liblinbox-dev 1.4.2-3 / usr / include / linbox / algorithms / varprec-cra-early-multip.h

/* linbox/blackbox/rational-reconstruction-base.h
 * Copyright (C) 2009 Anna Marszalek
 *
 * Written by Anna Marszalek <aniau@astronet.pl>
 *
 * ========LICENCE========
 * This file is part of the library LinBox.
 *
  * LinBox is free software: you can redistribute it and/or modify
 * it under the terms of the  GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
 * This library is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
 * License along with this library; if not, write to the Free Software
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
 * ========LICENCE========
 */

#ifndef __LINBOX_varprec_cra_multip_single_H
#define __LINBOX_varprec_cra_multip_single_H

#include "linbox/util/timer.h"
#include <stdlib.h>
#include "linbox/integer.h"
#include "linbox/field/gmp-rational.h"
#include "linbox/solutions/methods.h"
#include <vector>
#include <utility>
#include "linbox/algorithms/cra-early-single.h"
#include "linbox/algorithms/cra-full-multip.h"
#include "linbox/algorithms/lazy-product.h"


namespace LinBox
{

	/*
	 * Implements early terminated CRA with preconditioning of the result
	 * factor | result should be given
	 * does not detect bad factors but can run until full termination [for result] in this case
	 * factor may be changed by changeFactor function, residues stored in FullMultipCRA are recomputed in this case,
	 * FullMulpitCRA should consist of vectors of size 0, but errors happen.
	 */

	//typedef Givaro::ZRing<Integer> Integers;
	//typedef Integers::Element Integer;

	template<class Domain_Type>
	struct VarPrecEarlyMultipCRA: public EarlySingleCRA<Domain_Type>, FullMultipCRA<Domain_Type> {

		typedef GMPRationalField Rationals;
		typedef Rationals::Element Quotient;

		typedef Domain_Type                     Domain;
		typedef typename Domain::Element DomainElement;
		typedef VarPrecEarlyMultipCRA<Domain> Self_t;

	protected:
		BlasVector< Givaro::ZRing<Integer> > vfactor_;
		BlasVector< Givaro::ZRing<Integer> > vmultip_;

		std::vector< unsigned long > randv;
		Integer& result(Integer &d) {return d;}; // DON'T TOUCH
	public:
		VarPrecEarlyMultipCRA(const unsigned long EARLY = DEFAULT_EARLY_TERM_THRESHOLD, const double b=0.0,
				      const BlasVector<Givaro::ZRing<Integer> >& vf = BlasVector<Givaro::ZRing<Integer> >(Givaro::ZRing<Integer>()),
				      const BlasVector<Givaro::ZRing<Integer> >& vm = BlasVector<Givaro::ZRing<Integer> >(Givaro::ZRing<Integer>())) :
			EarlySingleCRA<Domain>(EARLY), FullMultipCRA<Domain>(b), vfactor_(vf), vmultip_(vm)
		{
			for (int i=0; i < (int)vfactor_.size(); ++i) {
				if (vfactor_[(size_t)i]==0) vfactor_[(size_t)i]=1;
			}
		}

		VarPrecEarlyMultipCRA(VarPrecEarlyMultipCRA& other) :
			EarlySingleCRA<Domain>(other.EARLY_TERM_THRESHOLD), FullMultipCRA<Domain>(other.LOGARITHMIC_UPPER_BOUND), vfactor_(other.vfactor_), vmultip_(other.vmultip_)
		{
			for (int i=0; i < vfactor_.size(); ++i) {
				if (vfactor_[(size_t)i]==0) vfactor_[(size_t)i]=1;
			}
		}

		int getThreshold(int& t) {return t = (int)EarlySingleCRA<Domain>::EARLY_TERM_THRESHOLD;}

		Integer& getModulus(Integer& m) {EarlySingleCRA<Domain>::getModulus(m);return m;}
		Integer& getResidue(Integer& r) {EarlySingleCRA<Domain>::getResidue(r);return r;}

		template<class Vect>
		Vect& getResidue(Vect& r) {
			Vect z(r.field()),vf(r.field()), vm(r.field());
			FullMultipCRA<Domain>::result(z);

			typename Vect::const_iterator it,itf,itm;
			Integer M; getModulus(M);
			getPreconditioner(vf,vm);

			//r.resize(vf.size());//vector of residues
			inverse(r, vf, M);
			normproductin(r, z, M);
			normproductin(r, vm, M);

			//FullMultipCRA<Domain>::getResidue(r);
			return r;
		}

		template<class Vect>
		void initialize (const Integer& D, const Vect& e) {
			srand48(BaseTimer::seed());
			vfactor_.resize( e.size(),1 );
			vmultip_.resize( e.size(),1 );
			randv. resize ( e.size() );

			for ( std::vector<unsigned long>::iterator int_p = randv. begin(); int_p != randv. end(); ++ int_p)
				*int_p = ((unsigned long)lrand48()) % 20000 - 10000;

			std::vector<Integer> vz(vfactor_.size());
			inverse(vz,vfactor_,D);
			productin(vz,vmultip_,D);
			productin(vz,e,D);

			Integer z;
			dot(z,D,vz,randv);

			EarlySingleCRA<Domain>::initialize(D, z);
			FullMultipCRA<Domain>::initialize(D, e);
		}

		template<class Vect>
		void initialize (const Domain& D, Vect& e) {
			srand48(BaseTimer::seed());
			vfactor_.resize( e.size(),1 );
			vmultip_.resize( e.size(),1 );
			randv. resize ( e.size() );

			for ( std::vector<unsigned long>::iterator int_p = randv. begin(); int_p != randv. end(); ++ int_p)
				*int_p = ((unsigned long)lrand48()) % 20000 - 10000;

			std::vector<DomainElement> vz(vfactor_.size());
			inverse(vz,vfactor_,D);
			productin(vz,vmultip_,D);
			productin(vz,e,D);

			DomainElement z;
			dot(z,D,vz,randv);

			EarlySingleCRA<Domain>::initialize(D, z);
			FullMultipCRA<Domain>::initialize(D, e);
		}

		template<class Vect>
		void progress (const Integer& D, const Vect& e) {

			// Could be much faster
			// - do not compute twice the product of moduli
			// - reconstruct one element of e until Early Termination,
			//   then only, try a random linear combination.

			std::vector<Integer> vz(vfactor_.size());
			inverse(vz,vfactor_,D);
			productin(vz,vmultip_,D);
			productin(vz,e,D);

			Integer z;
			dot(z,D,vz,randv);

			EarlySingleCRA<Domain>::progress(D, z);
			FullMultipCRA<Domain>::progress(D, e);
		}

		template<class Vect>
		void progress (const Domain& D, const Vect& e) {
			//z = (e/ factor mod D)
			//!@todo Could be much faster
			// - do not compute twice the product of moduli
			// - reconstruct one element of e until Early Termination,
			//   then only, try a random linear combination.

			std::vector<DomainElement> vz(vfactor_.size());
			inverse(vz,vfactor_,D);
			productin(vz,vmultip_,D);
			productin(vz,e,D);
			DomainElement z;
			dot(z,D,vz,randv);

			EarlySingleCRA<Domain>::progress(D, z);
			FullMultipCRA<Domain>::progress(D, e);
		}

		template<class OKDomain>
		void progress (const Domain& D, const BlasVector<OKDomain>& e) {
			//z = (e/ factor mod D)
			//!@todo Could be much faster
			// - do not compute twice the product of moduli
			// - reconstruct one element of e until Early Termination,
			//   then only, try a random linear combination.

			BlasVector<Domain> vz(D,vfactor_.size());
			inverse(vz,vfactor_,D);
			productin(vz,vmultip_,D);
			productin(vz,e,D);
			DomainElement z;
			dot(z,D,vz,randv);

			EarlySingleCRA<Domain>::progress(D, z);
			FullMultipCRA<Domain>::progress(D, e);
		}

		bool terminated() {
			bool ET = EarlySingleCRA<Domain>::terminated();
			if (FullMultipCRA<Domain>::LOGARITHMIC_UPPER_BOUND> 1.0) ET = ET || FullMultipCRA<Domain>::terminated();
			return ET;
		}

		bool noncoprime(const Integer& i) const {
			return EarlySingleCRA<Domain>::noncoprime(i);
		}

		//Integer& getFactor(Integer& f) {
		//	return f = EarlySingleCRA<Domain>::factor_;
		//}

		//Integer& getMultip(Integer& m) {
		//	return m=EarlySingleCRA<Domain>::multip_;
		//}

		template<class Vect>
		Vect& getFactor(Vect& vf) const {
			//Integer f = getFactor(f);
			vf.clear();// vf.resize(vfactor_.size());
			typename Vect::const_iterator it = vfactor_.begin();
			for (;it != vfactor_.end(); ++it) {
				vf.push_back((*it));
			}
			return vf;// = vfactor_;
		}


		template<class Vect>
		Vect& getMultip(Vect& vm) const {
			//Integer m = getMultip(m);
			vm.clear(); //vm.resize(vmultip_ .size());
			typename Vect::const_iterator it = vmultip_.begin();
			for (;it != vmultip_.end(); ++it) {
				vm.push_back((*it));
			}
			return vm;
		}

		//Integer& getPreconditioner(Integer& f, Integer& m) const {
		//	getMultip(m);
		//	return getFactor(f);
		//}

		template<class Vect>
		Vect& getPreconditioner(Vect& vf, Vect& vm) const {
			getMultip(vm);
			getFactor(vf);
			return vf;
		}

		//Quotient& getPreconditioner(Quotient& q) {
		//	return q = EarlySingleCRA<Domain>::getPreconditioner(q);
		//}

		template<template<class, class> class Vect, template<class> class Alloc>
		Vect<Quotient, Alloc<Quotient> >& getPreconditioner(Vect<Quotient, Alloc<Quotient> >& vq) const {

			Vect<Integer, Alloc<Integer> > vf,vm;
			getPreconditioner(vf,vm);
			vq.clear();

			typename Vect<Integer, Alloc<Integer> >::const_iterator itf, itm;
			for (itf = vf.begin(), itm=vm.begin() ; itf != vf.end(); ++itf,++itm) {
				vq.push_back(Quotient(*itm,*itf));
			}

			return vq;
		}

		template<template<class, class> class Vect, template<class> class Alloc>
		Vect<Integer, Alloc<Integer> >& result(Vect<Integer, Alloc<Integer> >& r) {
			if ((FullMultipCRA<Domain>::LOGARITHMIC_UPPER_BOUND> 1.0) && ( FullMultipCRA<Domain>::terminated() )) {
				FullMultipCRA<Domain>::result(r);
				return r ;
			}
			else {
				//Integer M; getModulus(m);
				Vect<Integer, Alloc<Integer> > z,vf, vm;
				FullMultipCRA<Domain>::result(z);

				typename Vect<Integer, Alloc<Integer> >::const_iterator it,itf,itm;

				Integer M; getModulus(M);
				getPreconditioner(vf,vm);

				Vect<Integer, Alloc<Integer> > residue;
				getResidue(residue);
				//vector of residues

				r.clear();

				//getPreconditioner(vf,vm);
				itf = vf.begin(); itm = vm.begin();it = z.begin();

				for (; it!= z.end(); ++it, ++itf,++itm ) {
					r.push_back(*itf * (*it) / (*itm) );
				}
				return r;
			}
		}

		template<template<class,class> class Vect, template<class> class Alloc>
		Vect<Integer, Alloc<Integer> >& result(Vect<Integer, Alloc<Integer> >& num, Integer& den)
		{
			if ((FullMultipCRA<Domain>::LOGARITHMIC_UPPER_BOUND> 1.0) && ( FullMultipCRA<Domain>::terminated() )) {
				FullMultipCRA<Domain>::result(num);
				den = 1;
				return num;
			}
			else {
				Vect<Integer, Alloc<Integer> > z,vf, vm;
				FullMultipCRA<Domain>::result(z);//vector of non prec results

				typename Vect<Integer, Alloc<Integer> >::const_iterator it,itf,itm;
				typename Vect<Integer, Alloc<Integer> >::iterator itt;

				Integer M; getModulus(M);
				getPreconditioner(vf,vm);

				Vect<Integer, Alloc<Integer> > residue;//vector of residues
				getResidue(residue);

				num.clear();

				//getPreconditioner(vf,vm);
				itf = vf.begin(); itm = vm.begin();it = residue.begin();
				den = 1; Integer old_den =1;
				for (; it!= residue.end(); ++it, ++itf,++itm ) {
					lcm(den,den,*itm);Integer d = den/old_den;

					num.push_back((*itf) * (*it));
					if (den != old_den) {
						for (itt=num.begin(); itt !=num.end()-1 ;++itt) {
							*itt *= d;
						}
						*itt = *itt * (den/ (*itm));
					}
					old_den = den;
				}
				return num;
			}
		}

		BlasVector<Givaro::ZRing<Integer> >& result(BlasVector<Givaro::ZRing<Integer> >& num, Integer& den)
		{
			if ((FullMultipCRA<Domain>::LOGARITHMIC_UPPER_BOUND> 1.0) && ( FullMultipCRA<Domain>::terminated() )) {
				FullMultipCRA<Domain>::result(num);
				den = 1;
				return num;
			}
			else {
				Givaro::ZRing<Integer> Z;
				BlasVector<Givaro::ZRing<Integer> > z(Z),vf(Z), vm(Z);
				FullMultipCRA<Domain>::result(z);//vector of non prec results

				typename BlasVector<Givaro::ZRing<Integer> >::const_iterator it,itf,itm;
				typename BlasVector<Givaro::ZRing<Integer> >::iterator itt;

				Integer M; getModulus(M);
				getPreconditioner(vf,vm);

				BlasVector<Givaro::ZRing<Integer> > residue(Z);//vector of residues
				getResidue(residue);

				num.clear();

				//getPreconditioner(vf,vm);
				itf = vf.begin(); itm = vm.begin();it = residue.begin();
				den = 1; Integer old_den =1;
				for (; it!= residue.end(); ++it, ++itf,++itm ) {
					lcm(den,den,*itm);Integer d = den/old_den;

					num.push_back((*itf) * (*it));
					if (den != old_den) {
						for (itt=num.begin(); itt !=num.end()-1 ;++itt) {
							*itt *= d;
						}
						*itt = *itt * (den/ (*itm));
					}
					old_den = den;
				}
				return num;
			}
		}

		template<template<class, class> class Vect, template<class> class Alloc>
		Vect<Quotient, Alloc<Quotient> >& result(Vect<Quotient, Alloc<Quotient> >& q)
		{
			q.clear();
			if ((FullMultipCRA<Domain>::LOGARITHMIC_UPPER_BOUND> 1.0) && ( FullMultipCRA<Domain>::terminated() )) {
				std::vector<Integer> vz;
				FullMultipCRA<Domain>::result(vz);

				typename Vect<Integer, Alloc<Integer> >::const_iterator it = vz.begin();
				for (; it!= vz.end(); ++it) {
					q.push_back(Quotient(*it,1));
				}
				return q;
			}
			else {
				Vect<Integer, Alloc<Integer> > z,vf, vm;
				FullMultipCRA<Domain>::result(z);
				typename Vect<Integer, Alloc<Integer> >::const_iterator it = z.begin(),itf,itm;

				Integer M; getModulus(M);
				getPreconditioner(vf,vm);

				Vect<Integer, Alloc<Integer> > residue;//vector of residues
				inverse(residue, vf, M);
				normproductin(residue, z, M);
				normproductin(residue, vm, M);

				itf = vf.begin(); itm = vm.begin();

				for (; it!= z.end(); ++it, ++itf,++itm ) {
					q.push_back(Quotient(*itf * (*it) , *itm));
				}
				return q;
			}
		}


		template<class Vect>
		bool changePreconditioner(const Vect& vf, const Vect& vm) {
			//Warning does not detect unchanged preconditioners !!!
			//if ((factor_ == f) && (multip_==m)) return EarlySingleCRA<Domain>::terminated();

			typename Vect::const_iterator itf, itm, itf2, itm2;

			vfactor_ = vf;
			for (int i=0; i < (int)vfactor_.size(); ++i) {
				if (vfactor_[(size_t)i]==0) vfactor_[(size_t)i]=1;	//if factor ==0 set no factor
			}
			vmultip_ = vm;

			Givaro::ZRing<Integer> Z;
			Vect e(Z,vfactor_.size());

			//clear CRAEarlySingle;
			EarlySingleCRA<Domain>::occurency_ = 0;
			EarlySingleCRA<Domain>::nextM_ = 1;
			EarlySingleCRA<Domain>::primeProd_ = 1;
			EarlySingleCRA<Domain>::residue_ = 0;

			//Computation of residue_

			//std::vector< double >::iterator  _dsz_it = RadixSizes_.begin();
			std::vector< LazyProduct >::iterator _mod_it = FullMultipCRA<Domain>::RadixPrimeProd_.end();// list of prime products
			std::vector< BlasVector<Givaro::ZRing<Integer> > >::iterator _tab_it = FullMultipCRA<Domain>::RadixResidues_.end();// list of residues as vectors of size 1
			std::vector< bool >::iterator    _occ_it = FullMultipCRA<Domain>::RadixOccupancy_.end();//flags of occupied fields
			int n = (int)FullMultipCRA<Domain>::RadixOccupancy_.size();
			//std::vector<Integer> ri(1); LazyProduct mi; double di;
			//could be much faster if max occupandy is stored
			int prev_shelf=0, shelf = 0; Integer prev_residue_=0;
			--_occ_it; --_mod_it; --_tab_it;//last elemet
			for (int i=n; i > 0; --i, --_mod_it, --_tab_it, --_occ_it ) {
				//--_occ_it; --_mod_it; --_tab_it;
				++shelf;
				if (*_occ_it) {
					Integer D = _mod_it->operator()();
					Vect e_v(Z,vfactor_.size());
					inverse(e_v,vfactor_,D);
					productin(e_v,*_tab_it,D);
					productin(e_v,vmultip_,D);

					Integer z;
					dot(z,D, e_v, randv);


					prev_residue_ = EarlySingleCRA<Domain>::residue_;
					EarlySingleCRA<Domain>::progress(D,z);

					if (prev_residue_ == EarlySingleCRA<Domain>::residue_ ) {
						EarlySingleCRA<Domain>::occurency_ = EarlySingleCRA<Domain>::occurency_ + (unsigned int) (shelf - prev_shelf);
					}
					if ( EarlySingleCRA<Domain>::terminated() ) {
						return true;
					}
					prev_shelf = shelf;
				}
			}

			return EarlySingleCRA<Domain>::terminated();

			//forall results in FullMultipCRA do
			//	precondition
			//	progress to  EarlySingleCRA
			//	check for termination
			//recompute
		}

		bool changeVector() {
			for ( std::vector<unsigned long>::iterator int_p = randv. begin();int_p != randv. end(); ++ int_p)
				*int_p = ((unsigned long)lrand48()) % 20000;
			return changePreconditioner(vfactor_,vmultip_);
		}

	protected:
		template <template<class> class Alloc, template<class, class> class Vect1, class Vect2>
		DomainElement& dot (DomainElement& z, const Domain& D,
				    const Vect1<DomainElement, Alloc<DomainElement> >& v1,
				    const Vect2& v2)
		{
			D.assign(z,D.zero); DomainElement tmp;
			typename Vect1<DomainElement, Alloc<DomainElement> >::const_iterator v1_p;
			typename Vect2::const_iterator v2_p;

			for (v1_p  = v1. begin(), v2_p = v2. begin();v1_p != v1. end();++ v1_p, ++ v2_p)
				D.axpyin(z, (*v1_p), D.init(tmp, (*v2_p)));
			return z;
		}

		template<class Vect2>
		DomainElement& dot (DomainElement& z, const Domain& D,
				    const BlasVector<Domain>& v1,
				    const Vect2& v2)
		{
			D.assign(z,D.zero); DomainElement tmp;
			typename BlasVector<Domain>::const_iterator v1_p;
			typename Vect2::const_iterator v2_p;

			for (v1_p  = v1. begin(), v2_p = v2. begin();v1_p != v1. end();++ v1_p, ++ v2_p)
				D.axpyin(z, (*v1_p), D.init(tmp, (*v2_p)));
			return z;
		}


		template <template<class> class Alloc, template<class, class> class Vect1, class Vect2>
		Integer& dot (Integer& z, const Integer& D, const Vect1<Integer, Alloc<Integer> >& v1, const Vect2& v2)
		{
			z = 0;
			typename Vect1<Integer, Alloc<Integer> >::const_iterator v1_p;
			typename Vect2::const_iterator v2_p;
			for (v1_p  = v1. begin(), v2_p = v2. begin(); v1_p != v1. end(); ++ v1_p, ++ v2_p) {
				z = (z + (*v1_p)*(*v2_p))%D;
			}
			return z;
		}

		template<class Vect2>
		Integer& dot (Integer& z, const Integer& D, const BlasVector<Givaro::ZRing<Integer> >& v1, const Vect2& v2)
		{
			z = 0;
			typename BlasVector<Givaro::ZRing<Integer> >::const_iterator v1_p;
			typename Vect2::const_iterator v2_p;
			for (v1_p  = v1. begin(), v2_p = v2. begin(); v1_p != v1. end(); ++ v1_p, ++ v2_p) {
				z = (z + (*v1_p)*int64_t(*v2_p))%D;
			}
			return z;
		}


		template<class Vect1, class Vect2>
		Vect1& inverse(Vect1& vz,const Vect2& vf, const Domain D)
		{
			vz.clear();
			typename Vect2::const_iterator it = vf.begin();

			for (; it != vf.end(); ++it) {
				DomainElement z,i;
				D.assign(z,D.one);
				D.init(i,*it);
				if (!D.isZero(i)) D.inv(z,i);
				vz.push_back(z);
			}
			return vz;

		}

		template<class Vect1, class Vect2>
		Vect1& inverse(Vect1& vz,const Vect2& vf, const Integer& D)
		{
			vz.clear();
			typename Vect2::const_iterator it = vf.begin();

			for (; it != vf.end(); ++it) {
				Integer z=1;
				if ((*it) != 0) inv(z,*it,D);
				vz.push_back(z);
			}
			return vz;
		}

	public:
		template<class Vect1, class Vect2>
		Vect1& productin(Vect1& vz, const Vect2 &vm, const Domain D) {
			typename Vect1::iterator v1_p;
			typename Vect2::const_iterator v2_p;

			for (v1_p  = vz. begin(), v2_p = vm. begin(); v1_p != vz. end(); ++ v1_p, ++ v2_p) {
				DomainElement i;
				D.init(i,*v2_p);
				D.mulin(*v1_p , i);
			}
			return vz;
		}


		template<class Vect1, class Vect2>
		Vect1& productin(Vect1& vz, const Vect2 vm, const Integer& D) {
			typename Vect1::iterator v1_p;
			typename Vect2::const_iterator v2_p;

			for (v1_p  = vz. begin(), v2_p = vm. begin(); v1_p != vz. end(); ++ v1_p, ++ v2_p) {
				*v1_p = (*v1_p) * (*v2_p);
				*v1_p = (*v1_p) % D;
			}
			return vz;
		}

		template<class Vect1, class Vect2>
		Vect1& normproductin(Vect1& vz, const Vect2 vm, const Integer& D) {
			typename Vect1::iterator v1_p;
			typename Vect2::const_iterator v2_p;
			for (v1_p  = vz. begin(), v2_p = vm. begin(); v1_p != vz. end(); ++ v1_p, ++ v2_p) {
				*v1_p = (*v1_p) * (*v2_p);
				*v1_p = (*v1_p) % D;
				Integer tmp = (*v1_p >0) ? *v1_p - D : *v1_p + D;
				if (absCompare(*v1_p,tmp)> 0) *v1_p = tmp;
			}
			return vz;
		}

		template<class Vect1, class Vect2>
		Vect1& productin(Vect1& vz, const Vect2 vm) {
			typename Vect1::iterator v1_p;
			typename Vect2::const_iterator v2_p;

			for (v1_p  = vz. begin(), v2_p = vm. begin(); v1_p != vz. end(); ++ v1_p, ++ v2_p) {
				*v1_p *= (*v2_p);
			}

			return vz;
		}

	};

} //namespace LinBox

#endif //__LINBOX_varprec_cra_multip_single_H


// Local Variables:
// mode: C++
// tab-width: 8
// indent-tabs-mode: nil
// c-basic-offset: 8
// End:
// vim:sts=8:sw=8:ts=8:noet:sr:cino=>s,f0,{0,g0,(0,\:0,t0,+0,=s