Open ttnghia opened 3 years ago
If this is added, be sure to add a section to the libcudf DEVELOPER_GUIDE.md as well.
This issue has been labeled inactive-30d
due to no recent activity in the past 30 days. Please close this issue if no further response or action is needed. Otherwise, please respond with a comment indicating any updates or changes to the original issue and/or confirm this issue still needs to be addressed. This issue will be labeled inactive-90d
if there is no activity in the next 60 days.
This issue has been labeled inactive-90d
due to no recent activity in the past 90 days. Please close this issue if no further response or action is needed. Otherwise, please respond with a comment indicating any updates or changes to the original issue and/or confirm this issue still needs to be addressed.
Currently, various cudf APIs need to use a special constant to denote an invalid row index for certain operations. However, there is not any guideline on what constant should be used to mark an invalid index row. Therefore, different APIs implement their own
INVALID_INDEX
constant.For example:
lists/explode.cu
:constexpr size_type InvalidIndex = -1;
detail/scatter.cuh
:std::numeric_limits<size_type>::lowest()
is usedlists/extract.cu
:std::numeric_limits<size_type>::max()
is usedAnd maybe more. As such, the constants used for
INVALID_INDEX
is not agreed across cudf development. We should explicitly add a sentinel constant in some of cudf main header and replace the existing ones with it. By doing so, we can enforce consistency and reduce confusion (why this file uses-1
while other file usesnumeric_limit::lowest()
ornumeric_limit::max()
).