pqc/external/flint-2.4.3/flintxx/doc/genericxx.txt
2014-05-24 23:16:06 +02:00

479 lines
18 KiB
Plaintext

/*=============================================================================
vim: spell spelllang=en textwidth=79
This file is part of FLINT.
FLINT is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 2 of the License, or
(at your option) any later version.
FLINT 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 General Public License for more details.
You should have received a copy of the GNU General Public License
along with FLINT; if not, write to the Free Software
Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
=============================================================================*/
/******************************************************************************
Copyright (C) 2013 Tom Bachmann
******************************************************************************/
*******************************************************************************
Rules and standard methods
A typical expression template class begins with the following lines of code:
\begin{lstlisting}[language=c++]
template<class Operation, class Data>
class some_expression
: public expression<derived_wrapper<some_expression>, Operation, Data>
{
// ...
};
\end{lstlisting}
We document here methods this class inherits from its base, and how they
relate to rules.
There are the following public typedefs:
\begin{description}
\item[ev\_traits\_t] A specialisation of \code{detail::evaluation_traits}.
Used to compute the rule for evaluation.
\item[derived\_t] The specialised derived class.
\item[evaluated\_t] The resulting type of evaluating this expression.
\item[evaluation\_return\_t] The return type of \code{evaluate()}. This
differs from the above for immediates, where evaluation returns a
reference instead of a copy.
\item[data\_t] The same as \code{Data}.
\item[operation\_t] The same as \code{Operation}.
\end{description}
*******************************************************************************
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Standard methods
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
data_t& some_expression::_data()
const data_t& some_expression::_data() const
Obtain the data related to this expression template.
evaluated_t some_expression::create_temporary() const
Default instantiate a temporary. Override this if your class is not default
instantiable.
template<class T> T some_expression::to() const
Convert self to type \code{T} (after evaluating). Uses
\code{rules::conversion}.
void some_expression::print(std::ostream& o) const
Print self to \code{o}. Uses \code{rules::print} or
\code{rules::to_string}.
int some_expression::print(FILE* f = stdout) const
Print self to \code{f}. Uses \code{rules::cprint}.
int some_expression::print_pretty(FILE* f = stdout) const
Print self to \code{f}. Uses \code{rules::print_pretty}.
template<class T> int some_expression::print_pretty(const T& extra,
FILE* f = stdout) const
Print self to \code{f}. Uses \code{rules::print_pretty} with two arguments.
int some_expression::read(FILE* f = stdin)
Read self from \code{f}. Uses \code{rules::read}.
const evaluation_return_t some_expression::evaluate() const
Evaluate self.
template<class T> void some_expression::set(const T& t)
Assign \code{t} to self. Uses evaluation and/or \code{rules::assignment}.
template<class T> bool some_expression::equals(const T& t) const
Determine if \code{t} is equal to self. Uses \code{rules::equals} or
\code{rules::cmp}.
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Global functions
In addition to member functions, flintxx also provides a number of global
functions. In general these operate on sets of arguments at least one of
which derives from \code{expression}, and are conditionally enabled only if
the relevant operation is implemented (via a rule).
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
template<class Expr> std::ostream& operator<<(std::ostream& o, const Expr& e)
Print \code{e} to \code{o}. Uses the member \code{print}.
template<class Expr1, class Expr2> bool operator==(const Expr1&, const Expr2&)
template<class Expr1, class Expr2> bool operator!=(const Expr1&, const Expr2&)
Compare two expressions. Uses the member \code{equals}.
template<class Expr1, class Expr2> bool operator??(const Expr1&, const Expr2&)
Relational operators \code{< > <= =>} are implemented using
\code{rules::cmp}.
template<class Expr1, class Expr2> ?? operator??(const Expr1&, const Expr2&)
Arithmetic operators \code{+ - * / % & | ^ << >>} are implemented
by constructing new expression templates with operation
\code{operations::plus} etc.
template<class Expr1> ?? operator??(const Expr1&)
Unary operators \code{- ~} are implemented by constructing new expression
templates with operation \code{operations::negate} and
\code{operations::complement}.
template<class Expr1, class Expr2> ?? operator?=(const Expr1&, const Expr2&)
Arithmetic-assignment operators \code{+= -= *= /= %= |= &= ^=}.
template<class Expr1> int print(const Expr1&)
template<class Expr1> int print(FILE*f, const Expr1&)
template<class Expr1> int print_pretty(const Expr1&)
template<class Expr1> int print_pretty(FILE*f, const Expr1&)
template<class Expr1, class T> int print_pretty(const Expr1&, const T& extra)
template<class Expr1, class T> int print_pretty(FILE*f, const Expr1&,
const T& extra)
Forward to member.
template<class Expr1, class Expr2> void swap(Expr1& e1, Expr2& e2)
Swap \code{e1} and \code{e2} using \code{rules::swap}. Note that via ADL,
this can be used by STL containers.
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
flintxx classes
The flint wrapper classes share some other common interfaces. These have to
be enabled using the convenience macros in \code{flintxx/flint_classes.h}
(q.v.). Here \code{accessname} and \code{ctype} are specified via the
macros. For e.g. \code{fmpz_polyxx} these are \code{_poly} and
\code{fmpz_poly_struct}.
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
?? some_expression::accessname()
?? some_expression::accessname() const
Obtain a reference to the underlying C struct. This is only available on
immediate expressions.
some_expression_ref::some_expression_ref(some_expression&)
some_expression_srcref::some_expression_srcref(const some_expression&)
some_expression_srcref::some_expression_srcref(some_expression_ref)
Build a reference type. Note that these are \emph{implicit} constructors.
static some_expression_ref some_expression_ref::make(ctype*)
static some_expression_srcref some_expression_srcref::make(const ctype*)
Build a reference type from a pointer to the underlying C struct.
*******************************************************************************
Convenience macros
*******************************************************************************
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
flintxx/rules.h
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
FLINT_DEFINE_GET2(name, totype, fromtype1, fromtype2, eval)
Specialise a getter called \code{name}, which takes arguments \code{e1} of
type \code{fromtype1} and \code{e2} of type \code{fromtype2}. It returns
\code{totype} by executing \code{eval}.
FLINT_DEFINE_GET(name, totype, fromtype, eval)
Same as \code{FLINT_DEFINE_GET2(name, totype, fromtype, fromtype, eval)}.
FLINT_DEFINE_GET_COND(name, totye, cond, eval)
Specialise a getter called \code{name}, which takes an argument \code{from}
of type \code{T:cond} It returns \code{totype} by
executing \code{eval}.
FLINT_DEFINE_DOIT(name, totype, fromtype, eval)
Specialise a doit rule called \code{name}, which takes arguments
\code{to} of type \code{totype&} and \code{from} of type
\code{const fromtype&}, and executes \code{eval}.
FLINT_DEFINE_DOIT_COND(name, totype, cond, eval)
Same as above, but takes \code{const T& from} for any \code{T:cond}.
FLINT_DEFINE_DOIT_COND2(name, cond1, cond2, eval)
Same as \code{FLINT_DEFINE_DOIT_COND}, but takes \code{T& to} and
\code{const U& from} for any \code{T} satisfying \code{cond1<T>} and
\code{U} satisfying \code{cond2<U>}.
FLINT_DEFINE_PRINT_COND(cond, eval)
Specialise the \code{cprint} rule. This takes a arguments \code{FILE* to}
and \code{const T& from} for any \code{T:cond}. It
prints \code{from} to \code{to} and returns \code{int} by executing
\code{eval}.
FLINT_DEFINE_PRINT_PRETTY_COND(cond, eval)
Same as above, but with \code{print_pretty} instead of \code{cprint}.
FLINT_DEFINE_PRINT_PRETTY_COND2(cond, extratype, eval)
Same as above, but takes an additional argument \code{extratype extra}.
Useful e.g. when printing polynomials and taking an extra variable name.
FLINT_DEFINE_READ_COND(cond, eval)
Specialise the \code{read} rule. This takes a arguments \code{FILE* from}
and \code{T& to} for any \code{T:cond}. It
reads \code{to} from \code{from} and returns \code{int} by executing
\code{eval}.
FLINT_DEFINE_UNARY_EXPR_(name, rtype, type, eval)
Specialise the unary expression rule for \code{operations::name} with
nominal return type \code{rtype}. It takes
arguments \code{V& to} and \code{const type& from}. Here \code{V} is any
type which \code{rtype} can be evaluated into. Executes \code{eval}.
FLINT_DEFINE_UNARY_EXPR(name, type, eval)
Same as \code{FLINT_DEFINE_UNARY_EXPR_(name, type, type, eval)}.
FLINT_DEFINE_BINARY_EXPR2(name, rtype, type1, type2, eval)
Specialise the binary expression rule for \code{operations::name} of
nominal return type \code{rtype}, and arguments \code{type1} and
\code{type2}.
FLINT_DEFINE_BINARY_EXPR(name, type, eval)
Same as \code{FLINT_DEFINE_BINARY_EXPR2(name, type, type, type, eval)}.
FLINT_DEFINE_CBINARY_EXPR(name, type, eval)
Same as above, but with \code{commutative_binary_expression} instead of
\code{binary_expression}.
FLINT_DEFINE_BINARY_EXPR_COND(name, type, cond, eval)
FLINT_DEFINE_CBINARY_EXPR_COND(name, type, cond, eval)
Specialise the (commutative) binary expression rule for
\code{operations::name} of nominal return type \code{type}, and arguments
\code{type} and \code{T:cond}.
FLINT_DEFINE_BINARY_EXPR_COND2(name, rettype, cond1, cond2, eval)
FLINT_DEFINE_CBINARY_EXPR_COND2(name, rettype, cond1, cond2, eval)
Specialise the (commutative) binary expression rule for
\code{operations::name} of nominal return type \code{rettype}, and arguments
\code{T:cond1} and \code{U:cond2}.
FLINT_DEFINE_THREEARY_EXPR_COND3(name, rettype, cond1, cond2, cond3, eval)
FLINT_DEFINE_FOURARY_EXPR_COND4(name, retttype, cond1 ... cond4, eval)
FLINT_DEFINE_FIVEARY_EXPR_COND5(name, rettype, cond1 ... cond5, eval)
FLINT_DEFINE_SIXARY_EXPR_COND6(name, rettype, cond1 ... cond6, eval)
FLINT_DEFINE_SEVENARY_EXPR_COND7(name, rettype, cond1 ... cond7, eval)
Specialise higher order rules, similarly to the above.
FLINT_DEFINE_THREEARY_EXPR(name, retttype, T1, T2, T3, eval)
Specialise a threeary expression rule unconditionally.
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
flintxx/expression.h
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
FLINT_DEFINE_UNNOP(name)
FLINT_DEFINE_BINOP(name)
FLINT_DEFINE_THREEARY(name)
FLINT_DEFINE_FOURARY(name)
FLINT_DEFINE_FIVEARY(name)
FLINT_DEFINE_SIXARY(name)
FLINT_DEFINE_SEVENARY(name)
Introduce a new n-ary operation \code{operations::##name##_op} and make it
available. This has to be called in namespace \code{flint}.
FLINT_DEFINE_UNNOP_HERE(name)
FLINT_DEFINE_BINOP_HERE(name)
FLINT_DEFINE_THREEARY_HERE(name)
FLINT_DEFINE_FOURARY_HERE(name)
FLINT_DEFINE_FIVEARY_HERE(name)
FLINT_DEFINE_SIXARY_HERE(name)
FLINT_DEFINE_SEVENARY_HERE(name)
Make the n-ary operation \code{operations::##name##_op} available in the
current namespace.
FLINT_DEFINE_THREEARY_HERE_2DEFAULT(name, type1, val1, type2, val2)
Make the threeary operation \code{name} available in current namespace,
but with only two arguments, the second of which is of type \code{type1} and
defaults to \code{val1}, and the third argument always (implicitly) of type
\code{type2} and value \code{val2}.
The suggested usage of this macro is to first call
\code{FLINT_DEFINE_THREEARY_HERE} (or \code{FLINT_DEFINE_THREEARY}),
and then call \code{FLINT_DEFINE_THREEARY_HERE_2DEFAULT}. The effect will be an
operation which can be invoked with 1, 2 or 3 arguments.
FLINT_UNOP_ENABLE_RETTYPE(name, T1)
FLINT_BINOP_ENABLE_RETTYPE(name, T1, T2)
FLINT_THREEARY_ENABLE_RETTYPE(name, T1, T2, T3)
FLINT_FOURARY_ENABLE_RETTYPE(name, T1, T2, T3, T4)
FLINT_FIVEARY_ENABLE_RETTYPE(name, T1, T2, T3, T4, T5)
FLINT_SIXARY_ENABLE_RETTYPE(name, T1, T2, T3, T4, T5, T6)
FLINT_SEVENARY_ENABLE_RETTYPE(name, T1, T2, T3, T4, T5, T6, T7)
Obtain the resulting type of invoking \code{name} with arguments of types
\code{T1}, ..., \code{Tn} if this is possible. Otherwise results in an
(SFINAE) error.
FLINT_UNOP_BUILD_RETTYPE(name, rettype, T)
Obtain the resulting type (i.e. expression template) of invoking
\code{name} with argument type \code{T}, assuming the nominal return type
is \code{rettype}. This version is sometimes necessary to break cyclic
dependencies.
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
flintxx/flint\_classes.h
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
FLINTXX_DEFINE_BASICS(name)
Add standard constructors (forwarded to \code{data_t}, and implicit ones
for reference types). Here \code{name} is the name of the expression
template class.
FLINTXX_DEFINE_C_REF(name, ctype, accessname)
Enable the reference types scheme.
FLINTXX_DEFINE_FORWARD_STATIC(funcname)
Add a statically forwarded constructor (similar to \code{make} for
reference types) which invokes a static constructor of the same name of
\code{data_t}.
FLINTXX_DEFINE_MEMBER_UNOP_RTYPE(rettype, name)
Add a no-argument member function which applies self to the lazy function
\code{name}, where \code{name} has nominal return type \code{rettype}.
(The return type has to be specified to break circular dependencies.)
FLINTXX_DEFINE_MEMBER_UNOP(name)
Same as above, but where the nominal return type is the (evaluated type of
the) current expression template class.
FLINTXX_DEFINE_MEMBER_BINOP(name)
FLINTXX_DEFINE_MEMBER_3OP(name)
FLINTXX_DEFINE_MEMBER_4OP(name)
FLINTXX_DEFINE_MEMBER_5OP(name)
Add a member function which \code{n-1} arguments, the result of which is to
invoke \code{name} on self and the arguments (in that order).
FLINTXX_COND_S(Base)
FLINTXX_COND_T(Base)
Expands to a condition (which can be passed to e.g.
\code{FLINT_DEFINE_CBINARY_EXPR_COND2}) appropriate for testing a
source/target of type \code{Base}.
FLINTXX_DEFINE_TO_STR(Base, eval)
Add a \code{to_string} rule which works well with the \code{*_get_str}
functions in FLINT.
FLINTXX_DEFINE_SWAP(Base, eval)
Add a swap rule.
FLINTXX_DEFINE_CONVERSION_TMP(totype, Base, eval)
Define a conversion rule from \code{Base} to \code{totype}, which
default-constructs a temporary object \code{to} of type \code{totype},
then executes \code{eval}, and then returns \code{to}.
FLINTXX_DEFINE_CMP(Base, eval)
FLINTXX_DEFINE_EQUALS(Base, eval)
Define a cmp/equality rule.
FLINTXX_DEFINE_ASSIGN_STR(Base, eval)
Define a string assignment rule (used by many polynomial classes).
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
flintxx/matrix.h
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
FLINTXX_DEFINE_MATRIX_METHODS(Traits)
Inside a matrix expression template class definition, given the unified
access traits \code{Traits} appropriate for this class, define the standard
methods \code{rows, cols, create_temporary}.
FLINTXX_DEFINE_TEMPORARY_RULES(Matrix)
Given a matrix expression template class \code{Matrix}, define appropriate
temporary instantiation rule, disable temporary merging, etc.
*******************************************************************************
Helper functions
*******************************************************************************
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
flintxx/flint\_exception.h
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
void execution_check(bool worked, const std::string& where,
const std::string& context)
If \code{worked} is true, do nothing. Else raise a \code{flint_exception}
with message \code{context + " computation failed: " + where}.
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
permxx.h
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
slong* maybe_perm_data(permxx* p)
Return \code{0} if \code{p == 0}, and else the underlying data.
It is helpful to use this together with \code{traits::is_maybe_perm} as
condition.