gmane.comp.parsers.spirit.devel

Dean Michael Berris | 22 Sep 08:02

[fusion] Short-circuited Exception-based View Iteration

Hi Everyone,

I'm not sure if the title makes sense, but I have a problem which can
be solved with a handful of preprocessor macros, but would like to see
solved with Fusion.

Let's say I have a vector of values:

fusion::vector<int, int, int, int> ints(0, 1, 2, 3);

I then have a function such as:

struct function {
  void operator()(int value) {
    if (value < 2)
      throw runtime_error("less than 3");
    cout << value << endl;
  };
};

Would there be a way for me to iterate through every element of 'ints'
and apply the function object instance to each element, and stop the
iteration only when function does not throw?

I know using exceptions for control flow is pretty frowned upon, and I
can certainly use it the other way around -- to short-circuit the
iteration only when it throws.

If you're interested in the use case for this particular question, I'm
looking for a way to make the dispatcher library I'm working on

(Continue reading)

Tobias Schwinger | 22 Sep 14:26

Re: [fusion] Short-circuited Exception-based View Iteration

Dean Michael Berris wrote:
> Hi Everyone,
> 
> I'm not sure if the title makes sense, but I have a problem which can
> be solved with a handful of preprocessor macros, but would like to see
> solved with Fusion.
> 
> Let's say I have a vector of values:
> 
> fusion::vector<int, int, int, int> ints(0, 1, 2, 3);
> 
> I then have a function such as:
> 
> struct function {
>   void operator()(int value) {
>     if (value < 2)
>       throw runtime_error("less than 3");
>     cout << value << endl;
>   };
> };
> 
> Would there be a way for me to iterate through every element of 'ints'
> and apply the function object instance to each element, and stop the
> iteration only when function does not throw?

No.

> I know using exceptions for control flow is pretty frowned upon, 

...and pretty inefficient, unless to terminate a deep recursion.

(Continue reading)

Joel de Guzman | 22 Sep 22:54

Re: [fusion] Short-circuited Exception-based View Iteration

Tobias Schwinger wrote:

> Reading the rest of your post, it seems you're basically looking for 
> algorithms like these:
> 
>     visit_if(sequence_reference,runtime_predicate,action)
>     visit_where_first(sequence_reference,runtime_predicate,action)

Seems those are specializations of:

     any(seq, f)

??

Description

For a sequence seq and unary function object f, any returns true if f returns 
true for at least one element of seq.

Regards,
--

-- 
Joel de Guzman
http://www.boost-consulting.com
http://spirit.sf.net

-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2005.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/

(Continue reading)

Tobias Schwinger | 22 Sep 23:17

Re: [fusion] Short-circuited Exception-based View Iteration

Joel de Guzman wrote:
> Tobias Schwinger wrote:
> 
>> Reading the rest of your post, it seems you're basically looking for 
>> algorithms like these:
>>
>>     visit_if(sequence_reference,runtime_predicate,action)
>>     visit_where_first(sequence_reference,runtime_predicate,action)
> 
> Seems those are specializations of:
> 
>      any(seq, f)
> 
> ??
> 
> Description
> 
> For a sequence seq and unary function object f, any returns true if f returns 
> true for at least one element of seq.

Yep, that is for the latter (based on the reasonable assumption that any 
checks by iterating the sequence in forward direction ).

The former would be 'count_if'.

Regards,
Tobias

-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft

(Continue reading)

Dean Michael Berris | 23 Sep 15:39

Re: [fusion] Short-circuited Exception-based View Iteration

Hi Joel!

On 9/22/07, Joel de Guzman <joel <at> boost-consulting.com> wrote:
> Tobias Schwinger wrote:
>
> > Reading the rest of your post, it seems you're basically looking for
> > algorithms like these:
> >
> >     visit_if(sequence_reference,runtime_predicate,action)
> >     visit_where_first(sequence_reference,runtime_predicate,action)
>
> Seems those are specializations of:
>
>      any(seq, f)
>
> ??
>
> Description
>
> For a sequence seq and unary function object f, any returns true if f returns
> true for at least one element of seq.
>

Almost... But 'any' will just return a boolean, but not which element
in seq (or a view of which elements in seq) where f returns true.
Perhaps 'filter' is what I'm looking for, but I'm just interested in
the first one where a predicate holds true, so even a find_if(...)
might suit my needs.

I'll explore a bit more.

(Continue reading)

Joel de Guzman | 23 Sep 16:40

Re: [fusion] Short-circuited Exception-based View Iteration

Dean Michael Berris wrote:

>>      any(seq, f)
>>
>> ??
>>
>> Description
>>
>> For a sequence seq and unary function object f, any returns true if f returns
>> true for at least one element of seq.
>>
> 
> Almost... But 'any' will just return a boolean, but not which element
> in seq (or a view of which elements in seq) where f returns true.

Do it in f where you know both.

> Perhaps 'filter' is what I'm looking for, but I'm just interested in
> the first one where a predicate holds true, so even a find_if(...)
> might suit my needs.

No. There's no way to do that. You can't have a runtime predicate
determine a compile time return type.

> I'll explore a bit more.
> 
> BTW, i haven't found a `find_first_if(...)` implementation (like what
> the STL has) which works with a homogeneous tuple -- somehow that
> container might be a special case that deserves attention (?) -- which
> returns a single value? Thoughts on this?

(Continue reading)

Dean Michael Berris | 24 Sep 06:39

Re: [fusion] Short-circuited Exception-based View Iteration

On 9/23/07, Joel de Guzman <joel <at> boost-consulting.com> wrote:
> Dean Michael Berris wrote:
> >
> > Almost... But 'any' will just return a boolean, but not which element
> > in seq (or a view of which elements in seq) where f returns true.
>
> Do it in f where you know both.
>

This may make f a bit bloated for my taste, but this will work. Thanks. :)

> > Perhaps 'filter' is what I'm looking for, but I'm just interested in
> > the first one where a predicate holds true, so even a find_if(...)
> > might suit my needs.
>
> No. There's no way to do that. You can't have a runtime predicate
> determine a compile time return type.
>

There might be special cases where this might happen though... Let me
illustrate below what I mean...

> > I'll explore a bit more.
> >
> > BTW, i haven't found a `find_first_if(...)` implementation (like what
> > the STL has) which works with a homogeneous tuple -- somehow that
> > container might be a special case that deserves attention (?) -- which
> > returns a single value? Thoughts on this?
> >
> > Maybe this will solve at least my problem... :D

(Continue reading)

shunsuke | 24 Sep 10:00

Re: [fusion] Short-circuited Exception-based View Iteration

Dean Michael Berris wrote:
> struct equals_2 {
>   template <typename T>
>   T operator() (T element) const {
>     if (element == 2) return element;
>   };
> };
> 
> typedef fusion::tuple<int, int, int> point_type;
> point_type point(1, 2, 3);
> assert(fusion::is_homogenous<point_type>::value == true);
> assert(fusion::find_first_if<equals_2>(point) == 2); // only works if
> 'point' is a homogeneous container

FWIW, a few weeks ago, I had a radical idea
about an iterator which runs through a tuple/fusion.

rectangle r; triangle t; circle c;
boost::tuple<rectangle*, triangle*, circle*> tup(&r, &t, &c);
BOOST_FOREACH (shape *s, oven::hetero<shape *>(tup)) {
     s->draw();
}

If interested, here is the implementation: http://tinyurl.com/2gojyw

Regards,

--

-- 
Shunsuke Sogame

(Continue reading)

Joel de Guzman | 24 Sep 10:20

Re: [fusion] Short-circuited Exception-based View Iteration

shunsuke wrote:
> Dean Michael Berris wrote:
>> struct equals_2 {
>>   template <typename T>
>>   T operator() (T element) const {
>>     if (element == 2) return element;
>>   };
>> };
>>
>> typedef fusion::tuple<int, int, int> point_type;
>> point_type point(1, 2, 3);
>> assert(fusion::is_homogenous<point_type>::value == true);
>> assert(fusion::find_first_if<equals_2>(point) == 2); // only works if
>> 'point' is a homogeneous container
> 
> FWIW, a few weeks ago, I had a radical idea
> about an iterator which runs through a tuple/fusion.
> 
> rectangle r; triangle t; circle c;
> boost::tuple<rectangle*, triangle*, circle*> tup(&r, &t, &c);
> BOOST_FOREACH (shape *s, oven::hetero<shape *>(tup)) {
>      s->draw();
> }
> 
> If interested, here is the implementation: http://tinyurl.com/2gojyw

I'm interested! This ought to be part of fusion 

Regards,
--

--

(Continue reading)

Joel de Guzman | 24 Sep 10:14

Re: [fusion] Short-circuited Exception-based View Iteration

Dean Michael Berris wrote:
> On 9/23/07, Joel de Guzman <joel <at> boost-consulting.com> wrote:
>> Dean Michael Berris wrote:
>>> Almost... But 'any' will just return a boolean, but not which element
>>> in seq (or a view of which elements in seq) where f returns true.
>> Do it in f where you know both.
>>
> 
> This may make f a bit bloated for my taste, but this will work. Thanks. :)

Perhaps you misunderstood what I mean. You can use function composition
to factor out the function in as high a granularity as you desire.
You do not have to stuff f with all the functionalities, you *compose*
f from basic primitives. That's the basic essence of functional programming.

Regards,
--

-- 
Joel de Guzman
http://www.boost-consulting.com
http://spirit.sf.net

-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2005.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/

Joel de Guzman | 24 Sep 10:17

Re: [fusion] Short-circuited Exception-based View Iteration

Dean Michael Berris wrote:

>> The world fusion (and for that matter, mpl) lives in has some peculiar
>> constraints. As I noted above, one is that you cannot determine a type
>> from a runtime predicate. You can, however, do it inside-out. IOTW,
>> let the predicate itself get the info that you need. If you really
>> need to return a type, somehow, you can, for example, use a pointer
>> to a base class, hold the value in a boost::any, etc. In other words
>> use type erasure. If I were you, I'd use the visitor, f, itself to
>> do the action as soon as the correct type is obtained. No type
>> erasure magic.
>>
> 
> This makes sense, though there might be a possibility where a
> homogeneous tuple would be able to return just a single value:
> 
> struct equals_2 {
>   template <typename T>
>   T operator() (T element) const {
>     if (element == 2) return element;
>   };
> };
> 
> typedef fusion::tuple<int, int, int> point_type;
> point_type point(1, 2, 3);
> assert(fusion::is_homogenous<point_type>::value == true);
> assert(fusion::find_first_if<equals_2>(point) == 2); // only works if
> 'point' is a homogeneous container

If that was the case, I'd use boost::array which is a nice data structure

(Continue reading)

Dean Michael C. Berris | 24 Sep 10:22

Re: [Spirit-devel] [fusion] Short-circuited Exception-based ViewIteration

> -----Original Message-----
> From: spirit-devel-bounces <at> lists.sourceforge.net [mailto:spirit-devel-
> bounces <at> lists.sourceforge.net] On Behalf Of Joel de Guzman
> Sent: Monday, September 24, 2007 1:18 AM
> To: spirit-devel <at> lists.sourceforge.net
> Cc: boost <at> lists.boost.org
> Subject: Re: [Spirit-devel] [fusion] Short-circuited Exception-based
> ViewIteration
> 
> Dean Michael Berris wrote:
> >
> > This makes sense, though there might be a possibility where a
> > homogeneous tuple would be able to return just a single value:
> >
> > struct equals_2 {
> >   template <typename T>
> >   T operator() (T element) const {
> >     if (element == 2) return element;
> >   };
> > };
> >
> > typedef fusion::tuple<int, int, int> point_type;
> > point_type point(1, 2, 3);
> > assert(fusion::is_homogenous<point_type>::value == true);
> > assert(fusion::find_first_if<equals_2>(point) == 2); // only works
if
> > 'point' is a homogeneous container
> 
> If that was the case, I'd use boost::array which is a nice data
structure

(Continue reading)

Dean Michael Berris | 23 Sep 15:29

Re: [Spirit-devel] [fusion] Short-circuited Exception-based View Iteration

Hi Tobias!

On 9/22/07, Tobias Schwinger <tschwinger <at> isonews2.com> wrote:
> Dean Michael Berris wrote:
> >
> > Would there be a way for me to iterate through every element of 'ints'
> > and apply the function object instance to each element, and stop the
> > iteration only when function does not throw?
>
> No.
>

I thought so. That confirms my suspicion that there isn't an obvious
way of doing it. 

> > I know using exceptions for control flow is pretty frowned upon,
>
> ...and pretty inefficient, unless to terminate a deep recursion.
>
> But to answer you're question, you can of course wrap another function
> around yours that throws if it doesn't catch anything .
>

This makes sense, but a little too convoluted for my taste. It's one
of the options, and I don't know if there's even an alternative
that'll be easy to use aside from the above suggestion. Thanks for the
idea. ;)

> Reading the rest of your post, it seems you're basically looking for
> algorithms like these:

(Continue reading)

Return

Return to gmane.comp.parsers.spirit.devel.

Advertisement

Project Web Page

List for developers of Spirit

Search Archive

Language

Change language

Options

Current view: Threads only / Showing only 20 lines / Not hiding cited text.
Change to All messages, whole messages, or hide cited text.

Post a message
NNTP Newsgroup
Classic Gmane web interface
XML

XML

RSS Feed
List Information

About Gmane