Conversation
form_interface is now split between form_reader_interface and form_writer_interface for example.
been replaced by PersistenceReader and PersistenceWriter.
in persistence directory.
Removed a few files that were restored in the merge with main accidentally.
|
@phlexbot format |
|
Review the full CodeQL report for details. |
|
@phlexbot format |
Format Fixes Applied✅ clang-format fixes pushed (commit c47a8c6) |
Storage[Reader|Writer] member variables.
…/phlex into form_read_write_split
wwuoneway
left a comment
There was a problem hiding this comment.
Thanks @aolivier23 and @gemmeren for the constructive comments and productive conversation. I agree with the proposed path: let's proceed by renaming output_item_config (to something more neutral like Itemconfig or item_config) to resolve the immediate naming conflict/confusion, and then complete/merge this PR. We can address the remaining know issues regarding configuration/getToken() and the technology versioning in a follow-up PR to keep this structural split manageable.
friendly to reader components. We eventually plan to remove the need for ItemConfig from readers entirely, but that will be a future PR.
|
Realized I didn't rename the function |
|
@phlexbot format |
Format Fixes Applied✅ clang-format fixes pushed (commit 39bb52e) |
weren't using it after the read/write split.
…/phlex into form_read_write_split
|
@phlexbot format |
Format Fixes Applied✅ clang-format fixes pushed (commit 888c71e) |
wwuoneway
left a comment
There was a problem hiding this comment.
thanks for the updates. The changes look fine, and given the scope and complexity already included in this PR, I'm comfortable moving forward. Regarding the CI failures, the clang-tidy is confirmed as unrelated to these specific changes. For code coverage, the cause is unclear, but I agree it shouldn't block this merge.
gemmeren
left a comment
There was a problem hiding this comment.
LGTM, the only comment is that if we consider the use of output_config by reader to be temporary, then we don't really need to rename it. Up to you.
This PR splits
form_interfaceinto two components:form_reader_interfaceandform_writer_interface. This split saves FORM time by avoiding lock contention between read and write processes and prepares for RNTuple support that does not have members for anRNTupleWriterwhen only reading from a container.All of the layers that power the former
form_interfaceare split in this way. A handful of functions shared between reading and writing have been made free functions inform::experimental::detail.