www.digitalmars.com         C & C++   DMDScript  

digitalmars.D - shouldn't this throw ? iota(5).sliced(2,2)

reply Timothee Cour via Digitalmars-d <digitalmars-d puremagic.com> writes:
what's the rationale for:
Range length must be greater than or equal to the sum of shift and the
product of lengths
instead of:
Range length must be equal to the sum of shift and the product of lengths
?

It seems more lax and bug-prone.
Jan 06 2016
parent reply Ilya Yaroshenko <ilyayaroshenko gmail.com> writes:
On Thursday, 7 January 2016 at 04:47:16 UTC, Timothee Cour wrote:

 what's the rationale for:
 Range length must be greater than or equal to the sum of shift 
 and the
 product of lengths
 instead of:
 Range length must be equal to the sum of shift and the product 
 of lengths
 ?

 It seems more lax and bug-prone.


std.ndslice will contain a module or few for allocations, script 
like code, and concatenations. User defined range may not have 
slicing, but must be able to be used with `sliced` multiple times 
with different lengths.
-- Ilya
Jan 06 2016
parent reply Timothee Cour via Digitalmars-d <digitalmars-d puremagic.com> writes:
I'm not sure I understand your argument.
My problem is that iota(5) has 5 elements which is more than 2*2, so I
would expect
iota(5).sliced(2,2)
or
iota(7).sliced(2,3).sliced(1,2)
to throw, as in pretty much any other tensor library:

*matlab:*
reshape(0:4, 2,2)
Error using reshape
To RESHAPE the number of elements must not change.

python with numpy:
import numpy as np
np.arange(5).reshape((2, 2))
ValueError: total size of new array must be unchanged

eigen:
http://eigen.tuxfamily.org/dox-devel/unsupported/TensorMorphing_8h_source.html
eigen_assert(internal::array_prod(m_impl.dimensions()) ==
internal::array_prod(op.dimensions()));

This behavior will 100% cause hard to find bugs.
Why not instead allow this behavior only when a template argument is given:
ReshapeAllowsDownsize.yes




On Wed, Jan 6, 2016 at 10:39 PM, Ilya Yaroshenko via Digitalmars-d <
digitalmars-d puremagic.com> wrote:


 On Thursday, 7 January 2016 at 04:47:16 UTC, Timothee Cour wrote:


 what's the rationale for:
 Range length must be greater than or equal to the sum of shift and the
 product of lengths
 instead of:
 Range length must be equal to the sum of shift and the product of lengths
 ?

 It seems more lax and bug-prone.


 std.ndslice will contain a module or few for allocations, script like
 code, and concatenations. User defined range may not have slicing, but must
 be able to be used with `sliced` multiple times with different lengths.
 -- Ilya
Jan 08 2016
parent reply Ilya <ilyayaroshenko gmail.com> writes:
On Saturday, 9 January 2016 at 00:15:10 UTC, Timothee Cour wrote:

 I'm not sure I understand your argument.
 My problem is that iota(5) has 5 elements which is more than 
 2*2, so I
 would expect
 iota(5).sliced(2,2)
 or
 iota(7).sliced(2,3).sliced(1,2)
 to throw, as in pretty much any other tensor library:

 [...]


auto a = iota(5).sliced(5);
auto b = a.reshape(2, 2); // throws ReshapeException
--ilya
Jan 08 2016
parent reply Timothee Cour via Digitalmars-d <digitalmars-d puremagic.com> writes:
ok good, 'reshape' behaves as it should; but what's the rationale for not
throwing on 'iota(5).sliced(2,2) ' ?

in light of what i wrote above, it's surprising behavior and will cause
bugs. What are the advantages of allowing it to not throw?


On Fri, Jan 8, 2016 at 6:09 PM, Ilya via Digitalmars-d <
digitalmars-d puremagic.com> wrote:


 On Saturday, 9 January 2016 at 00:15:10 UTC, Timothee Cour wrote:


 I'm not sure I understand your argument.
 My problem is that iota(5) has 5 elements which is more than 2*2, so I
 would expect
 iota(5).sliced(2,2)
 or
 iota(7).sliced(2,3).sliced(1,2)
 to throw, as in pretty much any other tensor library:

 [...]


 auto a = iota(5).sliced(5);
 auto b = a.reshape(2, 2); // throws ReshapeException
 --ilya
Jan 08 2016
parent reply Ilya <ilyayaroshenko gmail.com> writes:
On Saturday, 9 January 2016 at 04:15:08 UTC, Timothee Cour wrote:

 ok good, 'reshape' behaves as it should; but what's the 
 rationale for not
 throwing on 'iota(5).sliced(2,2) ' ?

 in light of what i wrote above, it's surprising behavior and 
 will cause bugs. What are the advantages of allowing it to not 
 throw?


 On Fri, Jan 8, 2016 at 6:09 PM, Ilya via Digitalmars-d < 
 digitalmars-d puremagic.com> wrote:


 On Saturday, 9 January 2016 at 00:15:10 UTC, Timothee Cour 
 wrote:


 I'm not sure I understand your argument.
 My problem is that iota(5) has 5 elements which is more than 
 2*2, so I
 would expect
 iota(5).sliced(2,2)
 or
 iota(7).sliced(2,3).sliced(1,2)
 to throw, as in pretty much any other tensor library:

 [...]


 auto a = iota(5).sliced(5);
 auto b = a.reshape(2, 2); // throws ReshapeException
 --ilya




Agreed. I will add optional exception (by default sliced will 
trow).
--Ilya
Jan 08 2016
parent Timothee Cour via Digitalmars-d <digitalmars-d puremagic.com> writes:
On Fri, Jan 8, 2016 at 10:07 PM, Ilya via Digitalmars-d <
digitalmars-d puremagic.com> wrote:


 On Saturday, 9 January 2016 at 04:15:08 UTC, Timothee Cour wrote:


 ok good, 'reshape' behaves as it should; but what's the rationale for not
 throwing on 'iota(5).sliced(2,2) ' ?

 in light of what i wrote above, it's surprising behavior and will cause
 bugs. What are the advantages of allowing it to not throw?


 On Fri, Jan 8, 2016 at 6:09 PM, Ilya via Digitalmars-d <
 digitalmars-d puremagic.com> wrote:

 On Saturday, 9 January 2016 at 00:15:10 UTC, Timothee Cour wrote:

 I'm not sure I understand your argument.

 My problem is that iota(5) has 5 elements which is more than 2*2, so I
 would expect
 iota(5).sliced(2,2)
 or
 iota(7).sliced(2,3).sliced(1,2)
 to throw, as in pretty much any other tensor library:

 [...]


 auto a = iota(5).sliced(5);
 auto b = a.reshape(2, 2); // throws ReshapeException
 --ilya




 Agreed. I will add optional exception (by default sliced will trow).
 --Ilya


thanks!! glad we could agree on this!
Jan 08 2016