16 Mar
2007
16 Mar
'07
1:30 p.m.
Nicolas Ballas <ballas(a)lrde.epita.fr> writes:
Index: ChangeLog
from Nicolas Ballas <ballas(a)lrde.epita.fr>
sparse_encode working.
* oln/core/encode/sparse_encode.hh: Update
``Update'' is quite general; please write more descriptive sentences.
And don't forget the ending period. :)
Index: oln/core/encode/sparse_encode.hh
--- oln/core/encode/sparse_encode.hh (revision 879)
+++ oln/core/encode/sparse_encode.hh (working copy)
@@ -33,6 +33,8 @@
# include <oln/core/sparse/sparse_image.hh>
+# include <vector>
+
namespace oln
{
@@ -45,13 +47,44 @@
*/
template <typename I>
sparse_image<typename I::point, typename I::value>
- sparse_encode(const Image<I> input)
+ sparse_encode(const Image<I>& input)
{
sparse_image<typename I::point, typename I::value> output;
typename I::piter p (input.points());
- // typename I::coord len = 1;
+ unsigned len = 1;
+ typename I::coord old = 1; /*!< old point first dim coordinate */
typename I::point rstart; /*!< range pointstart */
- typename I::value rvalue; /*!< range value */
+ std::vector<typename I::value> values; /*!< range value */
80 columns; aligned arguments.
+
+ p.start();
+ if (!p.is_valid())
+ return output;
+
+ rstart = p;
+ //FIXME: is it generall ?
Typo; missing space after `//'.
+ old = (p.to_point())[0];
+ values.push_back(input(p.to_point()));
Can't we write
input(p)
directly instead?
+ p.next();
+ while (p.is_valid())
+ {
+ if ((p.to_point())[0] - 1 == old)
+ {
+ ++len;
+ values.push_back(input(p.to_point()));
+ }
+ else
+ {
+ output.insert(rstart, len, values);
+ rstart = p;
+ len = 1;
+ values.clear();
+ values.push_back(input(p.to_point()));
+ }
+ old = (p.to_point())[0];
+ p.next();
+ }
+ output.insert(rstart, len, values);
+ return output;
}
}
Good!
It's a good idea to write a test for each kind of image you add to the
library, even if this test is dead simple. It helps to track
regression, notably using the Build Farm (https://build.lrde.org/).