[ntuple] finalize automatic schema evolution between collections #20007
Conversation
Collection items must have "_0" as item field name, which can be violated in RVectorField::CreateUntyped().
jblomer
force-pushed
the
ntuple-evolution-collections
branch
from
0c12fd5 to
30e1202
Compare
Test Results 22 files 22 suites 3d 17h 37m 22s ⏱️ Results for commit f420cd5. ♻️ This comment has been updated with latest results. |
jblomer
force-pushed
the
ntuple-evolution-collections
branch
from
30e1202 to
f420cd5
Compare
|
|
||
| void ROOT::RSetField::ReconcileOnDiskField(const RNTupleDescriptor &desc) | ||
| { | ||
| static const std::vector<std::string> prefixesRegular = {"std::set<", "std::unordered_set<", "std::map<"}; |
There was a problem hiding this comment.
I guess we discussed this, but I forgot the reasoning again, sorry: Why can we read into a kUnorderedSet from a std::map, but not from a std::unordered_map?
There was a problem hiding this comment.
I think from the point of view of I/O we could. But to make it happen, we would need an unordered_set<pair>, which is missing a default hash function...
There was a problem hiding this comment.
from a
std::map, but not from astd::unordered_map
we would need anunordered_set<pair>, which is missing a default hash function...
Why is this an issue for a std::unordered_map?
And is it a fundamental issue or just an issue for the test? I.e. if a user had written a map and wanted to read it into a unordered_set they would have have provided the necessary hash overload, isn't it?
|
|
||
| void ROOT::RMapField::ReconcileOnDiskField(const RNTupleDescriptor &desc) | ||
| { | ||
| static const std::vector<std::string> prefixesRegular = {"std::map<", "std::unordered_map<", "std::set<"}; |
There was a problem hiding this comment.
similarly here: std::unordered_set?
No description provided.