adonis0147 opened a new pull request, #25737:
URL: https://github.com/apache/doris/pull/25737
## Proposed changes
Issue Number: close #25735
The implementation of `S3FileSystem::create_file_impl` doesn't create a file
on S3 file system.
```cpp
Status S3FileSystem::create_file_impl(const Path& file, FileWriterPtr*
writer,
const FileWriterOptions* opts) {
GET_KEY(key, file);
*writer = std::make_unique<S3FileWriter>(
key, std::static_pointer_cast<S3FileSystem>(shared_from_this()),
opts);
return Status::OK();
}
```
However, the implementation of `LocalFileSystem::create_file_impl` does so.
After calling the function `open`, an empty file was created on the local file
system.
```cpp
Status LocalFileSystem::create_file_impl(const Path& file, FileWriterPtr*
writer,
const FileWriterOptions* opts) {
int fd = ::open(file.c_str(), O_TRUNC | O_WRONLY | O_CREAT | O_CLOEXEC,
0666);
if (-1 == fd) {
return Status::IOError("failed to open {}: {}", file.native(),
errno_to_str());
}
*writer = std::make_unique<LocalFileWriter>(
file, fd,
std::static_pointer_cast<LocalFileSystem>(shared_from_this()));
return Status::OK();
}
```
They are inconsistent semantics.
Besides, In `DocumentsWriter::writeSegment`, the function calls
`directory->createOutput((segmentName + ".frq").c_str())` and
`S3FileSystem::create_file_impl` will be called finally. The interface
`createOutput` is expected to create a new and empty file in the current
directory. Therefore, the implementation is wrong and it makes errors occur
when `InvertedIndexColumnWriter` finishes.
```
// Creates a new, empty file in the directory with the given name.
// Returns a stream writing this file.
virtual IndexOutput* createOutput(const char* name) = 0;
```
## Further comments
If this is a relatively large or complex change, kick off the discussion at
[[email protected]](mailto:[email protected]) by explaining why you
chose the solution you did and what alternatives you considered, etc...
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
