */ size_t findAdj( Elem[] buf, Elem pat, Pred2E pred = Pred2E.init ); } else { template findAdj_( Elem, Pred = IsEqual!(Elem) ) { static assert( isCallableType!(Pred) ); size_t fn( Elem[] buf, Pred pred = Pred.init ) { if( buf.length < 2 ) return buf.length; Elem sav = buf[0]; foreach( size_t pos, Elem cur; buf[1 .. $] ) { if( pred( cur, sav ) ) return pos; sav = cur; } return buf.length; } } template findAdj( Buf ) { size_t findAdj( Buf buf ) { return findAdj_!(ElemTypeOf!(Buf)).fn( buf ); } } template findAdj( Buf, Pred ) { size_t findAdj( Buf buf, Pred pred ) { return findAdj_!(ElemTypeOf!(Buf), Pred).fn( buf, pred ); } } debug( UnitTest ) { unittest { assert( findAdj( "aabcdef" ) == 0 ); assert( findAdj( "abcddef" ) == 3 ); assert( findAdj( "abcdeff" ) == 5 ); assert( findAdj( "abcdefg" ) == 7 ); } } } //////////////////////////////////////////////////////////////////////////////// // Contains //////////////////////////////////////////////////////////////////////////////// version( DDoc ) { /** * Performs a linear scan of buf from $(LB)0 .. buf.length$(RP), returning * true if an element matching pat is found. Comparisons will be performed * using the supplied predicate or '<' if none is supplied. * * Params: * buf = The array to search. * pat = The pattern to search for. * pred = The evaluation predicate, which should return true if e1 is * equal to e2 and false if not. This predicate may be any * callable type. * * Returns: * True if an element equivalent to pat is found, false if not. */ size_t contains( Elem[] buf, Elem pat, Pred2E pred = Pred2E.init ); /** * Performs a linear scan of buf from $(LB)0 .. buf.length$(RP), returning * true if a sequence matching pat is found. Comparisons will be performed * using the supplied predicate or '<' if none is supplied. * * Params: * buf = The array to search. * pat = The pattern to search for. * pred = The evaluation predicate, which should return true if e1 is * equal to e2 and false if not. This predicate may be any * callable type. * * Returns: * True if an element equivalent to pat is found, false if not. */ size_t contains( Elem[] buf, Elem[] pat, Pred2E pred = Pred2E.init ); } else { template contains( Buf, Pat ) { size_t contains( Buf buf, Pat pat ) { return find( buf, pat ) != buf.length; } } template contains( Buf, Pat, Pred ) { size_t contains( Buf buf, Pat pat, Pred pred ) { return find( buf, pat, pred ) != buf.length; } } debug( UnitTest ) { unittest { // find element assert( !contains( "", 'a' ) ); assert( contains( "abc", 'a' ) ); assert( contains( "abc", 'b' ) ); assert( contains( "abc", 'c' ) ); assert( !contains( "abc", 'd' ) ); // null parameters assert( !contains( "", "" ) ); assert( !contains( "a", "" ) ); assert( !contains( "", "a" ) ); // exact match assert( contains( "abc", "abc" ) ); // simple substring match assert( contains( "abc", "a" ) ); assert( contains( "abca", "a" ) ); assert( contains( "abc", "b" ) ); assert( contains( "abc", "c" ) ); assert( !contains( "abc", "d" ) ); // multi-char substring match assert( contains( "abc", "ab" ) ); assert( contains( "abcab", "ab" ) ); assert( contains( "abc", "bc" ) ); assert( !contains( "abc", "ac" ) ); assert( contains( "abrabracadabra", "abracadabra" ) ); } } } //////////////////////////////////////////////////////////////////////////////// // Mismatch //////////////////////////////////////////////////////////////////////////////// version( DDoc ) { /** * Performs a parallel linear scan of bufA and bufB from $(LB)0 .. N$(RP) * where N = min$(LP)bufA.length, bufB.length$(RP), returning the index of * the first element in bufA which does not match the corresponding element * in bufB or N if no mismatch occurs. Comparisons will be performed using * the supplied predicate or '==' if none is supplied. * * Params: * bufA = The array to evaluate. * bufB = The array to match against. * pred = The evaluation predicate, which should return true if e1 is * equal to e2 and false if not. This predicate may be any * callable type. * * Returns: * The index of the first mismatch or N if the first N elements of bufA * and bufB match, where N = min$(LP)bufA.length, bufB.length$(RP). */ size_t mismatch( Elem[] bufA, Elem[] bufB, Pred2E pred = Pred2E.init ); } else { template mismatch_( Elem, Pred = IsEqual!(Elem) ) { static assert( isCallableType!(Pred) ); size_t fn( Elem[] bufA, Elem[] bufB, Pred pred = Pred.init ) { size_t posA = 0, posB = 0; while( posA < bufA.length && posB < bufB.length ) { if( !pred( bufB[posB], bufA[posA] ) ) break; ++posA, ++posB; } return posA; } } template mismatch( BufA, BufB ) { size_t mismatch( BufA bufA, BufB bufB ) { return mismatch_!(ElemTypeOf!(BufA)).fn( bufA, bufB ); } } template mismatch( BufA, BufB, Pred ) { size_t mismatch( BufA bufA, BufB bufB, Pred pred ) { return mismatch_!(ElemTypeOf!(BufA), Pred).fn( bufA, bufB, pred ); } } debug( UnitTest ) { unittest { assert( mismatch( "a", "abcdefg" ) == 1 ); assert( mismatch( "abcdefg", "a" ) == 1 ); assert( mismatch( "x", "abcdefg" ) == 0 ); assert( mismatch( "abcdefg", "x" ) == 0 ); assert( mismatch( "xbcdefg", "abcdefg" ) == 0 ); assert( mismatch( "abcdefg", "xbcdefg" ) == 0 ); assert( mismatch( "abcxefg", "abcdefg" ) == 3 ); assert( mismatch( "abcdefg", "abcxefg" ) == 3 ); assert( mismatch( "abcdefx", "abcdefg" ) == 6 ); assert( mismatch( "abcdefg", "abcdefx" ) == 6 ); } } } //////////////////////////////////////////////////////////////////////////////// // Count //////////////////////////////////////////////////////////////////////////////// version( DDoc ) { /** * Performs a linear scan of buf from $(LB)0 .. buf.length$(RP), returning * a count of the number of elements matching pat. Comparisons will be * performed using the supplied predicate or '==' if none is supplied. * * Params: * buf = The array to scan. * pat = The pattern to match. * pred = The evaluation predicate, which should return true if e1 is * equal to e2 and false if not. This predicate may be any * callable type. * * Returns: * The number of elements matching pat. */ size_t count( Elem[] buf, Elem pat, Pred2E pred = Pred2E.init ); } else { template count_( Elem, Pred = IsEqual!(Elem) ) { static assert( isCallableType!(Pred) ); size_t fn( Elem[] buf, Elem pat, Pred pred = Pred.init ) { size_t cnt = 0; foreach( size_t pos, Elem cur; buf ) { if( pred( cur, pat ) ) ++cnt; } return cnt; } } template count( Buf, Pat ) { size_t count( Buf buf, Pat pat ) { return count_!(ElemTypeOf!(Buf)).fn( buf, pat ); } } template count( Buf, Pat, Pred ) { size_t count( Buf buf, Pat pat, Pred pred ) { return count_!(ElemTypeOf!(Buf), Pred).fn( buf, pat, pred ); } } debug( UnitTest ) { unittest { assert( count( "gbbbi", 'a' ) == 0 ); assert( count( "gbbbi", 'g' ) == 1 ); assert( count( "gbbbi", 'b' ) == 3 ); assert( count( "gbbbi", 'i' ) == 1 ); assert( count( "gbbbi", 'd' ) == 0 ); } } } //////////////////////////////////////////////////////////////////////////////// // Count-If //////////////////////////////////////////////////////////////////////////////// version( DDoc ) { /** * Performs a linear scan of buf from $(LB)0 .. buf.length$(RP), returning * a count of the number of elements where pred returns true. * * Params: * buf = The array to scan. * pred = The evaluation predicate, which should return true if the * element is a valid match and false if not. This predicate * may be any callable type. * * Returns: * The number of elements where pred returns true. */ size_t countIf( Elem[] buf, Pred1E pred = Pred1E.init ); } else { template countIf_( Elem, Pred ) { static assert( isCallableType!(Pred) ); size_t fn( Elem[] buf, Pred pred ) { size_t cnt = 0; foreach( size_t pos, Elem cur; buf ) { if( pred( cur ) ) ++cnt; } return cnt; } } template countIf( Buf, Pred ) { size_t countIf( Buf buf, Pred pred ) { return countIf_!(ElemTypeOf!(Buf), Pred).fn( buf, pred ); } } debug( UnitTest ) { unittest { assert( countIf( "gbbbi", ( char c ) { return c == 'a'; } ) == 0 ); assert( countIf( "gbbbi", ( char c ) { return c == 'g'; } ) == 1 ); assert( countIf( "gbbbi", ( char c ) { return c == 'b'; } ) == 3 ); assert( countIf( "gbbbi", ( char c ) { return c == 'i'; } ) == 1 ); assert( countIf( "gbbbi", ( char c ) { return c == 'd'; } ) == 0 ); } } } //////////////////////////////////////////////////////////////////////////////// // Replace //////////////////////////////////////////////////////////////////////////////// version( DDoc ) { /** * Performs a linear scan of buf from $(LB)0 .. buf.length$(RP), replacing * occurrences of pat with val. Comparisons will be performed using the * supplied predicate or '==' if none is supplied. * * Params: * buf = The array to scan. * pat = The pattern to match. * val = The value to substitute. * pred = The evaluation predicate, which should return true if e1 is * equal to e2 and false if not. This predicate may be any * callable type. * * Returns: * The number of elements replaced. */ size_t replace( Elem[] buf, Elem pat, Elem val, Pred2E pred = Pred2E.init ); } else { template replace_( Elem, Pred = IsEqual!(Elem) ) { static assert( isCallableType!(Pred) ); size_t fn( Elem[] buf, Elem pat, Elem val, Pred pred = Pred.init ) { size_t cnt = 0; foreach( size_t pos, inout Elem cur; buf ) { if( pred( cur, pat ) ) { cur = val; ++cnt; } } return cnt; } } template replace( Buf, Elem ) { size_t replace( Buf buf, Elem pat, Elem val ) { return replace_!(ElemTypeOf!(Buf)).fn( buf, pat, val ); } } template replace( Buf, Elem, Pred ) { size_t replace( Buf buf, Elem pat, Elem val, Pred pred ) { return replace_!(ElemTypeOf!(Buf), Pred).fn( buf, pat, val, pred ); } } debug( UnitTest ) { unittest { assert( replace( "gbbbi".dup, 'a', 'b' ) == 0 ); assert( replace( "gbbbi".dup, 'g', 'h' ) == 1 ); assert( replace( "gbbbi".dup, 'b', 'c' ) == 3 ); assert( replace( "gbbbi".dup, 'i', 'j' ) == 1 ); assert( replace( "gbbbi".dup, 'd', 'e' ) == 0 ); } } } //////////////////////////////////////////////////////////////////////////////// // Replace-If //////////////////////////////////////////////////////////////////////////////// version( DDoc ) { /** * Performs a linear scan of buf from $(LB)0 .. buf.length$(RP), replacing * elements where pred returns true with val. * * Params: * buf = The array to scan. * val = The value to substitute. * pred = The evaluation predicate, which should return true if the * element is a valid match and false if not. This predicate * may be any callable type. * * Returns: * The number of elements replaced. */ size_t replaceIf( Elem[] buf, Elem val, Pred2E pred = Pred2E.init ); } else { template replaceIf_( Elem, Pred ) { static assert( isCallableType!(Pred) ); size_t fn( Elem[] buf, Elem val, Pred pred ) { size_t cnt = 0; foreach( size_t pos, inout Elem cur; buf ) { if( pred( cur ) ) { cur = val; ++cnt; } } return cnt; } } template replaceIf( Buf, Elem, Pred ) { size_t replaceIf( Buf buf, Elem val, Pred pred ) { return replaceIf_!(ElemTypeOf!(Buf), Pred).fn( buf, val, pred ); } } debug( UnitTest ) { unittest { assert( replaceIf( "gbbbi".dup, 'b', ( char c ) { return c == 'a'; } ) == 0 ); assert( replaceIf( "gbbbi".dup, 'h', ( char c ) { return c == 'g'; } ) == 1 ); assert( replaceIf( "gbbbi".dup, 'c', ( char c ) { return c == 'b'; } ) == 3 ); assert( replaceIf( "gbbbi".dup, 'j', ( char c ) { return c == 'i'; } ) == 1 ); assert( replaceIf( "gbbbi".dup, 'e', ( char c ) { return c == 'd'; } ) == 0 ); } } } //////////////////////////////////////////////////////////////////////////////// // Remove //////////////////////////////////////////////////////////////////////////////// version( DDoc ) { /** * Performs a linear scan of buf from $(LB)0 .. buf.length$(RP), moving all * elements matching pat to the end of the sequence. The relative order of * elements not matching pat will be preserved. Comparisons will be * performed using the supplied predicate or '==' if none is supplied. * * Params: * buf = The array to scan. This parameter is not marked 'inout' * to allow temporary slices to be modified. As buf is not resized * in any way, omitting the 'inout' qualifier has no effect on the * result of this operation, even though it may be viewed as a * side-effect. * pat = The pattern to match against. * pred = The evaluation predicate, which should return true if e1 is * equal to e2 and false if not. This predicate may be any * callable type. * * Returns: * The number of elements that do not match pat. */ size_t remove( Elem[] buf, Elem pat, Pred2E pred = Pred2E.init ); } else { template remove_( Elem, Pred = IsEqual!(Elem) ) { static assert( isCallableType!(Pred) ); size_t fn( Elem[] buf, Elem pat, Pred pred = Pred.init ) { // NOTE: Indexes are passed instead of references because DMD does // not inline the reference-based version. void exch( size_t p1, size_t p2 ) { Elem t = buf[p1]; buf[p1] = buf[p2]; buf[p2] = t; } size_t cnt = 0; for( size_t pos = 0, len = buf.length; pos < len; ++pos ) { if( pred( buf[pos], pat ) ) ++cnt; else exch( pos, pos - cnt ); } return buf.length - cnt; } } template remove( Buf, Pat ) { size_t remove( Buf buf, Pat pat ) { return remove_!(ElemTypeOf!(Buf)).fn( buf, pat ); } } template remove( Buf, Pat, Pred ) { size_t remove( Buf buf, Pat pat, Pred pred ) { return remove_!(ElemTypeOf!(Buf), Pred).fn( buf, pat, pred ); } } debug( UnitTest ) { unittest { void test( char[] buf, char pat, size_t num ) { assert( remove( buf, pat ) == num ); foreach( pos, cur; buf ) { assert( pos < num ? cur != pat : cur == pat ); } } test( "abcdefghij".dup, 'x', 10 ); test( "xabcdefghi".dup, 'x', 9 ); test( "abcdefghix".dup, 'x', 9 ); test( "abxxcdefgh".dup, 'x', 8 ); test( "xaxbcdxxex".dup, 'x', 5 ); } } } //////////////////////////////////////////////////////////////////////////////// // Remove-If //////////////////////////////////////////////////////////////////////////////// version( DDoc ) { /** * Performs a linear scan of buf from $(LB)0 .. buf.length$(RP), moving all * elements that satisfy pred to the end of the sequence. The relative * order of elements that do not satisfy pred will be preserved. * * Params: * buf = The array to scan. This parameter is not marked 'inout' * to allow temporary slices to be modified. As buf is not resized * in any way, omitting the 'inout' qualifier has no effect on the * result of this operation, even though it may be viewed as a * side-effect. * pred = The evaluation predicate, which should return true if the * element satisfies the condition and false if not. This * predicate may be any callable type. * * Returns: * The number of elements that do not satisfy pred. */ size_t removeIf( Elem[] buf, Pred1E pred ); } else { template removeIf_( Elem, Pred ) { static assert( isCallableType!(Pred) ); size_t fn( Elem[] buf, Pred pred ) { // NOTE: Indexes are passed instead of references because DMD does // not inline the reference-based version. void exch( size_t p1, size_t p2 ) { Elem t = buf[p1]; buf[p1] = buf[p2]; buf[p2] = t; } size_t cnt = 0; for( size_t pos = 0, len = buf.length; pos < len; ++pos ) { if( pred( buf[pos] ) ) ++cnt; else exch( pos, pos - cnt ); } return buf.length - cnt; } } template removeIf( Buf, Pred ) { size_t removeIf( Buf buf, Pred pred ) { return removeIf_!(ElemTypeOf!(Buf), Pred).fn( buf, pred ); } } debug( UnitTest ) { unittest { void test( char[] buf, bool delegate( char ) dg, size_t num ) { assert( removeIf( buf, dg ) == num ); foreach( pos, cur; buf ) { assert( pos < num ? !dg( cur ) : dg( cur ) ); } } test( "abcdefghij".dup, ( char c ) { return c == 'x'; }, 10 ); test( "xabcdefghi".dup, ( char c ) { return c == 'x'; }, 9 ); test( "abcdefghix".dup, ( char c ) { return c == 'x'; }, 9 ); test( "abxxcdefgh".dup, ( char c ) { return c == 'x'; }, 8 ); test( "xaxbcdxxex".dup, ( char c ) { return c == 'x'; }, 5 ); } } } //////////////////////////////////////////////////////////////////////////////// // Unique //////////////////////////////////////////////////////////////////////////////// version( DDoc ) { /** * Performs a linear scan of buf from $(LB)0 .. buf.length$(RP), moving all * but the first element of each consecutive group of duplicate elements to * the end of the sequence. The relative order of all remaining elements * will be preserved. Comparisons will be performed using the supplied * predicate or '==' if none is supplied. * * Params: * buf = The array to scan. This parameter is not marked 'inout' * to allow temporary slices to be modified. As buf is not resized * in any way, omitting the 'inout' qualifier has no effect on the * result of this operation, even though it may be viewed as a * side-effect. * pred = The evaluation predicate, which should return true if e1 is * equal to e2 and false if not. This predicate may be any * callable type. * * Returns: * The number of unique elements in buf. */ size_t unique( Elem[] buf, Pred2E pred = Pred2E.init ); } else { template unique_( Elem, Pred = IsEqual!(Elem) ) { static assert( isCallableType!(Pred) ); size_t fn( Elem[] buf, Pred pred = Pred.init ) { // NOTE: Indexes are passed instead of references because DMD does // not inline the reference-based version. void exch( size_t p1, size_t p2 ) { Elem t = buf[p1]; buf[p1] = buf[p2]; buf[p2] = t; } if( buf.length < 2 ) return buf.length; size_t cnt = 0; Elem pat = buf[0]; for( size_t pos = 1, len = buf.length; pos < len; ++pos ) { if( pred( buf[pos], pat ) ) ++cnt; else { pat = buf[pos]; exch( pos, pos - cnt ); } } return buf.length - cnt; } } template unique( Buf ) { size_t unique( Buf buf ) { return unique_!(ElemTypeOf!(Buf)).fn( buf ); } } template unique( Buf, Pred ) { size_t unique( Buf buf, Pred pred ) { return unique_!(ElemTypeOf!(Buf), Pred).fn( buf, pred ); } } debug( UnitTest ) { unittest { void test( char[] buf, char[] pat ) { assert( unique( buf ) == pat.length ); foreach( pos, cur; pat ) { assert( buf[pos] == cur ); } } test( "abcdefghij".dup, "abcdefghij" ); test( "aabcdefghi".dup, "abcdefghi" ); test( "bcdefghijj".dup, "bcdefghij" ); test( "abccdefghi".dup, "abcdefghi" ); test( "abccdddefg".dup, "abcdefg" ); } } } //////////////////////////////////////////////////////////////////////////////// // Partition //////////////////////////////////////////////////////////////////////////////// version( DDoc ) { /** * Partitions buf such that all elements that satisfy pred will be placed * before the elements that do not satisfy pred. The algorithm is not * required to be stable. * * Params: * buf = The array to partition. This parameter is not marked 'inout' * to allow temporary slices to be sorted. As buf is not resized * in any way, omitting the 'inout' qualifier has no effect on * the result of this operation, even though it may be viewed * as a side-effect. * pred = The evaluation predicate, which should return true if the * element satisfies the condition and false if not. This * predicate may be any callable type. * * Returns: * The number of elements that satisfy pred. */ size_t partition( Elem[] buf, Pred1E pred ); } else { template partition_( Elem, Pred = IsLess!(Elem) ) { static assert( isCallableType!(Pred ) ); size_t fn( Elem[] buf, Pred pred ) { // NOTE: Indexes are passed instead of references because DMD does // not inline the reference-based version. void exch( size_t p1, size_t p2 ) { Elem t = buf[p1]; buf[p1] = buf[p2]; buf[p2] = t; } if( buf.length < 2 ) return 0; size_t l = 0, r = buf.length, i = l, j = r - 1; while( true ) { while( i < r && pred( buf[i] ) ) ++i; while( j > l && !pred( buf[j] ) ) --j; if( i >= j ) break; exch( i++, j-- ); } return i; } } template partition( Buf, Pred ) { size_t partition( Buf buf, Pred pred ) { return partition_!(ElemTypeOf!(Buf), Pred).fn( buf, pred ); } } debug( UnitTest ) { unittest { void test( char[] buf, bool delegate( char ) dg, size_t num ) { assert( partition( buf, dg ) == num ); for( size_t pos = 0; pos < buf.length; ++pos ) { assert( pos < num ? dg( buf[pos] ) : !dg( buf[pos] ) ); } } test( "abcdefg".dup, ( char c ) { return c < 'a'; }, 0 ); test( "gfedcba".dup, ( char c ) { return c < 'a'; }, 0 ); test( "abcdefg".dup, ( char c ) { return c < 'h'; }, 7 ); test( "gfedcba".dup, ( char c ) { return c < 'h'; }, 7 ); test( "abcdefg".dup, ( char c ) { return c < 'd'; }, 3 ); test( "gfedcba".dup, ( char c ) { return c < 'd'; }, 3 ); test( "bbdaabc".dup, ( char c ) { return c < 'c'; }, 5 ); } } } //////////////////////////////////////////////////////////////////////////////// // Select //////////////////////////////////////////////////////////////////////////////// version( DDoc ) { /** * Partitions buf with num - 1 as a pivot such that the first num elements * will be less than or equal to the remaining elements in the array. * Comparisons will be performed using the supplied predicate or '<' if * none is supplied. The algorithm is not required to be stable. * * Params: * buf = The array to partition. This parameter is not marked 'inout' * to allow temporary slices to be sorted. As buf is not resized * in any way, omitting the 'inout' qualifier has no effect on * the result of this operation, even though it may be viewed * as a side-effect. * num = The number of elements which are considered significant in * this array, where num - 1 is the pivot around which partial * sorting will occur. For example, if num is buf.length / 2 * then select will effectively partition the array around its * median value, with the elements in the first half of the array * evaluating as less than or equal to the elements in the second * half. * pred = The evaluation predicate, which should return true if e1 is * less than e2 and false if not. This predicate may be any * callable type. * * Returns: * The index of the pivot point, which will be the lesser of num - 1 and * buf.length. */ size_t select( Elem[] buf, Num num, Pred2E pred = Pred2E.init ); } else { template select_( Elem, Pred = IsLess!(Elem) ) { static assert( isCallableType!(Pred ) ); size_t fn( Elem[] buf, size_t num, Pred pred = Pred.init ) { // NOTE: Indexes are passed instead of references because DMD does // not inline the reference-based version. void exch( size_t p1, size_t p2 ) { Elem t = buf[p1]; buf[p1] = buf[p2]; buf[p2] = t; } if( buf.length < 2 ) return buf.length; size_t l = 0, r = buf.length - 1, k = num; while( r > l ) { size_t i = l, j = r - 1; Elem v = buf[r]; while( true ) { while( i < r && pred( buf[i], v ) ) ++i; while( j > l && pred( v, buf[j] ) ) --j; if( i >= j ) break; exch( i++, j-- ); } exch( i, r ); if( i >= k ) r = i - 1; if( i <= k ) l = i + 1; } return num - 1; } } template select( Buf, Num ) { size_t select( Buf buf, Num num ) { return select_!(ElemTypeOf!(Buf)).fn( buf, num ); } } template select( Buf, Num, Pred ) { size_t select( Buf buf, Num num, Pred pred ) { return select_!(ElemTypeOf!(Buf), Pred).fn( buf, num, pred ); } } debug( UnitTest ) { unittest { char[] buf = "efedcaabca".dup; size_t num = buf.length / 2; size_t pos = select( buf, num ); assert( pos == num - 1 ); foreach( cur; buf[0 .. pos] ) assert( cur <= buf[pos] ); foreach( cur; buf[pos .. $] ) assert( cur >= buf[pos] ); } } } //////////////////////////////////////////////////////////////////////////////// // Sort //////////////////////////////////////////////////////////////////////////////// version( DDoc ) { /** * Sorts buf using the supplied predicate or '<' if none is supplied. The * algorithm is not required to be stable. The current implementation is * based on quicksort, but uses a three-way partitioning scheme to improve * performance for ranges containing duplicate values (Bentley and McIlroy, * 1993). * * Params: * buf = The array to sort. This parameter is not marked 'inout' to * allow temporary slices to be sorted. As buf is not resized * in any way, omitting the 'inout' qualifier has no effect on * the result of this operation, even though it may be viewed * as a side-effect. * pred = The evaluation predicate, which should return true if e1 is * less than e2 and false if not. This predicate may be any * callable type. */ void sort( Elem[] buf, Pred2E pred = Pred2E.init ); } else { template sort_( Elem, Pred = IsLess!(Elem) ) { static assert( isCallableType!(Pred ) ); void fn( Elem[] buf, Pred pred = Pred.init ) { bool equiv( Elem p1, Elem p2 ) { return !pred( p1, p2 ) && !pred( p2, p1 ); } // NOTE: Indexes are passed instead of references because DMD does // not inline the reference-based version. void exch( size_t p1, size_t p2 ) { Elem t = buf[p1]; buf[p1] = buf[p2]; buf[p2] = t; } void quicksort( size_t l, size_t r ) { if( r <= l ) return; // This implementation of quicksort improves upon the classic // algorithm by partitioning the array into three parts, one // each for keys smaller than, equal to, and larger than the // partitioning element, v: // // |--less than v--|--equal to v--|--greater than v--[v] // l j i r // // This approach sorts ranges containing duplicate elements // more quickly. During processing, the following situation // is maintained: // // |--equal--|--less--|--[###]--|--greater--|--equal--[v] // l p i j q r // // Please note that this implementation varies from the typical // algorithm by replacing the use of signed index values with // unsigned values. Elem v = buf[r]; size_t i = l, j = r, p = l, q = r; while( true ) { while( pred( buf[i], v ) ) ++i; while( pred( v, buf[--j] ) ) if( j == l ) break; if( i >= j ) break; exch( i, j ); if( equiv( buf[i], v ) ) exch( p++, i ); if( equiv( v, buf[j] ) ) exch( --q, j ); ++i; } exch( i, r ); if( p < i ) { j = i - 1; for( size_t k = l; k < p; k++, j-- ) exch( k, j ); quicksort( l, j ); } if( ++i < q ) { for( size_t k = r - 1; k >= q; k--, i++ ) exch( k, i ); quicksort( i, r ); } } if( buf.length > 1 ) { quicksort( 0, buf.length - 1 ); } } } template sort( Buf ) { void sort( Buf buf ) { return sort_!(ElemTypeOf!(Buf)).fn( buf ); } } template sort( Buf, Pred ) { void sort( Buf buf, Pred pred ) { return sort_!(ElemTypeOf!(Buf), Pred).fn( buf, pred ); } } debug( UnitTest ) { unittest { void test( char[] buf ) { sort( buf ); char sav = buf[0]; foreach( cur; buf ) { assert( cur >= sav ); sav = cur; } } test( "mkcvalsidivjoaisjdvmzlksvdjioawmdsvmsdfefewv".dup ); test( "asdfasdfasdfasdfasdfasdfasdfasdfasdfasdfasdf".dup ); test( "the quick brown fox jumped over the lazy dog".dup ); test( "abcdefghijklmnopqrstuvwxyz".dup ); test( "zyxwvutsrqponmlkjihgfedcba".dup ); } } } //////////////////////////////////////////////////////////////////////////////// // Lower Bound //////////////////////////////////////////////////////////////////////////////// version( DDoc ) { /** * Performs a binary search of buf, returning the index of the first * location where pat may be inserted without disrupting sort order. If * the sort order of pat precedes all elements in buf then 0 will be * returned. If the sort order of pat succeeds the largest element in buf * then buf.length will be returned. Comparisons will be performed using * the supplied predicate or '<' if none is supplied. * * Params: * buf = The sorted array to search. * pat = The pattern to search for. * pred = The evaluation predicate, which should return true if e1 is * less than e2 and false if not. This predicate may be any * callable type. * * Returns: * The index of the first match or buf.length if no match was found. */ size_t lbound( Elem[] buf, Elem pat, Pred2E pred = Pred2E.init ); } else { template lbound_( Elem, Pred = IsLess!(Elem) ) { static assert( isCallableType!(Pred) ); size_t fn( Elem[] buf, Elem pat, Pred pred = Pred.init ) { size_t beg = 0, end = buf.length, mid = end / 2; while( beg < end ) { if( pred( buf[mid], pat ) ) beg = mid + 1; else end = mid; mid = beg + ( end - beg ) / 2; } return mid; } } template lbound( Buf, Pat ) { size_t lbound( Buf buf, Pat pat ) { return lbound_!(ElemTypeOf!(Buf)).fn( buf, pat ); } } template lbound( Buf, Pat, Pred ) { size_t lbound( Buf buf, Pat pat, Pred pred ) { return lbound_!(ElemTypeOf!(Buf), Pred).fn( buf, pat, pred ); } } debug( UnitTest ) { unittest { assert( lbound( "bcefg", 'a' ) == 0 ); assert( lbound( "bcefg", 'h' ) == 5 ); assert( lbound( "bcefg", 'd' ) == 2 ); assert( lbound( "bcefg", 'e' ) == 2 ); } } } //////////////////////////////////////////////////////////////////////////////// // Upper Bound //////////////////////////////////////////////////////////////////////////////// version( DDoc ) { /** * Performs a binary search of buf, returning the index of the first * location beyond where pat may be inserted without disrupting sort order. * If the sort order of pat precedes all elements in buf then 0 will be * returned. If the sort order of pat succeeds the largest element in buf * then buf.length will be returned. Comparisons will be performed using * the supplied predicate or '<' if none is supplied. * * Params: * buf = The sorted array to search. * pat = The pattern to search for. * pred = The evaluation predicate, which should return true if e1 is * less than e2 and false if not. This predicate may be any * callable type. * * Returns: * The index of the first match or buf.length if no match was found. */ size_t ubound( Elem[] buf, Elem pat, Pred2E pred = Pred2E.init ); } else { template ubound_( Elem, Pred = IsLess!(Elem) ) { static assert( isCallableType!(Pred) ); size_t fn( Elem[] buf, Elem pat, Pred pred = Pred.init ) { size_t beg = 0, end = buf.length, mid = end / 2; while( beg < end ) { if( !pred( pat, buf[mid] ) ) beg = mid + 1; else end = mid; mid = beg + ( end - beg ) / 2; } return mid; } } template ubound( Buf, Pat ) { size_t ubound( Buf buf, Pat pat ) { return ubound_!(ElemTypeOf!(Buf)).fn( buf, pat ); } } template ubound( Buf, Pat, Pred ) { size_t ubound( Buf buf, Pat pat, Pred pred ) { return ubound_!(ElemTypeOf!(Buf), Pred).fn( buf, pat, pred ); } } debug( UnitTest ) { unittest { assert( ubound( "bcefg", 'a' ) == 0 ); assert( ubound( "bcefg", 'h' ) == 5 ); assert( ubound( "bcefg", 'd' ) == 2 ); assert( ubound( "bcefg", 'e' ) == 3 ); } } } //////////////////////////////////////////////////////////////////////////////// // Binary Search //////////////////////////////////////////////////////////////////////////////// version( DDoc ) { /** * Performs a binary search of buf, returning true if an element equivalent * to pat is found. Comparisons will be performed using the supplied * predicate or '<' if none is supplied. * * Params: * buf = The sorted array to search. * pat = The pattern to search for. * pred = The evaluation predicate, which should return true if e1 is * less than e2 and false if not. This predicate may be any * callable type. * * Returns: * True if an element equivalent to pat is found, false if not. */ bool bsearch( Elem[] buf, Elem pat, Pred2E pred = Pred2E.init ); } else { template bsearch_( Elem, Pred = IsLess!(Elem) ) { static assert( isCallableType!(Pred) ); bool fn( Elem[] buf, Elem pat, Pred pred = Pred.init ) { size_t pos = lbound( buf, pat, pred ); return pos < buf.length && !( pat < buf[pos] ); } } template bsearch( Buf, Pat ) { bool bsearch( Buf buf, Pat pat ) { return bsearch_!(ElemTypeOf!(Buf)).fn( buf, pat ); } } template bsearch( Buf, Pat, Pred ) { bool bsearch( Buf buf, Pat pat, Pred pred ) { return bsearch_!(ElemTypeOf!(Buf), Pred).fn( buf, pat, pred ); } } debug( UnitTest ) { unittest { assert( !bsearch( "bcefg", 'a' ) ); assert( bsearch( "bcefg", 'b' ) ); assert( bsearch( "bcefg", 'c' ) ); assert( !bsearch( "bcefg", 'd' ) ); assert( bsearch( "bcefg", 'e' ) ); assert( bsearch( "bcefg", 'f' ) ); assert( bsearch( "bcefg", 'g' ) ); assert( !bsearch( "bcefg", 'h' ) ); } } } //////////////////////////////////////////////////////////////////////////////// // Includes //////////////////////////////////////////////////////////////////////////////// version( DDoc ) { /** * Performs a parallel linear scan of setA and setB from $(LB)0 .. N$(RP) * where N = min$(LP)setA.length, setB.length$(RP), returning true if setA * includes all elements in setB and false if not. Both setA and setB are * required to be sorted, and duplicates in setB require an equal number of * duplicates in setA. Comparisons will be performed using the supplied * predicate or '<' if none is supplied. * * Params: * setA = The sorted array to evaluate. * setB = The sorted array to match against. * pred = The evaluation predicate, which should return true if e1 is * less than e2 and false if not. This predicate may be any * callable type. * * Returns: * true if setA includes all elements in setB, false if not. */ bool includes( Elem[] setA, Elem[] setB, Pred2E pred = Pred2E.init ); } else { template includes_( Elem, Pred = IsLess!(Elem) ) { static assert( isCallableType!(Pred ) ); bool fn( Elem[] setA, Elem[] setB, Pred pred = Pred.init ) { size_t posA = 0, posB = 0; while( posA < setA.length && posB < setB.length ) { if( pred( setB[posB], setA[posA] ) ) return false; else if( pred( setA[posA], setB[posB] ) ) ++posA; else ++posA, ++posB; } return posB == setB.length; } } template includes( BufA, BufB ) { bool includes( BufA setA, BufB setB ) { return includes_!(ElemTypeOf!(BufA)).fn( setA, setB ); } } template includes( BufA, BufB, Pred ) { bool includes( BufA setA, BufB setB, Pred pred ) { return includes_!(ElemTypeOf!(BufA), Pred).fn( setA, setB, pred ); } } debug( UnitTest ) { unittest { assert( includes( "abcdefg", "a" ) ); assert( includes( "abcdefg", "g" ) ); assert( includes( "abcdefg", "d" ) ); assert( includes( "abcdefg", "abcdefg" ) ); assert( includes( "aaaabbbcdddefgg", "abbbcdefg" ) ); assert( !includes( "abcdefg", "aaabcdefg" ) ); assert( !includes( "abcdefg", "abcdefggg" ) ); assert( !includes( "abbbcdefg", "abbbbcdefg" ) ); } } } //////////////////////////////////////////////////////////////////////////////// // Union Of //////////////////////////////////////////////////////////////////////////////// version( DDoc ) { /** * Computes the union of setA and setB as a set operation and returns the * retult in a new sorted array. Both setA and setB are required to be * sorted. If either setA or setB contain duplicates, the result will * contain the larger number of duplicates from setA and setB. When an * overlap occurs, entries will be copied from setA. Comparisons will be * performed using the supplied predicate or '<' if none is supplied. * * Params: * setA = The first sorted array to evaluate. * setB = The second sorted array to evaluate. * pred = The evaluation predicate, which should return true if e1 is * less than e2 and false if not. This predicate may be any * callable type. * * Returns: * A new array containing the union of setA and setB. */ Elem[] unionOf( Elem[] setA, Elem[] setB, Pred2E pred = Pred2E.init ); } else { template unionOf_( Elem, Pred = IsLess!(Elem) ) { static assert( isCallableType!(Pred ) ); Elem[] fn( Elem[] setA, Elem[] setB, Pred pred = Pred.init ) { size_t posA = 0, posB = 0; Elem[] setU; while( posA < setA.length && posB < setB.length ) { if( pred( setA[posA], setB[posB] ) ) setU ~= setA[posA++]; else if( pred( setB[posB], setA[posA] ) ) setU ~= setB[posB++]; else setU ~= setA[posA++], posB++; } setU ~= setA[posA .. $]; setU ~= setB[posB .. $]; return setU; } } template unionOf( BufA, BufB ) { ElemTypeOf!(BufA)[] unionOf( BufA setA, BufB setB ) { return unionOf_!(ElemTypeOf!(BufA)).fn( setA, setB ); } } template unionOf( BufA, BufB, Pred ) { ElemTypeOf!(BufA)[] unionOf( BufA setA, BufB setB, Pred pred ) { return unionOf_!(ElemTypeOf!(BufA), Pred).fn( setA, setB, pred ); } } debug( UnitTest ) { unittest { assert( unionOf( "", "" ) == "" ); assert( unionOf( "abc", "def" ) == "abcdef" ); assert( unionOf( "abbbcd", "aadeefg" ) == "aabbbcdeefg" ); } } } //////////////////////////////////////////////////////////////////////////////// // Intersection Of //////////////////////////////////////////////////////////////////////////////// version( DDoc ) { /** * Computes the intersection of setA and setB as a set operation and * returns the retult in a new sorted array. Both setA and setB are * required to be sorted. If either setA or setB contain duplicates, the * result will contain the smaller number of duplicates from setA and setB. * All entries will be copied from setA. Comparisons will be performed * using the supplied predicate or '<' if none is supplied. * * Params: * setA = The first sorted array to evaluate. * setB = The second sorted array to evaluate. * pred = The evaluation predicate, which should return true if e1 is * less than e2 and false if not. This predicate may be any * callable type. * * Returns: * A new array containing the intersection of setA and setB. */ Elem[] intersectionOf( Elem[] setA, Elem[] setB, Pred2E pred = Pred2E.init ); } else { template intersectionOf_( Elem, Pred = IsLess!(Elem) ) { static assert( isCallableType!(Pred ) ); Elem[] fn( Elem[] setA, Elem[] setB, Pred pred = Pred.init ) { size_t posA = 0, posB = 0; Elem[] setI; while( posA < setA.length && posB < setB.length ) { if( pred( setA[posA], setB[posB] ) ) ++posA; else if( pred( setB[posB], setA[posA] ) ) ++posB; else setI ~= setA[posA++], posB++; } return setI; } } template intersectionOf( BufA, BufB ) { ElemTypeOf!(BufA)[] intersectionOf( BufA setA, BufB setB ) { return intersectionOf_!(ElemTypeOf!(BufA)).fn( setA, setB ); } } template intersectionOf( BufA, BufB, Pred ) { ElemTypeOf!(BufA)[] intersectionOf( BufA setA, BufB setB, Pred pred ) { return intersectionOf_!(ElemTypeOf!(BufA), Pred).fn( setA, setB, pred ); } } debug( UnitTest ) { unittest { assert( intersectionOf( "", "" ) == "" ); assert( intersectionOf( "abc", "def" ) == "" ); assert( intersectionOf( "abbbcd", "aabdddeefg" ) == "abd" ); } } } //////////////////////////////////////////////////////////////////////////////// // Missing From //////////////////////////////////////////////////////////////////////////////// version( DDoc ) { /** * Returns a new array containing all elements in setA which are not * present in setB. Both setA and setB are required to be sorted. * Comparisons will be performed using the supplied predicate or '<' * if none is supplied. * * Params: * setA = The first sorted array to evaluate. * setB = The second sorted array to evaluate. * pred = The evaluation predicate, which should return true if e1 is * less than e2 and false if not. This predicate may be any * callable type. * * Returns: * A new array containing the elements in setA that are not in setB. */ Elem[] missingFrom( Elem[] setA, Elem[] setB, Pred2E pred = Pred2E.init ); } else { template missingFrom_( Elem, Pred = IsLess!(Elem) ) { static assert( isCallableType!(Pred ) ); Elem[] fn( Elem[] setA, Elem[] setB, Pred pred = Pred.init ) { size_t posA = 0, posB = 0; Elem[] setM; while( posA < setA.length && posB < setB.length ) { if( pred( setA[posA], setB[posB] ) ) setM ~= setA[posA++]; else if( pred( setB[posB], setA[posA] ) ) ++posB; else ++posA, ++posB; } setM ~= setA[posA .. $]; return setM; } } template missingFrom( BufA, BufB ) { ElemTypeOf!(BufA)[] missingFrom( BufA setA, BufB setB ) { return missingFrom_!(ElemTypeOf!(BufA)).fn( setA, setB ); } } template missingFrom( BufA, BufB, Pred ) { ElemTypeOf!(BufA)[] missingFrom( BufA setA, BufB setB, Pred pred ) { return missingFrom_!(ElemTypeOf!(BufA), Pred).fn( setA, setB, pred ); } } debug( UnitTest ) { unittest { assert( missingFrom( "", "" ) == "" ); assert( missingFrom( "", "abc" ) == "" ); assert( missingFrom( "abc", "" ) == "abc" ); assert( missingFrom( "abc", "abc" ) == "" ); assert( missingFrom( "abc", "def" ) == "abc" ); assert( missingFrom( "abbbcd", "abd" ) == "bbc" ); assert( missingFrom( "abcdef", "bc" ) == "adef" ); } } } //////////////////////////////////////////////////////////////////////////////// // Difference Of //////////////////////////////////////////////////////////////////////////////// version( DDoc ) { /** * Returns a new array containing all elements in setA which are not * present in setB and the elements in setB which are not present in * setA. Both setA and setB are required to be sorted. Comparisons * will be performed using the supplied predicate or '<' if none is * supplied. * * Params: * setA = The first sorted array to evaluate. * setB = The second sorted array to evaluate. * pred = The evaluation predicate, which should return true if e1 is * less than e2 and false if not. This predicate may be any * callable type. * * Returns: * A new array containing the elements in setA that are not in setB * and the elements in setB that are not in setA. */ Elem[] differenceOf( Elem[] setA, Elem[] setB, Pred2E pred = Pred2E.init ); } else { template differenceOf_( Elem, Pred = IsLess!(Elem) ) { static assert( isCallableType!(Pred ) ); Elem[] fn( Elem[] setA, Elem[] setB, Pred pred = Pred.init ) { size_t posA = 0, posB = 0; Elem[] setD; while( posA < setA.length && posB < setB.length ) { if( pred( setA[posA], setB[posB] ) ) setD ~= setA[posA++]; else if( pred( setB[posB], setA[posA] ) ) setD ~= setB[posB++]; else ++posA, ++posB; } setD ~= setA[posA .. $]; setD ~= setB[posB .. $]; return setD; } } template differenceOf( BufA, BufB ) { ElemTypeOf!(BufA)[] differenceOf( BufA setA, BufB setB ) { return differenceOf_!(ElemTypeOf!(BufA)).fn( setA, setB ); } } template differenceOf( BufA, BufB, Pred ) { ElemTypeOf!(BufA)[] differenceOf( BufA setA, BufB setB, Pred pred ) { return differenceOf_!(ElemTypeOf!(BufA), Pred).fn( setA, setB, pred ); } } debug( UnitTest ) { unittest { assert( differenceOf( "", "" ) == "" ); assert( differenceOf( "", "abc" ) == "abc" ); assert( differenceOf( "abc", "" ) == "abc" ); assert( differenceOf( "abc", "abc" ) == "" ); assert( differenceOf( "abc", "def" ) == "abcdef" ); assert( differenceOf( "abbbcd", "abd" ) == "bbc" ); assert( differenceOf( "abd", "abbbcd" ) == "bbc" ); } } } //////////////////////////////////////////////////////////////////////////////// // Make Heap //////////////////////////////////////////////////////////////////////////////// version( DDoc ) { /** * Converts buf to a heap using the supplied predicate or '<' if none is * supplied. * * Params: * buf = The array to convert. This parameter is not marked 'inout' to * allow temporary slices to be sorted. As buf is not resized in * any way, omitting the 'inout' qualifier has no effect on the * result of this operation, even though it may be viewed as a * side-effect. * pred = The evaluation predicate, which should return true if e1 is * less than e2 and false if not. This predicate may be any * callable type. */ void makeHeap( Elem[] buf, Pred2E pred = Pred2E.init ); } else { template makeHeap_( Elem, Pred = IsLess!(Elem) ) { static assert( isCallableType!(Pred ) ); void fn( Elem[] buf, Pred pred = Pred.init ) { // NOTE: Indexes are passed instead of references because DMD does // not inline the reference-based version. void exch( size_t p1, size_t p2 ) { Elem t = buf[p1]; buf[p1] = buf[p2]; buf[p2] = t; } void fixDown( size_t pos, size_t end ) { if( end <= pos ) return; while( pos <= ( end - 1 ) / 2 ) { size_t nxt = 2 * pos + 1; if( nxt < end && pred( buf[nxt], buf[nxt + 1] ) ) ++nxt; if( !pred( buf[pos], buf[nxt] ) ) break; exch( pos, nxt ); pos = nxt; } } if( buf.length < 2 ) return; size_t end = buf.length - 1, pos = end / 2 + 2; do { fixDown( --pos, end ); } while( pos > 0 ); } } template makeHeap( Buf ) { void makeHeap( Buf buf ) { return makeHeap_!(ElemTypeOf!(Buf)).fn( buf ); } } template makeHeap( Buf, Pred ) { void makeHeap( Buf buf, Pred pred ) { return makeHeap_!(ElemTypeOf!(Buf), Pred).fn( buf, pred ); } } debug( UnitTest ) { unittest { void basic( char[] buf ) { if( buf.length < 2 ) return; size_t pos = 0, end = buf.length - 1; while( pos <= ( end - 1 ) / 2 ) { assert( buf[pos] >= buf[2 * pos + 1] ); if( 2 * pos + 1 < end ) { assert( buf[pos] >= buf[2 * pos + 2] ); } ++pos; } } void test( char[] buf ) { makeHeap( buf ); basic( buf ); } test( "mkcvalsidivjoaisjdvmzlksvdjioawmdsvmsdfefewv".dup ); test( "asdfasdfasdfasdfasdfasdfasdfasdfasdfasdfasdf".dup ); test( "the quick brown fox jumped over the lazy dog".dup ); test( "abcdefghijklmnopqrstuvwxyz".dup ); test( "zyxwvutsrqponmlkjihgfedcba".dup ); test( "ba".dup ); test( "a".dup ); test( "".dup ); } } } //////////////////////////////////////////////////////////////////////////////// // Push Heap //////////////////////////////////////////////////////////////////////////////// version( DDoc ) { /** * Adds val to buf by appending it and adjusting it up the heap. * * Params: * buf = The heap to modify. This parameter is marked 'inout' because * buf.length will be altered. * val = The element to push onto buf. * pred = The evaluation predicate, which should return true if e1 is * less than e2 and false if not. This predicate may be any * callable type. */ void pushHeap( inout Elem[] buf, Elem val, Pred2E pred = Pred2E.init ); } else { template pushHeap_( Elem, Pred = IsLess!(Elem) ) { static assert( isCallableType!(Pred ) ); void fn( inout Elem[] buf, Elem val, Pred pred = Pred.init ) { // NOTE: Indexes are passed instead of references because DMD does // not inline the reference-based version. void exch( size_t p1, size_t p2 ) { Elem t = buf[p1]; buf[p1] = buf[p2]; buf[p2] = t; } void fixUp( size_t pos ) { if( pos < 1 ) return; size_t par = ( pos - 1 ) / 2; while( pos > 0 && pred( buf[par], buf[pos] ) ) { exch( par, pos ); pos = par; par = ( pos - 1 ) / 2; } } buf ~= val; if( buf.length > 1 ) { fixUp( buf.length - 1 ); } } } template pushHeap( Buf, Val ) { void pushHeap( inout Buf buf, Val val ) { return pushHeap_!(ElemTypeOf!(Buf)).fn( buf, val ); } } template pushHeap( Buf, Val, Pred ) { void pushHeap( inout Buf buf, Val val, Pred pred ) { return pushHeap_!(ElemTypeOf!(Buf), Pred).fn( buf, val, pred ); } } debug( UnitTest ) { unittest { void basic( char[] buf ) { if( buf.length < 2 ) return; size_t pos = 0, end = buf.length - 1; while( pos <= ( end - 1 ) / 2 ) { assert( buf[pos] >= buf[2 * pos + 1] ); if( 2 * pos + 1 < end ) { assert( buf[pos] >= buf[2 * pos + 2] ); } ++pos; } } char[] buf; foreach( cur; "abcdefghijklmnopqrstuvwxyz" ) { pushHeap( buf, cur ); basic( buf ); } buf.length = 0; foreach( cur; "zyxwvutsrqponmlkjihgfedcba" ) { pushHeap( buf, cur ); basic( buf ); } } } } //////////////////////////////////////////////////////////////////////////////// // Pop Heap //////////////////////////////////////////////////////////////////////////////// version( DDoc ) { /** * Removes the top element from buf by swapping it with the bottom element, * adjusting it down the heap, and reducing the length of buf by one. * * Params: * buf = The heap to modify. This parameter is marked 'inout' because * buf.length will be altered. * pred = The evaluation predicate, which should return true if e1 is * less than e2 and false if not. This predicate may be any * callable type. */ void popHeap( inout Elem[] buf, Pred2E pred = Pred2E.init ); } else { template popHeap_( Elem, Pred = IsLess!(Elem) ) { static assert( isCallableType!(Pred ) ); void fn( inout Elem[] buf, Pred pred = Pred.init ) { // NOTE: Indexes are passed instead of references because DMD does // not inline the reference-based version. void exch( size_t p1, size_t p2 ) { Elem t = buf[p1]; buf[p1] = buf[p2]; buf[p2] = t; } void fixDown( size_t pos, size_t end ) { if( end <= pos ) return; while( pos <= ( end - 1 ) / 2 ) { size_t nxt = 2 * pos + 1; if( nxt < end && pred( buf[nxt], buf[nxt + 1] ) ) ++nxt; if( !pred( buf[pos], buf[nxt] ) ) break; exch( pos, nxt ); pos = nxt; } } if( buf.length > 1 ) { exch( 0, buf.length - 1 ); fixDown( 0, buf.length - 2 ); } if( buf.length > 0 ) { buf.length = buf.length - 1; } } } template popHeap( Buf ) { void popHeap( inout Buf buf ) { return popHeap_!(ElemTypeOf!(Buf)).fn( buf ); } } template popHeap( Buf, Pred ) { void popHeap( inout Buf buf, Pred pred ) { return popHeap_!(ElemTypeOf!(Buf), Pred).fn( buf, pred ); } } debug( UnitTest ) { unittest { void test( char[] buf ) { if( buf.length < 2 ) return; size_t pos = 0, end = buf.length - 1; while( pos <= ( end - 1 ) / 2 ) { assert( buf[pos] >= buf[2 * pos + 1] ); if( 2 * pos + 1 < end ) { assert( buf[pos] >= buf[2 * pos + 2] ); } ++pos; } } char[] buf = "zyxwvutsrqponmlkjihgfedcba".dup; while( buf.length > 0 ) { popHeap( buf ); test( buf ); } } } } //////////////////////////////////////////////////////////////////////////////// // Sort Heap //////////////////////////////////////////////////////////////////////////////// version( DDoc ) { /** * Sorts buf as a heap using the supplied predicate or '<' if none is * supplied. Calling makeHeap and sortHeap on an array in succession * has the effect of sorting the array using the heapsort algorithm. * * Params: * buf = The heap to sort. This parameter is not marked 'inout' to * allow temporary slices to be sorted. As buf is not resized * in any way, omitting the 'inout' qualifier has no effect on * the result of this operation, even though it may be viewed * as a side-effect. * pred = The evaluation predicate, which should return true if e1 is * less than e2 and false if not. This predicate may be any * callable type. */ void sortHeap( Elem[] buf, Pred2E pred = Pred2E.init ); } else { template sortHeap_( Elem, Pred = IsLess!(Elem) ) { static assert( isCallableType!(Pred ) ); void fn( Elem[] buf, Pred pred = Pred.init ) { // NOTE: Indexes are passed instead of references because DMD does // not inline the reference-based version. void exch( size_t p1, size_t p2 ) { Elem t = buf[p1]; buf[p1] = buf[p2]; buf[p2] = t; } void fixDown( size_t pos, size_t end ) { if( end <= pos ) return; while( pos <= ( end - 1 ) / 2 ) { size_t nxt = 2 * pos + 1; if( nxt < end && pred( buf[nxt], buf[nxt + 1] ) ) ++nxt; if( !pred( buf[pos], buf[nxt] ) ) break; exch( pos, nxt ); pos = nxt; } } if( buf.length < 2 ) return; size_t pos = buf.length - 1; while( pos > 0 ) { exch( 0, pos ); fixDown( 0, --pos ); } } } template sortHeap( Buf ) { void sortHeap( Buf buf ) { return sortHeap_!(ElemTypeOf!(Buf)).fn( buf ); } } template sortHeap( Buf, Pred ) { void sortHeap( Buf buf, Pred pred ) { return sortHeap_!(ElemTypeOf!(Buf), Pred).fn( buf, pred ); } } debug( UnitTest ) { unittest { char[] buf = "zyxwvutsrqponmlkjihgfedcba".dup; sortHeap( buf ); char sav = buf[0]; foreach( cur; buf ) { assert( cur >= sav ); sav = cur; } } } }