Follow up better type on pcl::index_t

[jingyangcarl]

Hi, as described in PR #3651, and issue #3351, it's recommended to create a new indices type, say pcl::index_t, instead of the plain container std::vector, std::vectorstd::uint32_t, or other types.

When I went through here:

pcl/common/include/pcl/conversions.h

Line 191 in 4316e11

std::uint32_t cloud_row_step = static_cast<std::uint32_t> (sizeof (PointT) * cloud.width);

pcl/common/include/pcl/conversions.h

Line 208 in 4316e11

for (std::uint32_t row = 0; row < msg.height; ++row)

The iterator here is defined as std::uint32_t. I think if we want to have a better indices type, should the iterator here be modified to some flexible type as well?

@kunaltyagi @taketwo

[kunaltyagi]

should the iterator here be modified to some flexible type as well

Technically, indices are not needed for input in IO. They are mostly for algorithms. We can use std::size_t for input and output.

Some points to consider:

• std::size_t technically fails for multi-dimension data (each dimension's max index_type multiplied together can be more than max uint64)
• common-use imaging sensors aren't even at 2^16 per dimension, so for images, 2^32 is ok (since the product for 2D will be 2^64, our max index size)
• Beyond 2^64 should not be the topic right now. We'll cross the bridge when we have to (probably never)

[stale]

Marking this as stale due to 30 days of inactivity. It will be closed in 7 days if no further activity occurs.